From 06139c5efd9905539d227f1a6c7a7e08ea215e91 Mon Sep 17 00:00:00 2001 From: Waleed Date: Thu, 1 Oct 2026 01:07:19 -0700 Subject: [PATCH 1/9] =?UTF-8?q?fix(seo):=20GEO=20hygiene=20=E2=80=94=20ind?= =?UTF-8?q?exing,=20canonical=20URLs,=20llms.txt,=20docs=20redirects,=20li?= =?UTF-8?q?brary=20consolidation=20(#8508)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * fix(seo): GEO hygiene — indexing, canonical URLs, llms.txt, docs redirects, library consolidation - Single source for public site URLs (@sim/utils/site) and a check:site-urls audit against apex links - Derive content canonicals from section basePath + slug instead of hand-written frontmatter - Replace robots.txt disallows for app surfaces with X-Robots-Tag noindex; stop blocking /_next/; drop the inherited homepage canonical and og:url; noindex deployed chats - Generate llms.txt / llms-full.txt from the content registries and comparison data - Docs: redirect removed locale URLs in one hop, fix 404 robots/canonical, git-based lastModified - Merge duplicate library articles with 301s and repoint inbound links; differentiate overlapping ones - Interlink comparisons and library; tighten integration page metadata and JSON-LD - One INTEGRATION_COUNT_LABEL; remove invalid review JSON-LD; raster Organization logo * fix(docs): always register the lastModified plugin so page types don't depend on git history * fix(docs): say the SDK default host redirects to www * improvement(landing): use named type sizes on the integration and comparison pages * fix(landing): match bare competitor names in descriptions; pluralize remaining actions --- apps/docs/app/[[...slug]]/page.tsx | 3 +- apps/docs/app/layout.tsx | 18 +- apps/docs/app/llms.txt/route.ts | 4 +- apps/docs/app/sitemap.ts | 1 + apps/docs/components/navbar/navbar.tsx | 3 +- apps/docs/components/structured-data.tsx | 1 - .../content/docs/api-reference/python.mdx | 10 +- .../content/docs/api-reference/typescript.mdx | 6 +- apps/docs/content/docs/chat/workflows.mdx | 2 +- apps/docs/content/docs/desktop/index.mdx | 2 +- .../docs/content/docs/files/passing-files.mdx | 4 +- .../docs/integrations/hubspot-setup.mdx | 2 +- apps/docs/content/docs/introduction/index.mdx | 2 +- apps/docs/content/docs/platform/costs.mdx | 2 +- .../workflows/deployment/agent-events.mdx | 2 +- .../docs/workflows/deployment/index.mdx | 4 +- apps/docs/lib/redirects.ts | 31 +- apps/docs/lib/urls.ts | 14 +- apps/docs/package.json | 1 + apps/docs/source.config.ts | 54 ++- .../(interfaces)/chat/[identifier]/page.tsx | 43 +- .../chat/components/header/header.tsx | 3 +- .../[workflowId]/[executionId]/page.tsx | 2 +- apps/sim/app/(landing)/blog/[slug]/page.tsx | 2 - apps/sim/app/(landing)/careers/careers.tsx | 5 +- .../(landing)/comparisons/[provider]/page.tsx | 33 +- .../comparison-link-row.tsx | 38 ++ .../components/comparison-link-row/index.ts | 1 + .../comparison-links/comparison-links.tsx | 27 ++ .../components/comparison-links/index.ts | 1 + .../(landing)/comparisons/library-links.ts | 73 ++++ apps/sim/app/(landing)/comparisons/page.tsx | 41 +- .../content-author-page.tsx | 15 +- .../content-index-page/content-index-page.tsx | 22 +- .../content-post-page/content-post-page.tsx | 61 +-- .../content-related-posts.tsx | 54 +++ .../components/content-post-page/index.ts | 1 + .../app/(landing)/components/content-utils.ts | 9 + .../app/(landing)/components/hero/hero.tsx | 7 +- .../home-structured-data.tsx | 36 +- apps/sim/app/(landing)/components/index.ts | 2 +- .../site-structured-data.tsx | 15 +- apps/sim/app/(landing)/demo/demo.tsx | 5 +- apps/sim/app/(landing)/demo/page.tsx | 4 +- .../integrations/(shell)/[slug]/page.tsx | 381 ++++++++++-------- .../(landing)/integrations/(shell)/page.tsx | 24 +- .../sim/app/(landing)/library/[slug]/page.tsx | 8 +- .../pricing-structured-data.tsx | 4 +- apps/sim/app/(landing)/pricing/pricing.tsx | 4 +- .../app/(landing)/privacy/privacy-content.tsx | 7 +- .../solutions/compliance/compliance.tsx | 7 +- .../solutions/engineering/engineering.tsx | 7 +- .../(landing)/solutions/finance/finance.tsx | 4 +- apps/sim/app/(landing)/solutions/hr/hr.tsx | 7 +- apps/sim/app/(landing)/solutions/it/it.tsx | 7 +- .../app/(landing)/solutions/sales/sales.tsx | 7 +- apps/sim/app/f/[token]/public-file-view.tsx | 3 +- apps/sim/app/llms-full.txt/route.ts | 344 ++++++++-------- apps/sim/app/llms.txt/llms.ts | 98 +++++ apps/sim/app/llms.txt/route.ts | 186 +++++---- apps/sim/app/manifest.ts | 4 +- apps/sim/app/robots.ts | 21 +- apps/sim/app/sitemap.ts | 1 - .../content/blog/agent-as-yjs-peer/index.mdx | 1 - apps/sim/content/blog/copilot/index.mdx | 1 - apps/sim/content/blog/emcn/index.mdx | 1 - apps/sim/content/blog/enterprise/index.mdx | 3 +- apps/sim/content/blog/executor/index.mdx | 1 - apps/sim/content/blog/mothership/index.mdx | 5 +- apps/sim/content/blog/multiplayer/index.mdx | 3 +- .../content/blog/secret-provenance/index.mdx | 1 - apps/sim/content/blog/series-a/index.mdx | 3 +- apps/sim/content/blog/v0-5/index.mdx | 5 +- .../content/customers/exp-realty/index.mdx | 1 - apps/sim/content/customers/rivian/index.mdx | 1 - .../index.mdx | 11 +- .../index.mdx | 1 - .../index.mdx | 5 +- .../index.mdx | 15 +- .../content/library/ai-agent-ideas/index.mdx | 1 - .../index.mdx | 33 +- .../library/ai-agent-observability/index.mdx | 11 +- .../index.mdx | 7 +- .../library/ai-agent-vs-chatbot/index.mdx | 13 +- .../index.mdx | 24 +- .../index.mdx | 7 +- .../ai-agents-in-procurement/index.mdx | 3 +- .../library/ai-agents-vs-rpa/index.mdx | 1 - .../index.mdx | 3 +- .../index.mdx | 322 --------------- .../index.mdx | 49 ++- .../index.mdx | 3 +- .../index.mdx | 1 - .../library/apache-2-0-vs-fair-code/index.mdx | 22 +- .../automation-anywhere-alternative/index.mdx | 218 ---------- .../best-ai-agent-builder-2026/index.mdx | 206 ---------- .../index.mdx | 7 +- .../index.mdx | 363 ----------------- .../index.mdx | 5 +- .../index.mdx | 9 +- .../best-ai-agent-platforms-2026/index.mdx | 62 ++- .../index.mdx | 9 +- .../index.mdx | 34 +- .../index.mdx | 31 +- .../index.mdx | 5 +- .../index.mdx | 9 +- .../index.mdx | 7 +- .../index.mdx | 1 - .../index.mdx | 3 +- .../best-ai-agents-for-slack/index.mdx | 72 +++- .../index.mdx | 15 +- .../index.mdx | 21 +- .../best-ai-automation-tools-2026/index.mdx | 7 +- .../index.mdx | 305 -------------- .../best-ai-workflow-builders/index.mdx | 101 ++++- .../index.mdx | 5 +- .../index.mdx | 5 +- .../index.mdx | 11 +- .../index.mdx | 10 +- .../index.mdx | 133 ------ .../index.mdx | 1 - .../index.mdx | 7 +- .../best-zapier-alternatives/index.mdx | 7 +- .../index.mdx | 5 +- .../index.mdx | 3 +- .../library/dify-alternatives/index.mdx | 17 +- .../index.mdx | 1 - .../index.mdx | 1 - .../index.mdx | 3 +- .../how-to-create-an-ai-agent/index.mdx | 3 +- .../index.mdx | 5 +- .../library/langgraph-alternatives/index.mdx | 5 +- .../index.mdx | 3 +- .../content/library/mcp-security/index.mdx | 3 +- .../library/n8n-alternatives/index.mdx | 171 ++------ .../open-source-ai-agent-platforms/index.mdx | 49 ++- .../library/openai-vs-n8n-vs-sim/index.mdx | 34 +- .../index.mdx | 5 +- .../index.mdx | 23 +- .../index.mdx | 7 +- .../index.mdx | 15 +- .../library/top-ai-assistants-2026/index.mdx | 1 - .../what-is-an-agentic-workflow/index.mdx | 3 +- .../index.mdx | 3 +- .../library/what-is-an-mcp-server/index.mdx | 3 +- .../index.mdx | 1 - .../index.mdx | 5 +- .../index.mdx | 7 +- apps/sim/ee/whitelabeling/index.ts | 2 +- apps/sim/ee/whitelabeling/metadata.ts | 58 +-- apps/sim/lib/blog/registry.ts | 2 + .../lib/compare/data/competitors/langchain.ts | 1 + apps/sim/lib/compare/data/competitors/make.ts | 1 + .../data/competitors/microsoft-copilot.ts | 1 + .../data/competitors/openai-agentkit.ts | 1 + .../data/competitors/power-automate.ts | 1 + .../lib/compare/data/competitors/stackai.ts | 1 + apps/sim/lib/compare/data/sim.ts | 37 +- apps/sim/lib/compare/data/types.ts | 8 + apps/sim/lib/content/registry-factory.test.ts | 3 +- apps/sim/lib/content/registry-factory.ts | 38 +- apps/sim/lib/content/schema.ts | 1 - apps/sim/lib/core/utils/urls.ts | 9 +- apps/sim/lib/customers/registry.ts | 3 + apps/sim/lib/help-links.ts | 4 +- apps/sim/lib/landing/constants.ts | 7 + apps/sim/lib/library/registry.ts | 3 + apps/sim/lib/navigation/paths.ts | 27 ++ apps/sim/next.config.ts | 81 ++-- apps/sim/proxy.ts | 8 +- .../cover.jpg | Bin 34893 -> 0 bytes .../automation-anywhere-alternative/cover.jpg | Bin 30429 -> 0 bytes .../best-ai-agent-builder-2026/cover.jpg | Bin 30845 -> 0 bytes .../cover.jpg | Bin 30678 -> 0 bytes .../cover.jpg | Bin 29204 -> 0 bytes .../cover.jpg | Bin 24146 -> 0 bytes bun.lock | 1 + package.json | 1 + packages/utils/package.json | 4 + packages/utils/src/site.ts | 12 + scripts/check-site-urls.ts | 65 +++ 181 files changed, 1883 insertions(+), 2890 deletions(-) create mode 100644 apps/sim/app/(landing)/comparisons/components/comparison-link-row/comparison-link-row.tsx create mode 100644 apps/sim/app/(landing)/comparisons/components/comparison-link-row/index.ts create mode 100644 apps/sim/app/(landing)/comparisons/components/comparison-links/comparison-links.tsx create mode 100644 apps/sim/app/(landing)/comparisons/components/comparison-links/index.ts create mode 100644 apps/sim/app/(landing)/comparisons/library-links.ts create mode 100644 apps/sim/app/(landing)/components/content-post-page/content-related-posts.tsx create mode 100644 apps/sim/app/(landing)/components/content-utils.ts create mode 100644 apps/sim/app/llms.txt/llms.ts delete mode 100644 apps/sim/content/library/ai-native-vs-traditional-workflow-automation/index.mdx delete mode 100644 apps/sim/content/library/automation-anywhere-alternative/index.mdx delete mode 100644 apps/sim/content/library/best-ai-agent-builder-2026/index.mdx delete mode 100644 apps/sim/content/library/best-ai-agent-builders-slack-crm-automation-2026/index.mdx delete mode 100644 apps/sim/content/library/best-ai-workflow-builders-small-teams-2026/index.mdx delete mode 100644 apps/sim/content/library/best-open-source-ai-agent-frameworks/index.mdx create mode 100644 apps/sim/lib/landing/constants.ts delete mode 100644 apps/sim/public/library/ai-native-vs-traditional-workflow-automation/cover.jpg delete mode 100644 apps/sim/public/library/automation-anywhere-alternative/cover.jpg delete mode 100644 apps/sim/public/library/best-ai-agent-builder-2026/cover.jpg delete mode 100644 apps/sim/public/library/best-ai-agent-builders-slack-crm-automation-2026/cover.jpg delete mode 100644 apps/sim/public/library/best-ai-workflow-builders-small-teams-2026/cover.jpg delete mode 100644 apps/sim/public/library/best-open-source-ai-agent-frameworks/cover.jpg create mode 100644 packages/utils/src/site.ts create mode 100644 scripts/check-site-urls.ts diff --git a/apps/docs/app/[[...slug]]/page.tsx b/apps/docs/app/[[...slug]]/page.tsx index 71853679de9..3ae36a2e477 100644 --- a/apps/docs/app/[[...slug]]/page.tsx +++ b/apps/docs/app/[[...slug]]/page.tsx @@ -116,6 +116,7 @@ export default async function Page(props: { params: Promise<{ slug?: string[] }> if (!page) notFound() const data = page.data as unknown as PageData & { + lastModified?: Date _openapi?: { method?: string } getAPIPageProps?: () => ApiPageProps } @@ -246,6 +247,7 @@ export default async function Page(props: { params: Promise<{ slug?: string[] }> title={data.title} description={data.description || ''} url={`${BASE_URL}${page.url}`} + dateModified={data.lastModified?.toISOString()} breadcrumb={breadcrumbs} /> ({ url: `${DOCS_BASE_URL}${page.url}`, + lastModified: 'lastModified' in page.data ? page.data.lastModified : undefined, })) } diff --git a/apps/docs/components/navbar/navbar.tsx b/apps/docs/components/navbar/navbar.tsx index c82d82421b0..ce61485d0ea 100644 --- a/apps/docs/components/navbar/navbar.tsx +++ b/apps/docs/components/navbar/navbar.tsx @@ -6,6 +6,7 @@ import { usePathname } from 'next/navigation' import { SearchTrigger } from '@/components/ui/search-trigger' import { SimWordmark } from '@/components/ui/sim-logo' import { ThemeToggle } from '@/components/ui/theme-toggle' +import { SIM_SITE_URL } from '@/lib/urls' import { cn } from '@/lib/utils' /** @@ -82,7 +83,7 @@ export function Navbar() {
- + Get started
diff --git a/apps/docs/components/structured-data.tsx b/apps/docs/components/structured-data.tsx index af105b5c834..a2764635083 100644 --- a/apps/docs/components/structured-data.tsx +++ b/apps/docs/components/structured-data.tsx @@ -25,7 +25,6 @@ export function StructuredData({ headline: title, description: description, url: url, - ...(dateModified && { datePublished: dateModified }), ...(dateModified && { dateModified }), author: { '@type': 'Organization', diff --git a/apps/docs/content/docs/api-reference/python.mdx b/apps/docs/content/docs/api-reference/python.mdx index fa143895ae6..5ab91aac6bf 100644 --- a/apps/docs/content/docs/api-reference/python.mdx +++ b/apps/docs/content/docs/api-reference/python.mdx @@ -30,7 +30,7 @@ from simstudio import SimStudioClient # Initialize the client client = SimStudioClient( api_key="your-api-key-here", - base_url="https://sim.ai" # optional, defaults to https://sim.ai + base_url="https://www.sim.ai" # optional; the default https://sim.ai redirects to this host ) # Execute a workflow @@ -53,7 +53,7 @@ SimStudioClient(api_key: str, base_url: str = "https://sim.ai") **Parameters:** - `api_key` (str): Your Sim API key -- `base_url` (str, optional): Base URL for the Sim API +- `base_url` (str, optional): Base URL for the Sim API (defaults to `https://sim.ai`, which redirects to `https://www.sim.ai`; set the `www` host explicitly) #### Methods @@ -678,7 +678,7 @@ def stream_workflow(): def generate(): response = requests.post( - 'https://sim.ai/api/v2/workflows/WORKFLOW_ID/execute', + 'https://www.sim.ai/api/v2/workflows/WORKFLOW_ID/execute', headers={ 'Content-Type': 'application/json', 'X-API-Key': os.getenv('SIM_API_KEY') @@ -732,7 +732,7 @@ Configure the client using environment variables: # Development configuration client = SimStudioClient( api_key=os.getenv("SIM_API_KEY"), - base_url=os.getenv("SIM_BASE_URL", "https://sim.ai") + base_url=os.getenv("SIM_BASE_URL", "https://www.sim.ai") ) ``` @@ -748,7 +748,7 @@ Configure the client using environment variables: client = SimStudioClient( api_key=api_key, - base_url=os.getenv("SIM_BASE_URL", "https://sim.ai") + base_url=os.getenv("SIM_BASE_URL", "https://www.sim.ai") ) ``` diff --git a/apps/docs/content/docs/api-reference/typescript.mdx b/apps/docs/content/docs/api-reference/typescript.mdx index a65ea5d96ec..88068da4506 100644 --- a/apps/docs/content/docs/api-reference/typescript.mdx +++ b/apps/docs/content/docs/api-reference/typescript.mdx @@ -44,7 +44,7 @@ import { SimStudioClient } from 'simstudio-ts-sdk'; // Initialize the client const client = new SimStudioClient({ apiKey: 'your-api-key-here', - baseUrl: 'https://sim.ai' // optional, defaults to https://sim.ai + baseUrl: 'https://www.sim.ai' // optional; the default https://sim.ai redirects to this host }); // Execute a workflow @@ -68,7 +68,7 @@ new SimStudioClient(config: SimStudioConfig) **Configuration:** - `config.apiKey` (string): Your Sim API key -- `config.baseUrl` (string, optional): Base URL for the Sim API (defaults to `https://sim.ai`) +- `config.baseUrl` (string, optional): Base URL for the Sim API (defaults to `https://sim.ai`, which redirects to `https://www.sim.ai`; set the `www` host explicitly) #### Methods @@ -492,7 +492,7 @@ Configure the client using environment variables: const client = new SimStudioClient({ apiKey, - baseUrl: process.env.SIM_BASE_URL || 'https://sim.ai' + baseUrl: process.env.SIM_BASE_URL || 'https://www.sim.ai' }); ``` diff --git a/apps/docs/content/docs/chat/workflows.mdx b/apps/docs/content/docs/chat/workflows.mdx index cdfa33e0764..ec95fa219f2 100644 --- a/apps/docs/content/docs/chat/workflows.mdx +++ b/apps/docs/content/docs/chat/workflows.mdx @@ -69,7 +69,7 @@ Sim can deploy a workflow as any of the three deployment types: | Deployment type | What it creates | |----------------|----------------| -| **API** | A REST endpoint at `https://sim.ai/api/v2/workflows/{id}/execute` | +| **API** | A REST endpoint at `https://www.sim.ai/api/v2/workflows/{id}/execute` | | **Chat** | A hosted conversational interface with a shareable URL | | **MCP tool** | An MCP server that exposes the workflow as a tool | diff --git a/apps/docs/content/docs/desktop/index.mdx b/apps/docs/content/docs/desktop/index.mdx index 925c58938e0..134bc8574ce 100644 --- a/apps/docs/content/docs/desktop/index.mdx +++ b/apps/docs/content/docs/desktop/index.mdx @@ -18,7 +18,7 @@ Sim Desktop is the macOS app for your Sim workspace. Everything the web app does ## Download -**[Download Sim Desktop for macOS](https://sim.ai/api/desktop/update/download)** +**[Download Sim Desktop for macOS](https://www.sim.ai/api/desktop/update/download)** One universal build runs natively on both Apple Silicon and Intel Macs. It is signed and notarized by Sim, so Gatekeeper accepts it with no override. diff --git a/apps/docs/content/docs/files/passing-files.mdx b/apps/docs/content/docs/files/passing-files.mdx index 846566e4287..22d22c94df5 100644 --- a/apps/docs/content/docs/files/passing-files.mdx +++ b/apps/docs/content/docs/files/passing-files.mdx @@ -87,7 +87,7 @@ When calling a workflow via API that expects file input, include files in your r ```bash - curl -X POST "https://sim.ai/api/v2/workflows/YOUR_WORKFLOW_ID/execute" \ + curl -X POST "https://www.sim.ai/api/v2/workflows/YOUR_WORKFLOW_ID/execute" \ -H "Content-Type: application/json" \ -H "x-api-key: YOUR_API_KEY" \ -d '{ @@ -101,7 +101,7 @@ When calling a workflow via API that expects file input, include files in your r ```bash - curl -X POST "https://sim.ai/api/v2/workflows/YOUR_WORKFLOW_ID/execute" \ + curl -X POST "https://www.sim.ai/api/v2/workflows/YOUR_WORKFLOW_ID/execute" \ -H "Content-Type: application/json" \ -H "x-api-key: YOUR_API_KEY" \ -d '{ diff --git a/apps/docs/content/docs/integrations/hubspot-setup.mdx b/apps/docs/content/docs/integrations/hubspot-setup.mdx index 1ecd338525a..2253303131d 100644 --- a/apps/docs/content/docs/integrations/hubspot-setup.mdx +++ b/apps/docs/content/docs/integrations/hubspot-setup.mdx @@ -18,7 +18,7 @@ This guide covers installing the integration, connecting a HubSpot account, conf You need: -- A [Sim](https://sim.ai) account and a workspace where you have **Write** or **Admin** permission. +- A [Sim](https://www.sim.ai) account and a workspace where you have **Write** or **Admin** permission. - A HubSpot account. To grant the requested scopes, your HubSpot user needs permission to install apps (typically a super admin). ## Install the app and connect HubSpot diff --git a/apps/docs/content/docs/introduction/index.mdx b/apps/docs/content/docs/introduction/index.mdx index 63f54b5ba6d..f099f381c63 100644 --- a/apps/docs/content/docs/introduction/index.mdx +++ b/apps/docs/content/docs/introduction/index.mdx @@ -76,7 +76,7 @@ For anything not built in, [MCP support](/agents/mcp) connects any external serv ## Deployment options -- **Cloud-hosted.** Launch immediately at [sim.ai](https://sim.ai) with managed infrastructure, scaling, and observability. +- **Cloud-hosted.** Launch immediately at [sim.ai](https://www.sim.ai) with managed infrastructure, scaling, and observability. - **Self-hosted.** Deploy on your own infrastructure with Docker Compose or Kubernetes, with support for local models. ## Next steps diff --git a/apps/docs/content/docs/platform/costs.mdx b/apps/docs/content/docs/platform/costs.mdx index 8194da17c24..583f1aceb73 100644 --- a/apps/docs/content/docs/platform/costs.mdx +++ b/apps/docs/content/docs/platform/costs.mdx @@ -510,7 +510,7 @@ Pro and Team plan users can buy additional credits at any time in **Settings → ## Next Steps -- Review your current usage in [Settings → Subscription](https://sim.ai/settings/subscription) +- Review your current usage in [Settings → Subscription](https://www.sim.ai/settings/subscription) - Learn about [Logging](/logs-debugging/logging) to track run details - Explore the [External API](/api-reference/getting-started) for programmatic cost monitoring - Check out [workflow optimization techniques](/workflows#blocks) to reduce costs diff --git a/apps/docs/content/docs/workflows/deployment/agent-events.mdx b/apps/docs/content/docs/workflows/deployment/agent-events.mdx index d1aa8a7a273..56b222dbd65 100644 --- a/apps/docs/content/docs/workflows/deployment/agent-events.mdx +++ b/apps/docs/content/docs/workflows/deployment/agent-events.mdx @@ -41,7 +41,7 @@ On the workflow API, setting `includeThinking` or `includeToolCalls` **without** ### Workflow API ```bash -curl -N https://sim.ai/api/v2/workflows/{id}/execute \ +curl -N https://www.sim.ai/api/v2/workflows/{id}/execute \ -H "X-API-Key: $SIM_API_KEY" \ -H "Content-Type: application/json" \ -H "X-Sim-Stream-Protocol: agent-events-v1" \ diff --git a/apps/docs/content/docs/workflows/deployment/index.mdx b/apps/docs/content/docs/workflows/deployment/index.mdx index 62a9f20eb03..f46e7c43fb8 100644 --- a/apps/docs/content/docs/workflows/deployment/index.mdx +++ b/apps/docs/content/docs/workflows/deployment/index.mdx @@ -41,13 +41,13 @@ Every surface runs the same live snapshot. You manage them from the tabs of the The most common surface is the **API**. Once deployed, your workflow answers at: ``` -POST https://sim.ai/api/v2/workflows/{workflow-id}/execute +POST https://www.sim.ai/api/v2/workflows/{workflow-id}/execute ``` Send the workflow's [Input Format](/workflows/triggers/start) as the request body's `input` value. The response includes workflow output and execution metadata; see [API deployment](/workflows/deployment/api) for response modes. ```bash -curl -X POST https://sim.ai/api/v2/workflows/{workflow-id}/execute \ +curl -X POST https://www.sim.ai/api/v2/workflows/{workflow-id}/execute \ -H "X-API-Key: $SIM_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "input": { "message": "Refund request from customer #4821" } }' diff --git a/apps/docs/lib/redirects.ts b/apps/docs/lib/redirects.ts index a92193fcc9f..f79577ebc80 100644 --- a/apps/docs/lib/redirects.ts +++ b/apps/docs/lib/redirects.ts @@ -5,21 +5,21 @@ import type { NextConfig } from 'next' type DocsRedirect = Awaited>>[number] /** - * Every redirect the docs site serves, in match order — Next applies the first + * Every unprefixed path redirect, in match order — Next applies the first * matching rule. * * This lives outside `next.config.ts` so it can be read without evaluating that * module. `createMDX()` runs at import time and bundles `source.config.ts` * against `process.cwd()`, so importing the config from the root Vitest project * fails with `The entry point "source.config.ts" cannot be marked as external`. - * `scripts/openapi/docs-redirects.test.ts` reads this array directly to keep the + * `scripts/openapi/docs-redirects.test.ts` reads `DOCS_REDIRECTS` directly to keep the * `/api-reference/` rules honest against the specs. * * The whole table lives here rather than only the `/api-reference/` block: Next * applies the first matching rule, so splitting one ordered list across two * modules would make match order an emergent property of two files. */ -export const DOCS_REDIRECTS: DocsRedirect[] = [ +const PATH_REDIRECTS: DocsRedirect[] = [ ...Object.entries(integrationNavigation.redirects).map(([from, to]) => ({ source: `/integrations/${from}`, destination: `/integrations/${to}`, @@ -390,3 +390,28 @@ export const DOCS_REDIRECTS: DocsRedirect[] = [ permanent: false, }, ] + +/** + * Locale prefixes the docs served before translations were removed in #7247 + * (`lib/i18n.ts` declared `en`, `es`, `fr`, `de`, `ja`, `zh`; `en` was hidden). + * Search engines still hold these URLs, so each one 308s to its English page. + */ +const RETIRED_LOCALE_PREFIX = '/:lang(en|es|fr|de|ja|zh)' + +/** + * Every redirect the docs site serves, in match order — Next applies the first + * matching rule and does not chain them internally. + * + * Each path redirect is repeated under the retired locale prefix with the same + * destination, so `/fr/tools/x` lands on `/integrations/x` in one hop rather + * than stripping the locale and redirecting again. The trailing catch-all + * strips the prefix from every other path, which already resolves. + */ +export const DOCS_REDIRECTS: DocsRedirect[] = [ + ...PATH_REDIRECTS, + ...PATH_REDIRECTS.map((rule) => ({ + ...rule, + source: rule.source === '/' ? RETIRED_LOCALE_PREFIX : `${RETIRED_LOCALE_PREFIX}${rule.source}`, + })), + { source: `${RETIRED_LOCALE_PREFIX}/:path*`, destination: '/:path*', permanent: true }, +] diff --git a/apps/docs/lib/urls.ts b/apps/docs/lib/urls.ts index 8001a5d55ed..63fe9d0c772 100644 --- a/apps/docs/lib/urls.ts +++ b/apps/docs/lib/urls.ts @@ -1,9 +1,9 @@ -export const DOCS_BASE_URL = process.env.NEXT_PUBLIC_DOCS_URL ?? 'https://docs.sim.ai' +import { SIM_DOCS_URL, SIM_SITE_URL } from '@sim/utils/site' + +export const DOCS_BASE_URL = process.env.NEXT_PUBLIC_DOCS_URL ?? SIM_DOCS_URL + /** - * The public marketing site's fixed canonical origin — not `NEXT_PUBLIC_APP_URL`. - * That env var reflects wherever *this* deployment (self-hosted or otherwise) - * happens to run, but the footer's marketing links (`/blog`, `/enterprise`, - * `/models`, `/terms`, `/privacy`, …) only ever exist on sim.ai itself, so - * they must stay hardcoded to it regardless of where docs is hosted. + * The marketing site's canonical origin, never `NEXT_PUBLIC_APP_URL`: marketing + * links only exist on sim.ai, wherever docs is hosted. */ -export const SIM_SITE_URL = 'https://sim.ai' +export { SIM_SITE_URL } diff --git a/apps/docs/package.json b/apps/docs/package.json index 5e9b39d8a81..5396bf2cdf9 100644 --- a/apps/docs/package.json +++ b/apps/docs/package.json @@ -20,6 +20,7 @@ "dependencies": { "@sim/db": "workspace:*", "@sim/emcn": "workspace:*", + "@sim/utils": "workspace:*", "@sim/workflow-renderer": "workspace:*", "@xyflow/react": "12.11.3", "class-variance-authority": "^0.7.1", diff --git a/apps/docs/source.config.ts b/apps/docs/source.config.ts index 231ee6ff820..aecb8280293 100644 --- a/apps/docs/source.config.ts +++ b/apps/docs/source.config.ts @@ -1,9 +1,14 @@ +import { execFileSync } from 'node:child_process' +import path from 'node:path' import { defineConfig, defineDocs, frontmatterSchema } from 'fumadocs-mdx/config' +import lastModified from 'fumadocs-mdx/plugins/last-modified' import { curlJsonBodyGrammar } from './lib/shiki-curl-json' import { simShikiOptions } from './lib/shiki-theme' +const DOCS_DIR = 'content/docs' + export const docs = defineDocs({ - dir: 'content/docs', + dir: DOCS_DIR, docs: { schema: frontmatterSchema, postprocess: { @@ -12,7 +17,54 @@ export const docs = defineDocs({ }, }) +/** + * Last-commit author date of every file under {@link DOCS_DIR}, keyed by absolute path, from one + * `git log` pass instead of the plugin's per-file `git log -1` spawn (~500 files). `undefined` + * when git is missing or the clone is shallow: a shallow clone attributes every file untouched + * since its boundary commit to that commit, so pages carry no `lastModified` rather than a + * fabricated one. + */ +function readGitLastModified(): Map | undefined { + const git = (args: string[]) => + execFileSync('git', args, { + encoding: 'utf8', + maxBuffer: 64 * 1024 * 1024, + stdio: ['ignore', 'pipe', 'ignore'], + }).trim() + try { + if (git(['rev-parse', '--is-shallow-repository']) !== 'false') return undefined + const root = git(['rev-parse', '--show-toplevel']) + const dates = new Map() + const log = git([ + '-c', + 'core.quotepath=off', + 'log', + '--format=%x00%aI', + '--name-only', + '--', + DOCS_DIR, + ]) + let date: Date | undefined + for (const line of log.split('\n')) { + if (line.startsWith('\0')) date = new Date(line.slice(1)) + else if (line && date) { + const file = path.join(root, line) + if (!dates.has(file)) dates.set(file, date) + } + } + return dates + } catch { + return undefined + } +} + +const gitLastModified = readGitLastModified() + export default defineConfig({ + /** Always registered so the generated page types are the same with or without git history. */ + plugins: [ + lastModified({ versionControl: async (file) => gitLastModified?.get(path.resolve(file)) }), + ], mdxOptions: { /** * Shiki defaults to `github-light` / `github-dark`, whose blues and purples appear nowhere diff --git a/apps/sim/app/(interfaces)/chat/[identifier]/page.tsx b/apps/sim/app/(interfaces)/chat/[identifier]/page.tsx index f55345cbbbc..e159709caf2 100644 --- a/apps/sim/app/(interfaces)/chat/[identifier]/page.tsx +++ b/apps/sim/app/(interfaces)/chat/[identifier]/page.tsx @@ -9,18 +9,16 @@ import { OfficeEmbedInit } from '@/app/(interfaces)/chat/[identifier]/office-emb const logger = createLogger('ChatMetadata') +const NOINDEX: Metadata['robots'] = { index: false, follow: false } + /** - * Only fully public, active deployments are indexable. Auth-gated (password, - * email, SSO) and inactive/nonexistent chats are noindexed at the page level - * so Google never indexes an auth wall — narrower than blocking `/chat/` - * entirely in robots.ts, which would also hide genuinely public deployments. + * Deployed chats are never indexed: they are thin, client-rendered pages built + * by users, not Sim content. A public, active chat gets its own title and + * description for link previews; auth-gated, inactive, and unknown chats get a + * generic title so nothing behind the gate leaks. * - * Errors from the lookup fail toward noindex rather than throwing: unlike - * the identical query in app/api/chat/[identifier]/route.ts (which must - * surface failures to the caller), a metadata resolution error has no - * error.tsx boundary in this route to catch it — throwing here would take - * the whole page down instead of just skipping indexability, and "can't - * confirm this is safe to index" should default to not indexing it anyway. + * A lookup error falls back to the generic title rather than throwing: this + * route has no error.tsx boundary, so a throw would take the whole page down. */ export async function generateMetadata({ params, @@ -29,15 +27,29 @@ export async function generateMetadata({ }): Promise { const { identifier } = await params - let isIndexable = false try { const [deployment] = await db - .select({ authType: chat.authType, isActive: chat.isActive }) + .select({ + title: chat.title, + description: chat.description, + authType: chat.authType, + isActive: chat.isActive, + }) .from(chat) .where(and(eq(chat.identifier, identifier), isNull(chat.archivedAt))) .limit(1) - isIndexable = Boolean(deployment?.isActive && deployment.authType === 'public') + if (deployment?.isActive && deployment.authType === 'public') { + const { title } = deployment + const description = deployment.description || undefined + return { + title, + description, + openGraph: { title, description, type: 'website' }, + twitter: { card: 'summary', title, description }, + robots: NOINDEX, + } + } } catch (error) { logger.error('Failed to resolve chat deployment for metadata', { identifier, @@ -45,10 +57,7 @@ export async function generateMetadata({ }) } - return { - title: 'Chat', - ...(!isIndexable && { robots: { index: false, follow: false } }), - } + return { title: 'Chat', robots: NOINDEX } } export const dynamic = 'force-dynamic' diff --git a/apps/sim/app/(interfaces)/chat/components/header/header.tsx b/apps/sim/app/(interfaces)/chat/components/header/header.tsx index fd96bb39cd1..4afbe77ae14 100644 --- a/apps/sim/app/(interfaces)/chat/components/header/header.tsx +++ b/apps/sim/app/(interfaces)/chat/components/header/header.tsx @@ -4,6 +4,7 @@ import { SimWordmark } from '@sim/emcn' import Image from 'next/image' import Link from 'next/link' import { GithubIcon } from '@/components/icons' +import { SITE_URL } from '@/lib/core/utils/urls' import { useBrandConfig } from '@/ee/whitelabeling' interface ChatHeaderProps { @@ -61,7 +62,7 @@ export function ChatHeader({ chatConfig, starCount }: ChatHeaderProps) { {/* Only show Sim logo if no custom branding is set */} ) } diff --git a/apps/sim/app/(landing)/careers/careers.tsx b/apps/sim/app/(landing)/careers/careers.tsx index a9023f10f97..14efa03efbb 100644 --- a/apps/sim/app/(landing)/careers/careers.tsx +++ b/apps/sim/app/(landing)/careers/careers.tsx @@ -1,6 +1,7 @@ import { Suspense } from 'react' import type { SearchParams } from 'nuqs/server' import { getAshbyJobs } from '@/lib/ashby/jobs' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { filterPostings, groupByDepartment, @@ -50,8 +51,8 @@ export default async function Careers({ searchParams }: CareersProps) {

Careers at Sim, the open-source AI workspace where teams build, deploy, and manage AI agents. Sim is hiring engineers, designers, and go-to-market builders to help teams - automate real work across hundreds of integrations and every major LLM — visually, - conversationally, or with code. + automate real work across {INTEGRATION_COUNT_LABEL} integrations and every major LLM — + visually, conversationally, or with code.

Is Sim better than {competitor.name}?

@@ -267,7 +270,7 @@ export default async function ComparisonProviderPage({

Sim vs {competitor.name}: feature-by-feature comparison

@@ -338,7 +341,7 @@ export default async function ComparisonProviderPage({

Bottom line

@@ -354,10 +357,30 @@ export default async function ComparisonProviderPage({
+ {relatedPosts.length > 0 ? ( + <> +
+ +
+ +
+
+ + ) : null} +

Frequently asked questions

diff --git a/apps/sim/app/(landing)/comparisons/components/comparison-link-row/comparison-link-row.tsx b/apps/sim/app/(landing)/comparisons/components/comparison-link-row/comparison-link-row.tsx new file mode 100644 index 00000000000..4f7728df801 --- /dev/null +++ b/apps/sim/app/(landing)/comparisons/components/comparison-link-row/comparison-link-row.tsx @@ -0,0 +1,38 @@ +import Link from 'next/link' +import type { CompetitorProfile } from '@/lib/compare/data' +import { BrandIconTile } from '@/app/(landing)/comparisons/components/brand-icon-tile' +import { ChevronArrow } from '@/app/(landing)/components/chevron-arrow' + +interface ComparisonLinkRowProps { + competitor: CompetitorProfile +} + +/** One "Sim vs {Competitor}" link row: brand tile, title, one-liner, and chevron. */ +export function ComparisonLinkRow({ competitor }: ComparisonLinkRowProps) { + const Icon = competitor.brand?.icon + return ( + + {Icon ? ( + + ) : null} +
+

+ Sim vs {competitor.name} +

+

+ {competitor.oneLiner} +

+
+ + + ) +} diff --git a/apps/sim/app/(landing)/comparisons/components/comparison-link-row/index.ts b/apps/sim/app/(landing)/comparisons/components/comparison-link-row/index.ts new file mode 100644 index 00000000000..ce1add34e24 --- /dev/null +++ b/apps/sim/app/(landing)/comparisons/components/comparison-link-row/index.ts @@ -0,0 +1 @@ +export { ComparisonLinkRow } from './comparison-link-row' diff --git a/apps/sim/app/(landing)/comparisons/components/comparison-links/comparison-links.tsx b/apps/sim/app/(landing)/comparisons/components/comparison-links/comparison-links.tsx new file mode 100644 index 00000000000..07f97757124 --- /dev/null +++ b/apps/sim/app/(landing)/comparisons/components/comparison-links/comparison-links.tsx @@ -0,0 +1,27 @@ +import type { CompetitorProfile } from '@/lib/compare/data' +import { ComparisonLinkRow } from '@/app/(landing)/comparisons/components/comparison-link-row' + +interface ComparisonLinksProps { + competitors: CompetitorProfile[] +} + +/** "Compare Sim" section linking a library article to the comparison pages it relates to. */ +export function ComparisonLinks({ competitors }: ComparisonLinksProps) { + return ( +
+ +
    + {competitors.map((competitor) => ( +
  • + +
    • + ))} +
+
+ ) +} diff --git a/apps/sim/app/(landing)/comparisons/components/comparison-links/index.ts b/apps/sim/app/(landing)/comparisons/components/comparison-links/index.ts new file mode 100644 index 00000000000..51100d40535 --- /dev/null +++ b/apps/sim/app/(landing)/comparisons/components/comparison-links/index.ts @@ -0,0 +1 @@ +export { ComparisonLinks } from './comparison-links' diff --git a/apps/sim/app/(landing)/comparisons/library-links.ts b/apps/sim/app/(landing)/comparisons/library-links.ts new file mode 100644 index 00000000000..990cb23dd0b --- /dev/null +++ b/apps/sim/app/(landing)/comparisons/library-links.ts @@ -0,0 +1,73 @@ +import { escapeRegExp } from '@sim/utils/string' +import type { CompetitorProfile } from '@/lib/compare/data' +import type { ContentMeta } from '@/lib/content/schema' +import { getAllPostMeta } from '@/lib/library/registry' +import { ALL_COMPETITORS } from '@/app/(landing)/comparisons/utils' + +/** Most links either side renders, so the section stays a short "read next" list. */ +const MAX_LINKS = 3 + +/** A competitor named in an article's title or tags is its subject; one named only in the description is a mention. */ +const SUBJECT_SCORE = 2 +const MENTION_SCORE = 1 + +interface MentionPatterns { + /** Matches Title Case titles and tags, where a bare common-word name ("Make") is ambiguous. */ + subject: RegExp + /** Matches sentence-case descriptions, where the bare name is unambiguous. */ + mention: RegExp +} + +/** Whole-word, case-sensitive match on any of `phrases`. */ +function wordPattern(phrases: string[]): RegExp { + return new RegExp(`\\b(?:${phrases.map(escapeRegExp).join('|')})\\b`) +} + +function buildMentionPatterns(competitor: CompetitorProfile): MentionPatterns { + const aliases = competitor.mentions ?? [competitor.name] + return { + subject: wordPattern(aliases), + mention: wordPattern([competitor.name, ...aliases]), + } +} + +const COMPETITOR_PATTERNS = ALL_COMPETITORS.map((competitor) => ({ + competitor, + patterns: buildMentionPatterns(competitor), +})) + +function scoreRelevance(post: ContentMeta, { subject, mention }: MentionPatterns): number { + if (subject.test(post.title) || post.tags.some((tag) => subject.test(tag))) return SUBJECT_SCORE + return mention.test(post.description) ? MENTION_SCORE : 0 +} + +/** The highest-scoring items with a positive score, best first; ties keep input order. */ +function topByScore(items: T[], score: (item: T) => number): T[] { + return items + .map((item) => ({ item, score: score(item) })) + .filter(({ score }) => score > 0) + .sort((a, b) => b.score - a.score) + .slice(0, MAX_LINKS) + .map(({ item }) => item) +} + +/** + * Library articles about or naming `competitor`, articles where it is the + * subject first, then newest first. + */ +export async function getLibraryPostsForCompetitor( + competitor: CompetitorProfile +): Promise { + const patterns = buildMentionPatterns(competitor) + return topByScore(await getAllPostMeta(), (post) => scoreRelevance(post, patterns)) +} + +/** + * Comparison pages for the competitors a library article is about or names, + * subjects first, then in {@link ALL_COMPETITORS} order. + */ +export function getComparisonsForPost(post: ContentMeta): CompetitorProfile[] { + return topByScore(COMPETITOR_PATTERNS, ({ patterns }) => scoreRelevance(post, patterns)).map( + ({ competitor }) => competitor + ) +} diff --git a/apps/sim/app/(landing)/comparisons/page.tsx b/apps/sim/app/(landing)/comparisons/page.tsx index b7b789a05b7..5f62a634c09 100644 --- a/apps/sim/app/(landing)/comparisons/page.tsx +++ b/apps/sim/app/(landing)/comparisons/page.tsx @@ -1,12 +1,10 @@ import { cn } from '@sim/emcn' import type { Metadata } from 'next' -import Link from 'next/link' import { simProfile } from '@/lib/compare/data' import { SITE_URL } from '@/lib/core/utils/urls' import { buildLandingMetadata } from '@/lib/landing/seo' -import { BrandIconTile } from '@/app/(landing)/comparisons/components/brand-icon-tile' +import { ComparisonLinkRow } from '@/app/(landing)/comparisons/components/comparison-link-row' import { ALL_COMPETITORS, ensurePeriod, lowercaseFirst } from '@/app/(landing)/comparisons/utils' -import { ChevronArrow } from '@/app/(landing)/components/chevron-arrow' import { JsonLd } from '@/app/(landing)/components/json-ld' import { LandingFAQ } from '@/app/(landing)/components/landing-faq' import { LANDING_CONTENT_WIDTH, LANDING_GUTTER } from '@/app/(landing)/components/landing-layout' @@ -141,37 +139,12 @@ export default function ComparisonHubPage() { All comparisons
- {ALL_COMPETITORS.map((competitor) => { - const Icon = competitor.brand?.icon - return ( -
- - {Icon ? ( - - ) : null} -
-

- Sim vs {competitor.name} -

-

- {competitor.oneLiner} -

-
- - -
-
- ) - })} + {ALL_COMPETITORS.map((competitor) => ( +
+ +
+
+ ))}
diff --git a/apps/sim/app/(landing)/components/content-author-page/content-author-page.tsx b/apps/sim/app/(landing)/components/content-author-page/content-author-page.tsx index 4d77b215514..79521d7878e 100644 --- a/apps/sim/app/(landing)/components/content-author-page/content-author-page.tsx +++ b/apps/sim/app/(landing)/components/content-author-page/content-author-page.tsx @@ -2,6 +2,7 @@ import Image from 'next/image' import Link from 'next/link' import type { ContentMeta } from '@/lib/content/schema' import { BackLink } from '@/app/(landing)/components/back-link' +import { formatPostDate } from '@/app/(landing)/components/content-utils' import { JsonLd } from '@/app/(landing)/components/json-ld' interface ContentAuthorPageProps { @@ -68,22 +69,12 @@ export function ContentAuthorPage({ className='group flex items-start gap-6 p-6 transition-colors hover:bg-[var(--surface-hover)] md:items-center' > - {new Date(p.date).toLocaleDateString('en-US', { - month: 'short', - day: 'numeric', - year: 'numeric', - timeZone: 'UTC', - })} + {formatPostDate(p.date)}
- {new Date(p.date).toLocaleDateString('en-US', { - month: 'short', - day: 'numeric', - year: 'numeric', - timeZone: 'UTC', - })} + {formatPostDate(p.date)}

{p.title} diff --git a/apps/sim/app/(landing)/components/content-index-page/content-index-page.tsx b/apps/sim/app/(landing)/components/content-index-page/content-index-page.tsx index b2138a8d2ef..7fd2e234452 100644 --- a/apps/sim/app/(landing)/components/content-index-page/content-index-page.tsx +++ b/apps/sim/app/(landing)/components/content-index-page/content-index-page.tsx @@ -3,6 +3,7 @@ import Image from 'next/image' import Link from 'next/link' import { paginateContentPosts } from '@/lib/content/index-list' import type { ContentMeta } from '@/lib/content/schema' +import { formatPostDate } from '@/app/(landing)/components/content-utils' import { JsonLd } from '@/app/(landing)/components/json-ld' interface ContentIndexPageProps { @@ -81,12 +82,7 @@ export function ContentIndexPage({

- {new Date(p.date).toLocaleDateString('en-US', { - month: 'short', - day: 'numeric', - year: 'numeric', - timeZone: 'UTC', - })} + {formatPostDate(p.date)}

{p.title} @@ -110,22 +106,12 @@ export function ContentIndexPage({ className='group flex items-start gap-6 p-6 transition-colors hover:bg-[var(--surface-hover)] md:items-center' > - {new Date(p.date).toLocaleDateString('en-US', { - month: 'short', - day: 'numeric', - year: 'numeric', - timeZone: 'UTC', - })} + {formatPostDate(p.date)}
- {new Date(p.date).toLocaleDateString('en-US', { - month: 'short', - day: 'numeric', - year: 'numeric', - timeZone: 'UTC', - })} + {formatPostDate(p.date)}

{p.title} diff --git a/apps/sim/app/(landing)/components/content-post-page/content-post-page.tsx b/apps/sim/app/(landing)/components/content-post-page/content-post-page.tsx index 5223621a50b..5929f4d7c42 100644 --- a/apps/sim/app/(landing)/components/content-post-page/content-post-page.tsx +++ b/apps/sim/app/(landing)/components/content-post-page/content-post-page.tsx @@ -1,22 +1,15 @@ +import type { ReactNode } from 'react' import { Avatar, AvatarFallback, AvatarImage } from '@sim/emcn' import Image from 'next/image' import Link from 'next/link' import { FAQ } from '@/lib/content/faq' import type { ContentMeta, ContentPost } from '@/lib/content/schema' import { BackLink } from '@/app/(landing)/components/back-link' +import { ContentRelatedPosts } from '@/app/(landing)/components/content-post-page/content-related-posts' +import { formatPostDate } from '@/app/(landing)/components/content-utils' import { JsonLd } from '@/app/(landing)/components/json-ld' import { ShareButton } from '@/app/(landing)/components/share-button' -/** Renders an ISO date as "Jul 1, 2026". Pinned to UTC so the day matches the frontmatter date in every reader's timezone. */ -function formatDate(iso: string): string { - return new Date(iso).toLocaleDateString('en-US', { - month: 'short', - day: 'numeric', - year: 'numeric', - timeZone: 'UTC', - }) -} - interface ContentPostPageProps { /** Route base path, e.g. `/blog` or `/library`. */ basePath: string @@ -25,7 +18,8 @@ interface ContentPostPageProps { post: ContentPost related: ContentMeta[] graphJsonLd: Record - shareUrl: string + /** Section-specific content rendered below the article body, above related posts. */ + afterArticle?: ReactNode } /** @@ -39,7 +33,7 @@ export function ContentPostPage({ post, related, graphJsonLd, - shareUrl, + afterArticle, }: ContentPostPageProps) { const Article = post.Content const modifiedIso = post.updated ?? post.date @@ -92,7 +86,7 @@ export function ContentPostPage({ dateTime={post.date} itemProp='datePublished' > - {formatDate(post.date)} + {formatPostDate(post.date)} {showUpdated ? ( <> @@ -104,7 +98,7 @@ export function ContentPostPage({ dateTime={modifiedIso} itemProp='dateModified' > - Updated {formatDate(modifiedIso)} + Updated {formatPostDate(modifiedIso)} ) : ( @@ -134,7 +128,7 @@ export function ContentPostPage({ ))}

- +

@@ -152,41 +146,14 @@ export function ContentPostPage({
+ {afterArticle ? ( +
{afterArticle}
+ ) : null} + {related.length > 0 && ( <>
- + )}
diff --git a/apps/sim/app/(landing)/components/content-post-page/content-related-posts.tsx b/apps/sim/app/(landing)/components/content-post-page/content-related-posts.tsx new file mode 100644 index 00000000000..4aa2dbc6dc3 --- /dev/null +++ b/apps/sim/app/(landing)/components/content-post-page/content-related-posts.tsx @@ -0,0 +1,54 @@ +import Image from 'next/image' +import Link from 'next/link' +import type { ContentMeta } from '@/lib/content/schema' +import { formatPostDate } from '@/app/(landing)/components/content-utils' + +interface ContentRelatedPostsProps { + /** Route base path of the posts' section, e.g. `/blog` or `/library`. */ + basePath: string + posts: ContentMeta[] + /** Accessible name for the nav landmark. */ + label?: string +} + +/** Row of post cards (cover, date, title, description) linking to other posts in a content section. */ +export function ContentRelatedPosts({ + basePath, + posts, + label = 'Related posts', +}: ContentRelatedPostsProps) { + return ( + + ) +} diff --git a/apps/sim/app/(landing)/components/content-post-page/index.ts b/apps/sim/app/(landing)/components/content-post-page/index.ts index dde5612e99f..464d0804bde 100644 --- a/apps/sim/app/(landing)/components/content-post-page/index.ts +++ b/apps/sim/app/(landing)/components/content-post-page/index.ts @@ -1,2 +1,3 @@ export { ContentPostLoading } from './content-post-loading' export { ContentPostPage } from './content-post-page' +export { ContentRelatedPosts } from './content-related-posts' diff --git a/apps/sim/app/(landing)/components/content-utils.ts b/apps/sim/app/(landing)/components/content-utils.ts new file mode 100644 index 00000000000..0f3fac2937d --- /dev/null +++ b/apps/sim/app/(landing)/components/content-utils.ts @@ -0,0 +1,9 @@ +/** Renders an ISO date as "Jul 1, 2026". Pinned to UTC so the day matches the frontmatter date in every reader's timezone. */ +export function formatPostDate(iso: string): string { + return new Date(iso).toLocaleDateString('en-US', { + month: 'short', + day: 'numeric', + year: 'numeric', + timeZone: 'UTC', + }) +} diff --git a/apps/sim/app/(landing)/components/hero/hero.tsx b/apps/sim/app/(landing)/components/hero/hero.tsx index 0f04536719d..8854a654621 100644 --- a/apps/sim/app/(landing)/components/hero/hero.tsx +++ b/apps/sim/app/(landing)/components/hero/hero.tsx @@ -1,4 +1,5 @@ import { cn } from '@sim/emcn' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { HeroAnnouncementChip } from '@/app/(landing)/components/hero/components/hero-announcement-chip' import { LandingHeroHeader } from '@/app/(landing)/components/hero/components/hero-header' import { HeroPlatformStage } from '@/app/(landing)/components/hero/components/hero-platform-stage' @@ -31,9 +32,9 @@ export function Hero() { >

Sim is the open-source AI workspace where teams build, deploy, and manage AI agents for - their organization. Connect hundreds of integrations and every major LLM, then govern - access, spend, data, and deployment from one place. Build visually, conversationally, or - with code, and run Sim in your own cloud. + their organization. Connect {INTEGRATION_COUNT_LABEL} integrations and every major LLM, then + govern access, spend, data, and deployment from one place. Build visually, conversationally, + or with code, and run Sim in your own cloud.

diff --git a/apps/sim/app/(landing)/components/home-structured-data/home-structured-data.tsx b/apps/sim/app/(landing)/components/home-structured-data/home-structured-data.tsx index d70a54a25ab..b82a0c8895b 100644 --- a/apps/sim/app/(landing)/components/home-structured-data/home-structured-data.tsx +++ b/apps/sim/app/(landing)/components/home-structured-data/home-structured-data.tsx @@ -1,10 +1,11 @@ import { SITE_URL } from '@/lib/core/utils/urls' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { JsonLd } from '@/app/(landing)/components/json-ld' /** * Home-page JSON-LD - the entities specific to `/`: the `WebPage`, its * `BreadcrumbList`, the product `WebApplication` (`#software`, with offers / - * featureList / reviews), and the `SoftwareSourceCode`. + * featureList), and the `SoftwareSourceCode`. * * Rendered only by the landing root (`landing.tsx`), server-side before visible * content. The site-wide `Organization` / `WebSite` entities live in @@ -14,15 +15,15 @@ import { JsonLd } from '@/app/(landing)/components/json-ld' * Maintenance: * - Offer prices must match the Pricing component exactly. * - All claims must also appear as visible text on the page. - * - Do not add `aggregateRating` without real, verifiable review data. + * - Do not add `review` or `aggregateRating` without real, rated, verifiable + * review data; curated testimonials do not qualify for review snippets. */ /** * The home page's canonical description - the single string shared by the * ``, OG/Twitter cards (`page.tsx`), and the JSON-LD * `WebPage.description` below, so the three surfaces never drift. */ -export const HOME_PAGE_DESCRIPTION = - 'Sim is the open-source AI workspace where companies build, distribute, and govern AI agents. Hundreds of integrations, every major LLM, permission groups, spend limits, and self-hosting.' +export const HOME_PAGE_DESCRIPTION = `Sim is the open-source AI workspace where companies build, distribute, and govern AI agents. ${INTEGRATION_COUNT_LABEL} integrations, every major LLM, permission groups, spend limits, and self-hosting.` /** * The home page's canonical title - the single string shared by the @@ -61,8 +62,7 @@ const HOME_JSON_LD = { '@id': `${SITE_URL}#software`, url: SITE_URL, name: 'Sim, The AI Workspace', - description: - 'Sim is the open-source AI workspace where companies build, distribute, and govern AI agents in one place. Teams build agents visually, conversationally, or with code across hundreds of integrations and every major LLM, while administrators control model access, integration access, spend limits, and deployment. Trusted by over 100,000 builders. SOC2 compliant and self-hostable.', + description: `Sim is the open-source AI workspace where companies build, distribute, and govern AI agents in one place. Teams build agents visually, conversationally, or with code across ${INTEGRATION_COUNT_LABEL} integrations and every major LLM, while administrators control model access, integration access, spend limits, and deployment. Trusted by over 100,000 builders. SOC2 compliant and self-hostable.`, applicationCategory: 'BusinessApplication', applicationSubCategory: 'AI Workspace', operatingSystem: 'Web', @@ -110,7 +110,7 @@ const HOME_JSON_LD = { 'Chat: build and manage agents in natural language', 'Visual workflow builder', 'CLI access for coding agents and terminal workflows', - 'Hundreds of integrations', + `${INTEGRATION_COUNT_LABEL} integrations`, 'LLM orchestration (OpenAI, Anthropic, Google, xAI, Mistral, Perplexity)', 'Knowledge base creation', 'Table creation', @@ -128,28 +128,6 @@ const HOME_JSON_LD = { 'Configurable data retention', 'Self-hosting with Docker or Kubernetes', ], - review: [ - { - '@type': 'Review', - author: { '@type': 'Person', name: 'Hasan Toor' }, - reviewBody: - 'This startup just dropped the fastest way to build AI agents. This Figma-like canvas to build agents will blow your mind.', - url: 'https://x.com/hasantoxr/status/1912909502036525271', - }, - { - '@type': 'Review', - author: { '@type': 'Person', name: 'nizzy' }, - reviewBody: - 'This is the zapier of agent building. I always believed that building agents and using AI should not be limited to technical people. I think this solves just that.', - url: 'https://x.com/nizzyabi/status/1907864421227180368', - }, - { - '@type': 'Review', - author: { '@type': 'Organization', name: 'xyflow' }, - reviewBody: 'A very good looking agent workflow builder and open source!', - url: 'https://x.com/xyflowdev/status/1909501499719438670', - }, - ], }, { '@type': 'SoftwareSourceCode', diff --git a/apps/sim/app/(landing)/components/index.ts b/apps/sim/app/(landing)/components/index.ts index d0ee91b5b1d..54fc62209e3 100644 --- a/apps/sim/app/(landing)/components/index.ts +++ b/apps/sim/app/(landing)/components/index.ts @@ -4,7 +4,7 @@ export { ChevronArrow } from './chevron-arrow' export { ContentAuthorLoading, ContentAuthorPage } from './content-author-page' export { ContentImage } from './content-image' export { ContentIndexLoading, ContentIndexPage } from './content-index-page' -export { ContentPostLoading, ContentPostPage } from './content-post-page' +export { ContentPostLoading, ContentPostPage, ContentRelatedPosts } from './content-post-page' export { ContentTagsLoading, ContentTagsPage } from './content-tags-page' export { Cta } from './cta/cta' export { FeaturedCustomer } from './featured-customer' diff --git a/apps/sim/app/(landing)/components/site-structured-data/site-structured-data.tsx b/apps/sim/app/(landing)/components/site-structured-data/site-structured-data.tsx index 591920d098e..18df7749882 100644 --- a/apps/sim/app/(landing)/components/site-structured-data/site-structured-data.tsx +++ b/apps/sim/app/(landing)/components/site-structured-data/site-structured-data.tsx @@ -1,4 +1,5 @@ import { SITE_URL } from '@/lib/core/utils/urls' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { JsonLd } from '@/app/(landing)/components/json-ld' const SITE_JSON_LD = { @@ -10,8 +11,7 @@ const SITE_JSON_LD = { name: 'Sim', alternateName: 'Sim Studio', legalName: 'Sim, Inc', - description: - 'Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect hundreds of integrations and every major LLM to create agents that automate real work.', + description: `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect ${INTEGRATION_COUNT_LABEL} integrations and every major LLM to create agents that automate real work.`, url: SITE_URL, foundingDate: '2025', address: { @@ -25,10 +25,10 @@ const SITE_JSON_LD = { logo: { '@type': 'ImageObject', '@id': `${SITE_URL}#logo`, - url: `${SITE_URL}/logo/b%26w/text/b%26w.svg`, - contentUrl: `${SITE_URL}/logo/b%26w/text/b%26w.svg`, - width: 49.78314, - height: 24.276, + url: `${SITE_URL}/favicon/android-chrome-512x512.png`, + contentUrl: `${SITE_URL}/favicon/android-chrome-512x512.png`, + width: 512, + height: 512, caption: 'Sim Logo', }, image: { '@id': `${SITE_URL}#logo` }, @@ -59,8 +59,7 @@ const SITE_JSON_LD = { '@id': `${SITE_URL}#website`, url: SITE_URL, name: 'Sim, The AI Workspace | Build, Deploy & Manage AI Agents', - description: - 'Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect hundreds of integrations and every major LLM. Join 100,000+ builders.', + description: `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect ${INTEGRATION_COUNT_LABEL} integrations and every major LLM. Join 100,000+ builders.`, publisher: { '@id': `${SITE_URL}#organization` }, inLanguage: 'en-US', }, diff --git a/apps/sim/app/(landing)/demo/demo.tsx b/apps/sim/app/(landing)/demo/demo.tsx index 85113e9effc..8b62a0c72a5 100644 --- a/apps/sim/app/(landing)/demo/demo.tsx +++ b/apps/sim/app/(landing)/demo/demo.tsx @@ -1,3 +1,4 @@ +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { TrustedBy } from '@/app/(landing)/components/trusted-by' import { DemoBooking } from '@/app/(landing)/demo/components/demo-booking' @@ -45,8 +46,8 @@ export default function Demo() {

Operationalize AI with Sim, the AI agent workspace where teams build, deploy, and manage AI agents and workflows. A Sim specialist walks your team through building agents that - automate real work across hundreds of integrations and every major LLM, visually, - conversationally, or with code. + automate real work across {INTEGRATION_COUNT_LABEL} integrations and every major LLM, + visually, conversationally, or with code.

[i.type, i])) export const dynamicParams = true /** - * Returns up to `limit` related integration slugs. + * Returns up to `limit` related integration slugs from the same category, so + * the section links within the topical cluster (both CRMs, both devops tools). * * Scoring (additive): * +3 per shared operation name - strongest signal (same capability) * +2 per shared operation word - weaker signal (e.g. both have "create" ops) - * +2 same integration category - topical relevance (both CRMs, both devops) * +1 same auth type - comparable setup experience * - * Every integration gets a score, so the sidebar always has suggestions. * Ties are broken by alphabetical slug order for determinism. */ function getRelatedSlugs( @@ -68,7 +66,7 @@ function getRelatedSlugs( operations: Integration['operations'], authType: AuthType, integrationType: Integration['integrationType'], - limit = 6 + limit = 4 ): string[] { const currentOpNames = new Set(operations.map((o) => o.name.toLowerCase())) const currentOpWords = new Set( @@ -82,7 +80,7 @@ function getRelatedSlugs( return allIntegrations .reduce>((scored, i) => { - if (i.slug === slug) return scored + if (i.slug === slug || i.integrationType !== integrationType) return scored const sharedNames = i.operations.filter((o) => currentOpNames.has(o.name.toLowerCase()) ).length @@ -92,12 +90,8 @@ function getRelatedSlugs( .split(/\s+/) .some((w) => w.length > 3 && currentOpWords.has(w)) ).length - const sameCategory = i.integrationType === integrationType ? 2 : 0 const sameAuth = i.authType === authType ? 1 : 0 - scored.push({ - slug: i.slug, - score: sharedNames * 3 + sharedWords * 2 + sameCategory + sameAuth, - }) + scored.push({ slug: i.slug, score: sharedNames * 3 + sharedWords * 2 + sameAuth }) return scored }, []) .sort((a, b) => b.score - a.score || a.slug.localeCompare(b.slug)) @@ -108,11 +102,15 @@ function getRelatedSlugs( const AUTH_STEP: Record string> = { oauth: (name) => `Connect your ${name} account with one-click OAuth, with no credentials to copy.`, - 'api-key': (name) => - `Paste your ${name} API key to authenticate. You can find it in your ${name} account settings.`, + 'api-key': (name) => `Paste your ${name} API key to authenticate.`, none: () => 'No authentication is needed, so the block works as soon as you drop it in.', } +/** Default H1 and page name, e.g. "Slack integration for AI agents". */ +function integrationTitle(name: string): string { + return `${name} integration for AI agents` +} + /** Human-readable catalog refresh date for the visible last-updated line. */ const UPDATED_AT_DISPLAY = new Date(`${INTEGRATIONS_UPDATED_AT}T00:00:00Z`).toLocaleDateString( 'en-US', @@ -154,9 +152,17 @@ function mentionifyPromptForNames(prompt: string, names: readonly string[]): str return prompt.replace(regex, (match) => `@${match}`) } -/** Lowercases only the first character so acronyms in tool names survive. */ -function lowercaseFirst(value: string): string { - return value.charAt(0).toLowerCase() + value.slice(1) +/** + * Turns a Title Case tool name into a mid-sentence phrase (`Get Markets` → + * `get markets`). Only plain capitalized words are lowercased, so acronyms + * (`PR`), mixed-case brands (`GitHub`), and the integration's own name survive. + */ +function toPhrase(toolName: string, integrationName: string): string { + const keep = new Set(integrationName.split(/\s+/)) + return toolName + .split(' ') + .map((word) => (!keep.has(word) && /^[A-Z][a-z]+$/.test(word) ? word.toLowerCase() : word)) + .join(' ') } /** @@ -208,101 +214,122 @@ function toProseList(items: string[]): string { return `${items.slice(0, -1).join(', ')}, and ${items[items.length - 1]}` } +/** Human-readable authentication method, as stated in the at-a-glance facts. */ +const AUTH_LABEL: Record = { + oauth: 'OAuth', + 'api-key': 'API key', + none: 'None required', +} + +function pluralize(count: number, noun: string): string { + return `${count} ${noun}${count === 1 ? '' : 's'}` +} + +/** Tool and trigger counts as one phrase, e.g. `"19 tools and 1 trigger"`; `''` when both are zero. */ +function capabilityPhrase(integration: Integration): string { + return [ + integration.operations.length > 0 ? pluralize(integration.operations.length, 'tool') : null, + integration.triggers.length > 0 ? pluralize(integration.triggers.length, 'trigger') : null, + ] + .filter((part): part is string => part !== null) + .join(' and ') +} + +const META_DESCRIPTION_MAX = 160 + +/** + * Meta description built from registry facts: the block description, then a + * sample of the integration's actual tools, then its trigger count. Whole + * sentences are kept while they fit, so the description never cuts mid-word. + */ +function buildMetaDescription(integration: Integration): string { + const { name, description, operations, triggers } = integration + const sentences = [sentenceWithTerminalPunctuation(description)] + if (operations.length > 0) { + const sample = operations.slice(0, 3).map((o) => toPhrase(o.name, name)) + const remaining = operations.length - sample.length + sentences.push( + `Sim AI agents can ${toProseList(remaining > 0 ? [...sample, pluralize(remaining, `more ${name} action`)] : sample)}.` + ) + } + if (triggers.length > 0) { + sentences.push( + `${name} events can start agents through ${pluralize(triggers.length, 'trigger')}.` + ) + } + return sentences.reduce((text, sentence) => { + const next = `${text} ${sentence}` + return next.length <= META_DESCRIPTION_MAX ? next : text + }) +} + /** "a" vs "an" for a service name; U-names read as "you", so they take "a". */ function articleFor(name: string): string { return /^[aeio]/i.test(name) ? 'an' : 'a' } /** - * Generates the per-integration FAQ. Answers lead with a direct answer and - * carry integration-specific facts; catalog-generic questions live once on - * the /integrations index FAQ instead of repeating across every page. + * Generates the per-integration FAQ from catalog facts only: the block + * description, its tool and trigger names and descriptions, and its auth + * method. Catalog-generic questions live once on the /integrations index FAQ + * instead of repeating across every page. */ -function buildFAQs(integration: Integration, relatedNames: string[]): FAQItem[] { +function buildFAQs(integration: Integration): FAQItem[] { const { name, description, operations, triggers, authType } = integration - const faqDescription = sentenceWithTerminalPunctuation(description) const opCount = operations.length const triggerCount = triggers.length const topOpNames = operations.slice(0, 5).map((o) => o.name) const firstOp = operations[0] - const firstTrigger = triggers[0] - const pairings = relatedNames.slice(0, 2) - const toolsPhrase = `${opCount} ${name} tool${opCount === 1 ? '' : 's'}` - const triggersPhrase = `${triggerCount} real-time trigger${triggerCount === 1 ? '' : 's'}` - const capabilityPhrase = [ - opCount > 0 ? toolsPhrase : null, - triggerCount > 0 ? triggersPhrase : null, - ] - .filter((part): part is string => part !== null) - .join(' and ') + const capability = capabilityPhrase(integration) const triggerNames = triggers.map((t) => t.name) const triggerListPhrase = triggerCount > 6 ? `${triggerNames.slice(0, 6).join(', ')}, and ${triggerCount - 6} more` : toProseList(triggerNames) - const firstTriggerWhen = firstTrigger?.description.match(/^trigger workflow (when .+)$/i)?.[1] const connectFinalStep = firstOp - ? `Pick a tool such as "${firstOp.name}", wire up its inputs, and click Run, and your agent is live.` + ? `Pick a tool such as "${firstOp.name}", wire up its inputs, and click Run.` : triggerCount > 0 - ? `Choose the ${name} event you want to listen for, and your agent runs automatically from then on.` - : `Configure the block's inputs and click Run, and your agent is live.` + ? `Choose the ${name} event you want to listen for, and your agent runs whenever it occurs.` + : `Configure the block's inputs and click Run.` - const faqs: FAQItem[] = [ + return [ { question: `What is Sim's ${name} integration?`, - answer: `Sim's ${name} integration ${capabilityPhrase ? `adds ${capabilityPhrase} to` : `connects ${name} to`} the AI agents you build in Sim's visual workflow builder — you build it all visually. ${faqDescription}${ - pairings.length === 2 - ? ` Teams often pair ${name} with ${pairings[0]} and ${pairings[1]} in the same agent.` - : '' - }`, + answer: `Sim's ${name} integration ${capability ? `adds ${capability} to` : `connects ${name} to`} the AI agents you build in Sim's visual workflow builder. ${sentenceWithTerminalPunctuation(description)}`, }, ...(opCount > 0 ? [ { question: `What can I automate with ${name} in Sim?`, - answer: `You can ${toProseList(topOpNames.map(lowercaseFirst))} with ${name} in Sim${ + answer: `You can ${toProseList(topOpNames.map((n) => toPhrase(n, name)))} with ${name} in Sim${ opCount > 5 ? `, plus ${opCount - 5} more ${name} tools listed on this page` : '' - }. ${opCount === 1 ? 'It runs' : 'Each runs'} as a tool inside an AI agent block, so an agent can chain ${name} with ${ - pairings.length === 2 - ? `services like ${pairings[0]} and ${pairings[1]}` - : 'any other connected service' - } and apply LLM reasoning between steps.`, + }. ${opCount === 1 ? 'It runs' : 'Each runs'} as a tool inside an AI agent, so the agent can combine ${name} with any other connected service and apply LLM reasoning between steps.`, }, ] : []), { question: `How do I connect ${name} to Sim?`, - answer: `Connecting ${name} takes about five minutes: (1) Create a free account at sim.ai. (2) Create an agent in your workspace. (3) Drag ${articleFor(name)} ${name} block onto the workflow builder. (4) ${AUTH_STEP[authType](name)} (5) ${connectFinalStep}`, + answer: `(1) Create a free account at sim.ai. (2) Create an agent in your workspace. (3) Drag ${articleFor(name)} ${name} block onto the workflow builder. (4) ${AUTH_STEP[authType](name)} (5) ${connectFinalStep}`, }, ...(firstOp && opCount >= 2 ? [ { - question: `How do I ${lowercaseFirst(firstOp.name)} with ${name} in Sim?`, + question: `How do I ${toPhrase(firstOp.name, name)} with ${name} in Sim?`, answer: `Add ${articleFor(name)} ${name} block to your agent and select "${firstOp.name}" as the tool.${ firstOp.description ? ` ${sentenceWithTerminalPunctuation(firstOp.description)}` : '' - } Fill in the required fields. Inputs can reference outputs from earlier steps, such as text generated by an AI block or data fetched from another integration, and you build it all visually.`, + } Fill in the required fields. Inputs can reference outputs from earlier steps, such as text generated by an AI block or data fetched from another integration.`, }, ] : []), ...(triggerCount > 0 ? [ { - question: `How do I trigger a Sim agent from ${name} automatically?`, - answer: `Add ${articleFor(name)} ${name} trigger block to your agent and copy its generated webhook URL into ${name}'s webhook settings. Sim supports ${triggersPhrase} for ${name}: ${triggerListPhrase}. Once configured, every matching ${name} event starts your agent instantly, no polling, no delay.`, - }, - { - question: `What data does Sim receive when a ${name} event triggers an agent?`, - answer: `Sim receives the full event payload ${name} sends, typically the record or object that changed, plus metadata like the event type and timestamp.${ - firstTriggerWhen - ? ` For example, the "${firstTrigger.name}" trigger fires ${sentenceWithTerminalPunctuation(firstTriggerWhen)}` - : '' - } Every field in the payload is available as a variable you can pass to AI blocks, conditions, or other integrations.`, + question: `Can ${name} events start a Sim agent automatically?`, + answer: `Yes. Sim supports ${pluralize(triggerCount, 'trigger')} for ${name}: ${triggerListPhrase}. Add ${articleFor(name)} ${name} trigger to your agent, and every matching ${name} event starts a run with the event data available to the rest of the workflow.`, }, ] : []), ] - - return faqs } export async function generateStaticParams() { @@ -318,21 +345,21 @@ export async function generateMetadata({ const integration = bySlug.get(slug) if (!integration) return {} - const { name, description, operations } = integration + const { name, operations } = integration const opSample = operations .slice(0, 3) .map((o) => o.name) .join(', ') const categoryLabel = formatIntegrationType(integration.integrationType) const seo = INTEGRATION_SEO[slug] - const metaDesc = - seo?.description ?? - `Automate ${name} with AI agents in Sim. ${sentenceWithTerminalPunctuation(truncate(description, 100))} Free to start.` + const metaDesc = seo?.description ?? buildMetaDescription(integration) + const defaultTitle = integrationTitle(name) + const pageUrl = `${baseUrl}/integrations/${slug}` return { // A hand-authored SEO title is rendered verbatim (it carries its own brand - // suffix); otherwise the bare name flows through the root `%s | Sim` template. - title: seo?.title ? { absolute: seo.title } : `${name} Integration`, + // suffix); otherwise the default flows through the root `%s | Sim` template. + title: seo?.title ? { absolute: seo.title } : defaultTitle, description: metaDesc, keywords: seo?.keywords ?? [ `${name} automation`, @@ -352,21 +379,17 @@ export async function generateMetadata({ // og:image/twitter:image come from the sibling opengraph-image.tsx - // Next serves it at a hash-suffixed URL, so hardcoding it here 404s. openGraph: { - title: seo?.title ?? `${name} Integration | Sim AI Workspace`, - description: - seo?.description ?? - `Connect ${name} to ${INTEGRATION_COUNT - 1}+ tools using AI agents. ${sentenceWithTerminalPunctuation(truncate(description, 100))}`, - url: `${baseUrl}/integrations/${slug}`, + title: seo?.title ?? `${defaultTitle} | Sim`, + description: metaDesc, + url: pageUrl, type: 'website', }, twitter: { card: 'summary_large_image', - title: seo?.title ?? `${name} Integration | Sim`, - description: - seo?.description ?? - `Automate ${name} with AI agents in Sim. Connect to ${INTEGRATION_COUNT - 1}+ tools. Free to start.`, + title: seo?.title ?? `${defaultTitle} | Sim`, + description: metaDesc, }, - alternates: { canonical: `${baseUrl}/integrations/${slug}` }, + alternates: { canonical: pageUrl }, } } @@ -388,12 +411,9 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl const relatedIntegrations = relatedSlugs .map((s) => bySlug.get(s)) .filter((i): i is Integration => i !== undefined) - const faqs = - seo?.faqs ?? - buildFAQs( - integration, - relatedIntegrations.map((i) => i.name) - ) + const faqs = seo?.faqs ?? buildFAQs(integration) + const capability = capabilityPhrase(integration) + const pageUrl = `${baseUrl}/integrations/${slug}` const matchingTemplates = getTemplatesForBlock(integration.type) .sort( (a, b) => @@ -413,25 +433,40 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl name: 'Integrations', item: `${baseUrl}/integrations`, }, - { '@type': 'ListItem', position: 3, name, item: `${baseUrl}/integrations/${slug}` }, + { '@type': 'ListItem', position: 3, name, item: pageUrl }, ], } - const softwareAppJsonLd = { + const webPageJsonLd = { '@context': 'https://schema.org', - '@type': 'SoftwareApplication', - name: `${name} Integration`, + '@type': 'WebPage', + '@id': pageUrl, + url: pageUrl, + name: integrationTitle(name), description, - url: `${baseUrl}/integrations/${slug}`, - applicationCategory: 'BusinessApplication', - applicationSubCategory: categoryLabel, - operatingSystem: 'Web', - featureList: operations.map((o) => o.name), + isPartOf: { '@id': `${baseUrl}#website` }, + publisher: { '@id': `${baseUrl}#organization` }, + about: { '@type': 'Thing', name }, + inLanguage: 'en-US', + dateModified: INTEGRATIONS_UPDATED_AT, ...(integration.tags?.length ? { keywords: integration.tags.map((tag) => tag.replace(/-/g, ' ')).join(', ') } : {}), - dateModified: INTEGRATIONS_UPDATED_AT, - offers: { '@type': 'Offer', price: '0', priceCurrency: 'USD' }, + ...(operations.length + triggers.length > 0 + ? { + mainEntity: { + '@type': 'ItemList', + name: `${name} tools and triggers in Sim`, + numberOfItems: operations.length + triggers.length, + itemListElement: [...operations, ...triggers].map((item, index) => ({ + '@type': 'ListItem', + position: index + 1, + name: item.name, + description: item.description, + })), + }, + } + : {}), } const faqJsonLd = { @@ -447,7 +482,7 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl return (
- + {/* Hero */} @@ -472,7 +507,7 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl id='integration-heading' className='text-[28px] text-[var(--text-primary)] leading-[110%] tracking-[-0.02em] sm:text-[36px] lg:text-[44px]' > - {seo?.h1 ?? name} + {seo?.h1 ?? integrationTitle(name)}

@@ -482,19 +517,10 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

- {name} is a {categoryLabel} integration for Sim, the AI workspace where teams build and - deploy AI agents. Sim's {name} integration provides{' '} - {[ - operations.length > 0 - ? `${operations.length} ${name} tool${operations.length === 1 ? '' : 's'}` - : null, - triggers.length > 0 - ? `${triggers.length} real-time trigger${triggers.length === 1 ? '' : 's'}` - : null, - ] - .filter((part): part is string => part !== null) - .join(' and ') || `a ${name} connection`}{' '} - that AI agents can use inside Sim's visual workflow builder.{' '} + {name} is a Sim integration in the {categoryLabel} category. Sim is the AI workspace where + teams build and deploy AI agents. Sim's {name} integration provides{' '} + {capability || `a ${name} connection`} that AI agents can use inside Sim's visual + workflow builder.{' '} {authType === 'oauth' ? `${name} connects with one-click OAuth.` : authType === 'api-key' @@ -516,7 +542,7 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl > View docs - +

@@ -529,23 +555,47 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl {/* Border-railed content */}

- {/* Overview */} - {overviewBody && ( - <> -
-

- Overview -

-

- {overviewBody} -

-
-
- - )} + {/* Overview + at-a-glance facts */} +
+

+ What Sim agents can do with {name} +

+ {overviewBody && ( +

+ {overviewBody} +

+ )} +
+ {[ + { term: 'Category', detail: categoryLabel }, + { term: 'Authentication', detail: AUTH_LABEL[authType] }, + { term: 'Tools', detail: String(operations.length) }, + { term: 'Triggers', detail: String(triggers.length) }, + ].map(({ term, detail }) => ( +
+
{term}
+
{detail}
+
+ ))} +
+
Documentation
+
+ + {name} docs + +
+
+
+
+
{/* Install / Add to workspace (integration-specific) */} {landingContent?.install && ( @@ -553,11 +603,11 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

{landingContent.install.heading}

-

+

{landingContent.install.intro}

    @@ -573,7 +623,7 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

    {item.title}

    -

    +

    {item.body}

@@ -594,11 +644,11 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

Privacy & data

-

+

{landingContent.privacy.body}{' '}

AI-generated content

-

+

{landingContent.aiDisclaimer}

@@ -635,7 +685,7 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

How to automate {name} with Sim

@@ -673,7 +723,7 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

{title}

-

+

{body}

@@ -702,16 +752,16 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

- Real-time triggers + {name} triggers

-

+

{seo?.triggersIntro ?? ( <> - Connect {articleFor(name)} {name} webhook to Sim and your agent runs the instant - an event happens, no polling, no delay. + Add {articleFor(name)} {name} trigger to a Sim agent and it starts a run + whenever one of these {name} events occurs. )}

@@ -721,11 +771,11 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl
-

+

{trigger.name} -

+

{trigger.description && ( -

+

{trigger.description}

)} @@ -743,7 +793,7 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

Agent templates

@@ -810,7 +860,7 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl
-

+

{template.title}

@@ -836,10 +886,10 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

-

+

{lastTemplate.title}

-

+

{lastTemplate.prompt}

@@ -859,13 +909,13 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

- Supported tools + {name} tools

- {operations.length} {name} tool{operations.length === 1 ? '' : 's'} available in Sim - {seo?.toolsSubtitleSuffix ?? ''} + {pluralize(operations.length, `${name} tool`)} available to Sim agents + {seo?.toolsSubtitleSuffix ?? ''}.

@@ -873,11 +923,11 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl
-

+

{op.name} -

+

{op.description && ( -

+

{op.description}

)} @@ -921,7 +971,7 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl

Frequently asked questions

@@ -932,9 +982,18 @@ export default async function IntegrationPage({ params }: { params: Promise<{ sl {/* Related integrations - horizontal cards with vertical dividers (blog featured pattern) */} {relatedIntegrations.length > 0 && ( - <> +
+
+ +
+
)}
diff --git a/apps/sim/app/(landing)/integrations/(shell)/page.tsx b/apps/sim/app/(landing)/integrations/(shell)/page.tsx index 045896769ca..7209c912fbb 100644 --- a/apps/sim/app/(landing)/integrations/(shell)/page.tsx +++ b/apps/sim/app/(landing)/integrations/(shell)/page.tsx @@ -45,7 +45,7 @@ const CATALOG_FAQS: FAQItem[] = [ }, { question: 'Can external events trigger my agents automatically?', - answer: `Yes. ${TRIGGER_INTEGRATION_COUNT} Sim integrations include real-time webhook triggers. Add a trigger block to your agent, copy its webhook URL into the external service, and every matching event starts your agent instantly, no polling, no delay.`, + answer: `Yes. ${TRIGGER_INTEGRATION_COUNT} Sim integrations include triggers, delivered by webhook or by polling depending on the service. Add a trigger block to your agent, and every matching event in the external service starts a run.`, }, { question: 'How many integrations does Sim support?', @@ -99,7 +99,7 @@ export async function generateMetadata({ return withFilteredNoindex( { - title: 'Integrations', + title: 'Integrations for AI Agents', description: `Connect ${INTEGRATION_COUNT}+ apps and services in Sim's AI workspace. Build agents that automate real work with ${TOP_NAMES.join(', ')}, and more.`, keywords: [ 'AI workspace integrations', @@ -144,14 +144,9 @@ export default async function IntegrationsPage({ itemListElement: allIntegrations.map((integration, index) => ({ '@type': 'ListItem', position: index + 1, - item: { - '@type': 'SoftwareApplication', - name: integration.name, - description: integration.description, - url: `${baseUrl}/integrations/${integration.slug}`, - applicationCategory: 'BusinessApplication', - featureList: integration.operations.map((o) => o.name), - }, + name: integration.name, + description: integration.description, + url: `${baseUrl}/integrations/${integration.slug}`, })), } @@ -173,12 +168,19 @@ export default async function IntegrationsPage({ {/* Hero */}
+

+ Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. + Sim's catalog lists {INTEGRATION_COUNT} integrations that together give AI agents{' '} + {TOTAL_TOOL_COUNT.toLocaleString('en-US')} tools. {TRIGGER_INTEGRATION_COUNT} integrations + include triggers that start an agent from an external event, and {OAUTH_COUNT} connect + with one-click OAuth. +

- Integrations + Integrations for AI agents

Connect every tool your team uses. Build agents that automate real work across{' '} diff --git a/apps/sim/app/(landing)/library/[slug]/page.tsx b/apps/sim/app/(landing)/library/[slug]/page.tsx index 3b7e256976b..57df49bb272 100644 --- a/apps/sim/app/(landing)/library/[slug]/page.tsx +++ b/apps/sim/app/(landing)/library/[slug]/page.tsx @@ -1,8 +1,9 @@ import type { Metadata } from 'next' import { notFound } from 'next/navigation' -import { getBaseUrl } from '@/lib/core/utils/urls' import { getAllPostMeta, getPostBySlug, getRelatedPosts } from '@/lib/library/registry' import { buildPostGraphJsonLd, buildPostMetadata, LIBRARY_SECTION } from '@/lib/library/seo' +import { ComparisonLinks } from '@/app/(landing)/comparisons/components/comparison-links' +import { getComparisonsForPost } from '@/app/(landing)/comparisons/library-links' import { ContentPostPage } from '@/app/(landing)/components' /** Unknown slugs reach the section 404 while known pages remain pre-rendered. */ @@ -31,6 +32,7 @@ export default async function Page({ params }: { params: Promise<{ slug: string const post = await getPostBySlug(slug) if (!post || post.draft) notFound() const related = await getRelatedPosts(slug, 3) + const comparisons = getComparisonsForPost(post) return ( 0 ? : undefined + } /> ) } diff --git a/apps/sim/app/(landing)/pricing/components/pricing-structured-data/pricing-structured-data.tsx b/apps/sim/app/(landing)/pricing/components/pricing-structured-data/pricing-structured-data.tsx index 016d1f9cb79..ffd1869c317 100644 --- a/apps/sim/app/(landing)/pricing/components/pricing-structured-data/pricing-structured-data.tsx +++ b/apps/sim/app/(landing)/pricing/components/pricing-structured-data/pricing-structured-data.tsx @@ -1,5 +1,6 @@ import { CREDIT_TIERS } from '@/lib/billing/constants' import { SITE_URL } from '@/lib/core/utils/urls' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { JsonLd } from '@/app/(landing)/components/json-ld' import { COMPARISON_SECTIONS } from '@/app/workspace/[workspaceId]/upgrade/components/comparison-table/comparison-data' @@ -37,8 +38,7 @@ const PRICING_JSON_LD = { '@type': 'WebApplication', '@id': `${PAGE_URL}#application`, name: 'Sim', - description: - 'Sim is the open-source AI workspace where teams build, deploy, and manage AI agents, connecting hundreds of integrations and every major LLM.', + description: `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents, connecting ${INTEGRATION_COUNT_LABEL} integrations and every major LLM.`, applicationCategory: 'BusinessApplication', operatingSystem: 'Web', url: SITE_URL, diff --git a/apps/sim/app/(landing)/pricing/pricing.tsx b/apps/sim/app/(landing)/pricing/pricing.tsx index 6df741bc4af..631a9f64d98 100644 --- a/apps/sim/app/(landing)/pricing/pricing.tsx +++ b/apps/sim/app/(landing)/pricing/pricing.tsx @@ -1,3 +1,4 @@ +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { PricingPlans } from '@/app/(landing)/pricing/components/pricing-plans' import { PricingStructuredData } from '@/app/(landing)/pricing/components/pricing-structured-data' @@ -5,8 +6,7 @@ import { PricingStructuredData } from '@/app/(landing)/pricing/components/pricin * sr-only product summary - an atomic citation target for AI answer engines that * names Sim, the AI workspace, AI agents, and every plan tier. */ -const GEO_SUMMARY = - 'Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Pricing scales across four plans: Free to start, Pro for growing teams, Max for scaling businesses, and Enterprise for large organizations, each connecting hundreds of integrations and every major LLM.' +const GEO_SUMMARY = `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Pricing scales across four plans: Free to start, Pro for growing teams, Max for scaling businesses, and Enterprise for large organizations, each connecting ${INTEGRATION_COUNT_LABEL} integrations and every major LLM.` /** Server-rendered heading slot handed to the {@link PricingPlans} client island. */ const PRICING_HEADING = ( diff --git a/apps/sim/app/(landing)/privacy/privacy-content.tsx b/apps/sim/app/(landing)/privacy/privacy-content.tsx index ffc57df52e2..d4f1bcca832 100644 --- a/apps/sim/app/(landing)/privacy/privacy-content.tsx +++ b/apps/sim/app/(landing)/privacy/privacy-content.tsx @@ -1,4 +1,5 @@ import { Fragment, type ReactNode } from 'react' +import { SITE_URL } from '@/lib/core/utils/urls' import { type LegalPageConfig, ProseLink } from '@/app/(landing)/components/prose-page' const INLINE_PATTERN = @@ -48,7 +49,7 @@ export const PRIVACY_CONFIG: LegalPageConfig = { { kind: 'paragraph', content: richText( - 'This Privacy Policy describes how Sim ("we", "us", "our", or "the Service") collects, uses, discloses, and protects personal data, including data obtained from Google APIs (including Google Workspace APIs), and your rights and controls regarding that data. This Privacy Policy is provided for transparency and information purposes only, including to satisfy the information obligations in Articles 13 and 14 of the General Data Protection Regulation ("GDPR"). It does not create contractual obligations on you. Your use of the Service is governed by the [Terms of Service](https://sim.ai/terms).' + `This Privacy Policy describes how Sim ("we", "us", "our", or "the Service") collects, uses, discloses, and protects personal data, including data obtained from Google APIs (including Google Workspace APIs), and your rights and controls regarding that data. This Privacy Policy is provided for transparency and information purposes only, including to satisfy the information obligations in Articles 13 and 14 of the General Data Protection Regulation ("GDPR"). It does not create contractual obligations on you. Your use of the Service is governed by the [Terms of Service](${SITE_URL}/terms).` ), }, ], @@ -199,7 +200,7 @@ export const PRIVACY_CONFIG: LegalPageConfig = { { kind: 'paragraph', content: richText( - 'The [Cookie Policy](https://sim.ai/cookie-policy) lists the Cookies set by Sim and its providers, their purposes, lifetimes, providers, and the methods for changing or withdrawing a choice.' + `The [Cookie Policy](${SITE_URL}/cookie-policy) lists the Cookies set by Sim and its providers, their purposes, lifetimes, providers, and the methods for changing or withdrawing a choice.` ), }, ], @@ -582,7 +583,7 @@ export const PRIVACY_CONFIG: LegalPageConfig = { { kind: 'paragraph', content: richText( - 'You may change or withdraw consent at any time through the cookie preferences link. If Your browser or extension sends a Global Privacy Control signal, we treat it as a withdrawal of consent for analytics and marketing Cookies. The [Cookie Policy](https://sim.ai/cookie-policy) explains the technologies, providers, purposes, lifetimes, and available controls.' + `You may change or withdraw consent at any time through the cookie preferences link. If Your browser or extension sends a Global Privacy Control signal, we treat it as a withdrawal of consent for analytics and marketing Cookies. The [Cookie Policy](${SITE_URL}/cookie-policy) explains the technologies, providers, purposes, lifetimes, and available controls.` ), }, { diff --git a/apps/sim/app/(landing)/solutions/compliance/compliance.tsx b/apps/sim/app/(landing)/solutions/compliance/compliance.tsx index 1fa56c78384..d6f7c2d6b00 100644 --- a/apps/sim/app/(landing)/solutions/compliance/compliance.tsx +++ b/apps/sim/app/(landing)/solutions/compliance/compliance.tsx @@ -1,3 +1,4 @@ +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { PlatformHeroVisual, SolutionsPage, @@ -36,10 +37,8 @@ const COMPLIANCE_CONFIG: SolutionsPageConfig = { hero: { eyebrow: 'Compliance', heading: 'Automate evidence, control checks, and audit reports with AI agents in Sim.', - description: - 'Sim is the open-source AI workspace where compliance teams build AI agents for evidence collection and control monitoring. Stay audit-ready year-round, across hundreds of integrations.', - summary: - 'Sim is the open-source AI workspace where compliance teams build, deploy, and manage AI agents for evidence collection, control monitoring, and audit reports. Agents keep the organization audit-ready year-round across hundreds of integrations.', + description: `Sim is the open-source AI workspace where compliance teams build AI agents for evidence collection and control monitoring. Stay audit-ready year-round, across ${INTEGRATION_COUNT_LABEL} integrations.`, + summary: `Sim is the open-source AI workspace where compliance teams build, deploy, and manage AI agents for evidence collection, control monitoring, and audit reports. Agents keep the organization audit-ready year-round across ${INTEGRATION_COUNT_LABEL} integrations.`, visual: ( diff --git a/apps/sim/app/(landing)/solutions/engineering/engineering.tsx b/apps/sim/app/(landing)/solutions/engineering/engineering.tsx index 4723cc56a17..ff50d62defe 100644 --- a/apps/sim/app/(landing)/solutions/engineering/engineering.tsx +++ b/apps/sim/app/(landing)/solutions/engineering/engineering.tsx @@ -1,3 +1,4 @@ +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { PlatformHeroVisual, SolutionsPage, @@ -36,10 +37,8 @@ const ENGINEERING_CONFIG: SolutionsPageConfig = { hero: { eyebrow: 'Engineering', heading: 'Automate code review, on-call, and docs with AI agents in Sim.', - description: - 'Sim is the open-source AI workspace where engineering teams build AI agents for code review, on-call, and docs. Agents wire into GitHub, CI/CD, and hundreds of integrations across the software lifecycle.', - summary: - 'Sim is the open-source AI workspace where engineering teams build, deploy, and manage AI agents for code review, on-call triage, and documentation. Agents wire into GitHub, CI/CD, and hundreds of integrations across the software lifecycle.', + description: `Sim is the open-source AI workspace where engineering teams build AI agents for code review, on-call, and docs. Agents wire into GitHub, CI/CD, and ${INTEGRATION_COUNT_LABEL} integrations across the software lifecycle.`, + summary: `Sim is the open-source AI workspace where engineering teams build, deploy, and manage AI agents for code review, on-call triage, and documentation. Agents wire into GitHub, CI/CD, and ${INTEGRATION_COUNT_LABEL} integrations across the software lifecycle.`, visual: ( diff --git a/apps/sim/app/(landing)/solutions/finance/finance.tsx b/apps/sim/app/(landing)/solutions/finance/finance.tsx index 59706d608c5..610636276e2 100644 --- a/apps/sim/app/(landing)/solutions/finance/finance.tsx +++ b/apps/sim/app/(landing)/solutions/finance/finance.tsx @@ -1,3 +1,4 @@ +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { PlatformHeroVisual, SolutionsPage, @@ -37,8 +38,7 @@ const FINANCE_CONFIG: SolutionsPageConfig = { heading: 'Automate invoice processing, reconciliation, and close with AI agents in Sim.', description: 'Sim is the open-source AI workspace where finance teams build AI agents for invoice processing, reconciliation, and close. Human approvals, anomaly detection, and full audit trails guard every run.', - summary: - 'Sim is the open-source AI workspace where finance teams build, deploy, and manage AI agents for invoice processing, reconciliation, and financial reporting. Agents run with human approvals, anomaly detection, and full audit trails across hundreds of integrations.', + summary: `Sim is the open-source AI workspace where finance teams build, deploy, and manage AI agents for invoice processing, reconciliation, and financial reporting. Agents run with human approvals, anomaly detection, and full audit trails across ${INTEGRATION_COUNT_LABEL} integrations.`, visual: ( diff --git a/apps/sim/app/(landing)/solutions/hr/hr.tsx b/apps/sim/app/(landing)/solutions/hr/hr.tsx index 137269ffa54..5edc2ef1c8b 100644 --- a/apps/sim/app/(landing)/solutions/hr/hr.tsx +++ b/apps/sim/app/(landing)/solutions/hr/hr.tsx @@ -1,3 +1,4 @@ +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { PlatformHeroVisual, SolutionsPage, @@ -37,10 +38,8 @@ const HR_CONFIG: SolutionsPageConfig = { hero: { eyebrow: 'HR', heading: 'Automate onboarding, employee questions, and approvals with AI agents in Sim.', - description: - 'Sim is the open-source AI workspace where HR teams build AI agents for onboarding, employee questions, and approvals. Agents wire into your HRIS and hundreds of integrations to keep people operations moving.', - summary: - 'Sim is the open-source AI workspace where HR teams build, deploy, and manage AI agents for onboarding, employee questions, and approvals. Agents connect your HRIS and hundreds of integrations so people operations keep moving.', + description: `Sim is the open-source AI workspace where HR teams build AI agents for onboarding, employee questions, and approvals. Agents wire into your HRIS and ${INTEGRATION_COUNT_LABEL} integrations to keep people operations moving.`, + summary: `Sim is the open-source AI workspace where HR teams build, deploy, and manage AI agents for onboarding, employee questions, and approvals. Agents connect your HRIS and ${INTEGRATION_COUNT_LABEL} integrations so people operations keep moving.`, visual: ( diff --git a/apps/sim/app/(landing)/solutions/it/it.tsx b/apps/sim/app/(landing)/solutions/it/it.tsx index f7032917587..8132fc8388d 100644 --- a/apps/sim/app/(landing)/solutions/it/it.tsx +++ b/apps/sim/app/(landing)/solutions/it/it.tsx @@ -1,3 +1,4 @@ +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { PlatformHeroVisual, SolutionsPage, @@ -36,10 +37,8 @@ const IT_CONFIG: SolutionsPageConfig = { hero: { eyebrow: 'IT', heading: 'Automate ticket triage, access, and monitoring with AI agents in Sim.', - description: - 'Sim is the open-source AI workspace where IT teams build AI agents for ticket triage, access, and monitoring. Agents run with governance, access controls, and audit trails across hundreds of integrations.', - summary: - 'Sim is the open-source AI workspace where IT teams build, deploy, and manage AI agents for ticket triage, access provisioning, and infrastructure monitoring. Agents run with IT-grade governance and audit trails across hundreds of integrations and every major LLM.', + description: `Sim is the open-source AI workspace where IT teams build AI agents for ticket triage, access, and monitoring. Agents run with governance, access controls, and audit trails across ${INTEGRATION_COUNT_LABEL} integrations.`, + summary: `Sim is the open-source AI workspace where IT teams build, deploy, and manage AI agents for ticket triage, access provisioning, and infrastructure monitoring. Agents run with IT-grade governance and audit trails across ${INTEGRATION_COUNT_LABEL} integrations and every major LLM.`, visual: ( diff --git a/apps/sim/app/(landing)/solutions/sales/sales.tsx b/apps/sim/app/(landing)/solutions/sales/sales.tsx index b46a21647fd..cc476f00faf 100644 --- a/apps/sim/app/(landing)/solutions/sales/sales.tsx +++ b/apps/sim/app/(landing)/solutions/sales/sales.tsx @@ -1,3 +1,4 @@ +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { PlatformHeroVisual, SolutionsPage, @@ -38,10 +39,8 @@ const SALES_CONFIG: SolutionsPageConfig = { hero: { eyebrow: 'Sales', heading: 'Automate lead research, outreach, and CRM updates with AI agents in Sim.', - description: - 'Sim is the open-source AI workspace where sales teams build AI agents for lead research, outreach, and CRM updates. Agents wire into Salesforce, HubSpot, and hundreds of integrations to keep the pipeline current.', - summary: - 'Sim is the open-source AI workspace where sales teams build, deploy, and manage AI agents for lead research, personalized outreach, and CRM updates. Agents wire into Salesforce, HubSpot, and hundreds of integrations so the pipeline stays current.', + description: `Sim is the open-source AI workspace where sales teams build AI agents for lead research, outreach, and CRM updates. Agents wire into Salesforce, HubSpot, and ${INTEGRATION_COUNT_LABEL} integrations to keep the pipeline current.`, + summary: `Sim is the open-source AI workspace where sales teams build, deploy, and manage AI agents for lead research, personalized outreach, and CRM updates. Agents wire into Salesforce, HubSpot, and ${INTEGRATION_COUNT_LABEL} integrations so the pipeline stays current.`, visual: ( diff --git a/apps/sim/app/f/[token]/public-file-view.tsx b/apps/sim/app/f/[token]/public-file-view.tsx index 4749438df7f..4b57b43f56d 100644 --- a/apps/sim/app/f/[token]/public-file-view.tsx +++ b/apps/sim/app/f/[token]/public-file-view.tsx @@ -4,6 +4,7 @@ import { useMemo } from 'react' import { Chip, OverflowText, SimWordmark } from '@sim/emcn' import { Download } from '@sim/emcn/icons' import Link from 'next/link' +import { SITE_URL } from '@/lib/core/utils/urls' import type { WorkspaceFileRecord } from '@/lib/uploads/contexts/workspace' import { DesktopTitleBarLane } from '@/app/_shell/desktop-title-bar' import { buildProvenance } from '@/app/f/[token]/utils' @@ -72,7 +73,7 @@ export function PublicFileView({ {!brand.logoUrl && ( <> Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect 1,000+ integrations and every major LLM to create agents that automate real work. - -## Overview - -Sim is the AI workspace where teams create agents visually with the workflow builder, conversationally through Chat, or programmatically with the API. Over 100,000 builders use Sim — from startups to Fortune 500 companies. Teams connect their tools and data, build agents that automate real work across systems, and manage them with full observability. SOC2 compliant. - -## Product Details - -- **Product Name**: Sim -- **Category**: AI Workspace / AI Agent Builder -- **Deployment**: Cloud (SaaS) and Self-hosted options -- **Pricing**: Free tier, Pro ($25/month, 6K credits), Max ($100/month, 25K credits), Team plans available, Enterprise (custom) -- **Compliance**: SOC2 Type II - -## Core Concepts - -### Workspace -A workspace is the top-level container in Sim. It holds workflows, data sources, credentials, and execution history. Users can create multiple workspaces for different projects or teams. - -### Workflow -A workflow is a directed graph of blocks that defines an agentic process. Workflows can be triggered manually, on a schedule, or via webhooks. Each workflow has a unique ID and can be versioned. - -### Block -A block is an individual step in a workflow. Types include: -- **Agent Block**: Executes an LLM call with system prompts and tools -- **Function Block**: Runs custom JavaScript/TypeScript code -- **API Block**: Makes HTTP requests to external services -- **Condition Block**: Branches workflow based on conditions -- **Loop Block**: Iterates over arrays or until conditions are met -- **Router Block**: Routes to different paths based on LLM classification - -### Trigger -A trigger initiates workflow execution. Types include: -- **Manual**: User clicks "Run" button -- **Schedule**: Cron-based scheduling (e.g., every hour, daily at 9am) -- **Webhook**: HTTP endpoint that triggers on incoming requests -- **Event**: Triggered by external events (email received, Slack message, etc.) - -### Execution -An execution is a single run of a workflow. It includes: -- Input parameters -- Block-by-block execution logs -- Output data -- Token usage and cost tracking -- Duration and performance metrics - -## Capabilities - -### LLM Orchestration -Sim supports all major LLM providers: -- OpenAI (GPT-5.2, GPT-5.1, GPT-5, GPT-4o, GPT-4.1) -- Anthropic (Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4.5, Claude Haiku 4.5) -- Google (Gemini Pro 3, Gemini Pro 3 Preview, Gemini 2.5 Pro, Gemini 2.5 Flash) -- Mistral (Mistral Large, Mistral Medium) -- xAI (Grok) -- Perplexity -- Ollama or VLLM (self-hosted open-source models) -- Azure OpenAI -- Amazon Bedrock - -### Integrations -1,000+ pre-built integrations including: -- **Communication**: Slack, Discord, Email (Gmail, Outlook), SMS (Twilio) -- **Productivity**: Notion, Airtable, Google Sheets, Google Docs -- **Development**: GitHub, GitLab, Jira, Linear -- **Data**: PostgreSQL, MySQL, MongoDB, Supabase, Pinecone -- **Storage**: AWS S3, Google Cloud Storage, Dropbox -- **CRM**: Salesforce, HubSpot, Pipedrive - -### RAG (Retrieval-Augmented Generation) -Built-in support for: -- Document ingestion (PDF, DOCX, TXT, Markdown) -- Vector database integration (Pinecone, Weaviate, Qdrant) -- Semantic search and retrieval -- Chunking strategies (fixed size, semantic, recursive) - -### Tables -Built-in table creation and management: -- Structured data storage -- Queryable tables for agent workflows -- Native integrations - -### Code Execution -- Sandboxed JavaScript/TypeScript execution -- Access to npm packages -- Persistent state across executions -- Error handling and retry logic - -## Use Cases - -### Customer Support Automation -- Classify incoming tickets by urgency and topic -- Generate draft responses using RAG over knowledge base -- Route to appropriate team members -- Auto-close resolved tickets - -### Content Generation Pipeline -- Research topics using web search tools -- Generate outlines and drafts with LLMs -- Review and edit with human-in-the-loop -- Publish to CMS platforms - -### Data Processing Workflows -- Extract data from documents (invoices, receipts, forms) -- Transform and validate data -- Load into databases or spreadsheets -- Generate reports and summaries - -### Sales and Marketing Automation -- Enrich leads with company data -- Score leads based on fit criteria -- Generate personalized outreach emails -- Sync with CRM systems - -## Technical Architecture - -### Frontend -- Next.js 16 with App Router -- React Flow for the visual builder -- Tailwind CSS for styling -- Zustand for state management - -### Backend -- Node.js with TypeScript -- PostgreSQL for persistent storage -- Redis for caching and queues -- S3-compatible storage for files - -### Execution Engine -- Isolated execution per workflow run -- Parallel block execution where possible -- Retry logic with exponential backoff -- Real-time streaming of outputs - -## Getting Started - -1. **Sign Up**: Create a free account at ${baseUrl} -2. **Create Workspace**: Set up your first workspace -3. **Build Workflow**: Drag blocks onto the workflow builder and connect them -4. **Configure Blocks**: Set up LLM providers, tools, and integrations -5. **Test**: Run the workflow manually to verify -6. **Deploy**: Set up triggers for automated execution - -## Links - -- [Website](${baseUrl}): Product overview and primary entry point -- [Documentation](https://docs.sim.ai): Product guides and technical reference -- [API Reference](https://docs.sim.ai/api): API documentation -- [GitHub](https://github.com/simstudioai/sim): Open-source codebase -- [Slack](https://join.slack.com/t/sim-ott9864/shared_invite/zt-43lp8tc5v-0qrrqHGBKUsvQlpoouH~TA): Community workspace -- [X/Twitter](https://x.com/simdotai): Announcements and updates -- [LinkedIn](https://linkedin.com/company/simdotai): Company page +import { CREDIT_TIERS } from '@/lib/billing/constants' +import type { CompetitorProfile, Fact, Prose } from '@/lib/compare/data' +import { simProfile } from '@/lib/compare/data' +import { toSiteUrl } from '@/lib/core/utils/urls' +import { getAllCustomerStoryMeta, getCustomerStorySource } from '@/lib/customers/registry' +import { DOCS_URL } from '@/lib/help-links' +import { + getAllPostMeta as getAllLibraryPostMeta, + getPostSource as getLibraryPostSource, +} from '@/lib/library/registry' +import { COMPARISON_SECTIONS, getFactGroup } from '@/app/(landing)/comparisons/comparison-sections' +import { + ALL_COMPETITORS, + buildBottomLine, + getLatestVerifiedDate, + SIM_LATEST_VERIFIED, +} from '@/app/(landing)/comparisons/utils' +import { PLATFORM_MENU } from '@/app/(landing)/components/navbar/components/nav-menu-chip' +import { + LLMS_HEADER, + linkLine, + markdownResponse, + navMenuLines, + SOLUTION_LINES, + section, + toLlmsMarkdown, +} from '@/app/llms.txt/llms' + +export const dynamic = 'force-static' +export const revalidate = 86400 + +const CONCEPTS = [ + [ + 'Workspace', + 'The container for a team’s agents, workflows, knowledge bases, tables, files, credentials, and run history.', + ], + [ + 'Chat', + 'Talk to Sim in natural language to build, run, and manage everything in the workspace.', + ], + [ + 'Workflow', + 'The agent logic built in the visual builder: blocks connected into a graph that runs from a trigger.', + ], + [ + 'Block', + 'One step in a workflow, such as an Agent (LLM call with tools), Function (code), API request, Condition, Router, Loop, or Parallel.', + ], + [ + 'Trigger', + 'What starts a run: a manual run, a schedule, a webhook, an API call, a chat message, or an event in a connected app.', + ], + [ + 'Knowledge Base', + 'Documents uploaded or synced from sources such as Notion, Google Drive, and Confluence, searchable by agents.', + ], + ['Tables', 'A built-in database agents read and update while they work.'], + ['Logs', 'Every run traced block by block, with inputs, outputs, cost, and duration.'], +] as const + +function proseToMarkdown(prose: Prose): string { + return prose + .map((segment) => + typeof segment === 'string' ? segment : `[${segment.text}](${toSiteUrl(segment.href)})` + ) + .join('') +} -## Support +/** A fact as a table cell: its compact form, as the comparison table renders it. */ +function factCell(fact: Fact | undefined): string { + const value = fact ? (fact.shortValue ?? fact.value) : 'Unknown' + return value.replace(/\|/g, '\\|').replace(/\s*\n\s*/g, ' ') +} -- [Documentation](https://docs.sim.ai): Self-serve guides and reference -- [Community Slack](https://join.slack.com/t/sim-ott9864/shared_invite/zt-43lp8tc5v-0qrrqHGBKUsvQlpoouH~TA): Community support -- Email: help@sim.ai -- Security issues: security@sim.ai +function isoDate(date: Date): string { + return date.toISOString().slice(0, 10) +} -## Legal +/** The key facts of one `/comparisons/{id}` page as markdown, from the same profile data. */ +function comparisonMarkdown(competitor: CompetitorProfile): string { + const verdict = buildBottomLine(competitor) + const verified = new Date( + Math.max(SIM_LATEST_VERIFIED.getTime(), getLatestVerifiedDate(competitor).getTime()) + ) + const rows = COMPARISON_SECTIONS.flatMap((s) => { + const sim = getFactGroup(simProfile, s.group) + const other = getFactGroup(competitor, s.group) + return s.rows.map( + (row) => + `| ${s.title}: ${row.label} | ${factCell(sim[row.key])} | ${factCell(other[row.key])} |` + ) + }) -- [Terms of Service](${baseUrl}/terms): Legal terms -- [Privacy Policy](${baseUrl}/privacy): Data handling practices -- [Cookie Policy](${baseUrl}/cookie-policy): Cookies Sim sets, why, and how to change your choice -- [Security](${baseUrl}/.well-known/security.txt): Vulnerability disclosure policy -` + return [ + `### Sim vs ${competitor.name}`, + `URL: ${toSiteUrl(`/comparisons/${competitor.id}`)} · Facts last verified ${isoDate(verified)}`, + `${competitor.name}: ${competitor.oneLiner}`, + competitor.leadAnswer ? proseToMarkdown(competitor.leadAnswer) : '', + competitor.betterThanAnswer ? proseToMarkdown(competitor.betterThanAnswer) : '', + `- ${verdict.chooseSim}\n- ${verdict.chooseCompetitor}`, + competitor.standoutFeatures.length > 0 + ? `Standout features of ${competitor.name}:\n\n${competitor.standoutFeatures.map((f) => `- ${f.title}: ${f.description}`).join('\n')}` + : '', + competitor.limitations.length > 0 + ? `Documented limitations of ${competitor.name}:\n\n${competitor.limitations.map((l) => `- ${l.title}: ${l.description}`).join('\n')}` + : '', + [`| Feature | Sim | ${competitor.name} |`, '| --- | --- | --- |', ...rows].join('\n'), + ] + .filter(Boolean) + .join('\n\n') +} - return new Response(llmsFullContent, { - headers: { - 'Content-Type': 'text/markdown; charset=utf-8', - 'Cache-Control': 'public, max-age=86400, s-maxage=86400', - }, - }) +/** + * `/llms-full.txt`: the substantive public content in one markdown file for AI + * engines to ingest — product overview, the full text of every published + * library article and customer story, and the sourced facts behind every + * comparison page. Generated from the content registries and comparison data, + * so retired articles and new comparisons track automatically. + */ +export async function GET() { + const [libraryPosts, customerStories] = await Promise.all([ + getAllLibraryPostMeta(), + getAllCustomerStoryMeta(), + ]) + const toBody = (source: string | null) => toLlmsMarkdown(source ?? '', 2) + const [libraryBodies, customerBodies] = await Promise.all([ + Promise.all(libraryPosts.map((p) => getLibraryPostSource(p.slug).then(toBody))), + Promise.all(customerStories.map((s) => getCustomerStorySource(s.slug).then(toBody))), + ]) + + const [pro, max] = CREDIT_TIERS + + return markdownResponse( + [ + LLMS_HEADER, + `This file holds the full text of Sim’s public library, customer stories, and comparison facts. The link index is [llms.txt](${toSiteUrl('/llms.txt')}); the product documentation is at [${DOCS_URL}/llms-full.txt](${DOCS_URL}/llms-full.txt).`, + section('Overview', [ + 'Teams build agents in the visual workflow builder, by talking to Sim in Chat, or with code through the API and SDKs. Sim is open source under the Apache 2.0 license and runs as a managed cloud service or self-hosted with Docker or Kubernetes.', + '', + ...CONCEPTS.map(([term, definition]) => `- **${term}**: ${definition}`), + ]), + section('Platform', navMenuLines(PLATFORM_MENU)), + section('Pricing', [ + linkLine('Pricing', '/pricing'), + '- Free: $0 to start building agents.', + `- ${pro.name}: $${pro.dollars} per user per month, ${pro.credits.toLocaleString('en-US')} credits.`, + `- ${max.name}: $${max.dollars} per user per month, ${max.credits.toLocaleString('en-US')} credits.`, + '- Enterprise: custom limits, infrastructure, and governance for large organizations.', + ]), + section('Solutions', SOLUTION_LINES), + section( + 'Customer stories', + customerStories.map((story, i) => + [`### ${story.title}`, `URL: ${story.canonical}`, customerBodies[i]].join('\n\n') + ), + '\n\n' + ), + section('Comparisons', ALL_COMPETITORS.map(comparisonMarkdown), '\n\n'), + section( + 'Library', + libraryPosts.map((p, i) => + [ + `### ${p.title}`, + `URL: ${p.canonical} · Updated ${(p.updated ?? p.date).slice(0, 10)}`, + `> ${p.description}`, + libraryBodies[i], + ].join('\n\n') + ), + '\n\n' + ), + ], + revalidate + ) } diff --git a/apps/sim/app/llms.txt/llms.ts b/apps/sim/app/llms.txt/llms.ts new file mode 100644 index 00000000000..c8b28574bfa --- /dev/null +++ b/apps/sim/app/llms.txt/llms.ts @@ -0,0 +1,98 @@ +import { SITE_URL, toSiteUrl } from '@/lib/core/utils/urls' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' +import type { NavMenu } from '@/app/(landing)/components/navbar/components/nav-menu-chip' +import { COMPLIANCE_PAGE_DESCRIPTION } from '@/app/(landing)/solutions/compliance/compliance' +import { ENGINEERING_PAGE_DESCRIPTION } from '@/app/(landing)/solutions/engineering/engineering' +import { FINANCE_PAGE_DESCRIPTION } from '@/app/(landing)/solutions/finance/finance' +import { HR_PAGE_DESCRIPTION } from '@/app/(landing)/solutions/hr/hr' +import { IT_PAGE_DESCRIPTION } from '@/app/(landing)/solutions/it/it' +import { SALES_PAGE_DESCRIPTION } from '@/app/(landing)/solutions/sales/sales' + +/** Shared building blocks for `/llms.txt` and `/llms-full.txt` (https://llmstxt.org). */ + +const SIM_SUMMARY = `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect ${INTEGRATION_COUNT_LABEL} integrations and every major LLM to create agents that automate real work — visually, conversationally, or with code.` + +/** The H1 and blockquote summary every llms.txt variant opens with. */ +export const LLMS_HEADER = `# Sim\n\n> ${SIM_SUMMARY}` + +const SOLUTIONS = [ + { + title: 'AI agents for compliance', + path: '/solutions/compliance', + description: COMPLIANCE_PAGE_DESCRIPTION, + }, + { + title: 'AI agents for engineering', + path: '/solutions/engineering', + description: ENGINEERING_PAGE_DESCRIPTION, + }, + { + title: 'AI agents for finance', + path: '/solutions/finance', + description: FINANCE_PAGE_DESCRIPTION, + }, + { title: 'AI agents for HR', path: '/solutions/hr', description: HR_PAGE_DESCRIPTION }, + { title: 'AI agents for IT', path: '/solutions/it', description: IT_PAGE_DESCRIPTION }, + { title: 'AI agents for sales', path: '/solutions/sales', description: SALES_PAGE_DESCRIPTION }, +] as const + +/** One llms.txt entry per `/solutions/*` page. */ +export const SOLUTION_LINES = SOLUTIONS.map((s) => linkLine(s.title, s.path, s.description)) + +/** One llms.txt list entry: `- [title](url): description`. */ +export function linkLine(title: string, href: string, description?: string): string { + return `- [${title}](${toSiteUrl(href)})${description ? `: ${description}` : ''}` +} + +/** + * A `## heading` followed by its entries, or nothing when there are none. + * List lines join with a newline; pass `'\n\n'` for multi-paragraph blocks. + */ +export function section(heading: string, entries: readonly string[], separator = '\n'): string { + return entries.length > 0 ? `## ${heading}\n\n${entries.join(separator)}` : '' +} + +/** Every item of a navbar menu as an llms.txt link line. */ +export function navMenuLines(menu: NavMenu): string[] { + return menu.sections.flatMap((s) => + s.items.map((item) => + linkLine(item.brand ? `${item.brand} ${item.title}` : item.title, item.href, item.description) + ) + ) +} + +/** + * Prepares a post's raw markdown for llms-full.txt: headings demoted by + * `demoteBy` levels so they nest under the caller's title, and site-relative + * links and images made absolute. + */ +export function toLlmsMarkdown(source: string, demoteBy: number): string { + let inFence = false + return source + .split('\n') + .map((line) => { + if (/^\s*(```|~~~)/.test(line)) { + inFence = !inFence + return line + } + if (inFence) return line + return line + .replace( + /^(#{1,6}) /, + (_, hashes: string) => `${'#'.repeat(Math.min(6, hashes.length + demoteBy))} ` + ) + .replace(/\]\(\/(?!\/)/g, `](${SITE_URL}/`) + }) + .join('\n') + .trim() +} + +/** Joins non-empty blocks into one markdown document and serves it with shared cache headers. */ +export function markdownResponse(blocks: readonly string[], revalidateSeconds: number): Response { + return new Response(`${blocks.filter(Boolean).join('\n\n')}\n`, { + headers: { + 'Content-Type': 'text/markdown; charset=utf-8', + 'Cache-Control': `public, s-maxage=${revalidateSeconds}, stale-while-revalidate=${revalidateSeconds}`, + }, + }) +} diff --git a/apps/sim/app/llms.txt/route.ts b/apps/sim/app/llms.txt/route.ts index d368cfe8668..b0068d8cdec 100644 --- a/apps/sim/app/llms.txt/route.ts +++ b/apps/sim/app/llms.txt/route.ts @@ -1,69 +1,119 @@ -import { getBaseUrl } from '@/lib/core/utils/urls' - -export function GET() { - const baseUrl = getBaseUrl() - - const content = `# Sim - -> Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect 1,000+ integrations and every major LLM to create agents that automate real work. - -Sim lets teams create agents visually with the workflow builder, conversationally through Chat, or programmatically with the API. The workspace includes knowledge bases, tables, files, and full observability. - -## Preferred URLs - -- [Homepage](${baseUrl}): Product overview and primary entry point -- [Integrations directory](${baseUrl}/integrations): Public catalog of integrations and automation capabilities -- [Models directory](${baseUrl}/models): Public catalog of AI models, pricing, context windows, and capabilities -- [Blog](${baseUrl}/blog): Announcements, guides, and product context -- [Changelog](${baseUrl}/changelog): Product updates and release notes - -## Documentation - -- [Documentation](https://docs.sim.ai): Product guides and technical reference -- [Quickstart](https://docs.sim.ai/getting-started): Fastest path to getting started -- [API Reference](https://docs.sim.ai/api-reference): API documentation - -## Key Concepts - -- **Workspace**: The AI workspace — container for agents, workflows, data sources, and runs -- **Workflow**: Visual builder — directed graph of blocks defining agent logic -- **Block**: Individual step such as an LLM call, tool call, HTTP request, or code execution -- **Trigger**: Event or schedule that initiates a workflow run -- **Execution**: A single run of a workflow with logs and outputs -- **Knowledge Base**: Document store used for retrieval-augmented generation - -## Capabilities - -- AI workspace for teams -- AI agent creation and deployment -- Integrations across business tools, databases, and communication platforms -- Multi-model LLM orchestration -- Knowledge bases and retrieval-augmented generation -- Table creation and management -- Document creation and processing -- Scheduled and webhook-triggered runs - -## Use Cases - -- AI agent deployment and orchestration -- Knowledge bases and RAG pipelines -- Customer support automation -- Internal operations workflows across sales, marketing, legal, and finance - -## Additional Links - -- [GitHub Repository](https://github.com/simstudioai/sim): Open-source codebase -- [Docs](https://docs.sim.ai): Canonical documentation source -- [Terms of Service](${baseUrl}/terms): Legal terms -- [Privacy Policy](${baseUrl}/privacy): Data handling practices -- [Cookie Policy](${baseUrl}/cookie-policy): Cookies Sim sets, why, and how to change your choice -- [Sitemap](${baseUrl}/sitemap.xml): Public URL inventory -` - - return new Response(content, { - headers: { - 'Content-Type': 'text/markdown; charset=utf-8', - 'Cache-Control': 'public, max-age=86400, s-maxage=86400', - }, - }) +import { getAllPostMeta as getAllBlogPostMeta } from '@/lib/blog/registry' +import { toSiteUrl } from '@/lib/core/utils/urls' +import { getAllCustomerStoryMeta } from '@/lib/customers/registry' +import { DOCS_URL, SLACK_COMMUNITY_URL } from '@/lib/help-links' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' +import { getAllPostMeta as getAllLibraryPostMeta } from '@/lib/library/registry' +import { ALL_COMPETITORS } from '@/app/(landing)/comparisons/utils' +import { PLATFORM_MENU } from '@/app/(landing)/components/navbar/components/nav-menu-chip' +import { MODEL_PROVIDERS_WITH_CATALOGS } from '@/app/(landing)/models/utils' +import { + LLMS_HEADER, + linkLine, + markdownResponse, + navMenuLines, + SOLUTION_LINES, + section, +} from '@/app/llms.txt/llms' + +export const dynamic = 'force-static' +export const revalidate = 86400 + +/** + * `/llms.txt` per https://llmstxt.org: a curated, link-first index of the + * public site. Content sections are generated from the same registries the + * sitemap reads, so new and retired pages track automatically. Individual + * integration and model pages are left to their hubs and the sitemap. + */ +export async function GET() { + const [blogPosts, libraryPosts, customerStories] = await Promise.all([ + getAllBlogPostMeta(), + getAllLibraryPostMeta(), + getAllCustomerStoryMeta(), + ]) + + return markdownResponse( + [ + LLMS_HEADER, + 'Teams build agents in the visual workflow builder, by talking to Sim in Chat, or with code through the API and SDKs. The workspace includes knowledge bases, tables, files, and logs for every run. Sim is open source (Apache 2.0) and runs in the cloud or self-hosted.', + `The full text of the library, customer stories, and comparison facts is in [llms-full.txt](${toSiteUrl('/llms-full.txt')}).`, + section('Platform', [ + linkLine('Home', '/', 'Product overview and primary entry point'), + ...navMenuLines(PLATFORM_MENU), + linkLine('Pricing', '/pricing', 'Free, Pro, Max, and Enterprise plans'), + ]), + section('Solutions', SOLUTION_LINES), + section( + 'Customers', + customerStories.length > 0 + ? [ + linkLine( + 'Customer stories', + '/customers', + 'How teams build and run AI agents with Sim' + ), + ...customerStories.map((story) => + linkLine(story.title, story.canonical, story.description) + ), + ] + : [] + ), + section('Comparisons', [ + linkLine( + 'All comparisons', + '/comparisons', + 'Sourced, dated comparisons of Sim with AI agent and workflow automation platforms' + ), + ...ALL_COMPETITORS.map((c) => + linkLine(`Sim vs ${c.name}`, `/comparisons/${c.id}`, c.oneLiner) + ), + ]), + section('Library', [ + linkLine('Library', '/library', 'Comparisons, how-tos, and roundups on building AI agents'), + ...libraryPosts.map((p) => linkLine(p.title, p.canonical, p.description)), + ]), + section('Integrations and models', [ + linkLine( + 'Integrations', + '/integrations', + `${INTEGRATION_COUNT_LABEL} integrations, triggers, and tools agents can use` + ), + linkLine( + 'Models', + '/models', + 'Every supported model with pricing, context window, and capabilities' + ), + ...MODEL_PROVIDERS_WITH_CATALOGS.map((provider) => + linkLine(`${provider.name} models`, provider.href, provider.description) + ), + ]), + section('Docs', [ + linkLine('Docs index', `${DOCS_URL}/llms.txt`, 'llms.txt index of the Sim documentation'), + linkLine( + 'Docs full text', + `${DOCS_URL}/llms-full.txt`, + 'Full text of the Sim documentation' + ), + linkLine('Documentation', DOCS_URL, 'Guides, SDKs, and API reference'), + ]), + section('Blog', [ + linkLine('Blog', '/blog', 'Announcements, engineering deep dives, and product context'), + ...blogPosts.map((p) => linkLine(p.title, p.canonical, p.description)), + ]), + section('Optional', [ + linkLine('Changelog', '/changelog', 'Product updates and release notes'), + linkLine('GitHub', 'https://github.com/simstudioai/sim', 'Open-source codebase'), + linkLine('Community Slack', SLACK_COMMUNITY_URL, 'Community workspace'), + linkLine('Terms of Service', '/terms'), + linkLine('Privacy Policy', '/privacy'), + linkLine('Cookie Policy', '/cookie-policy'), + linkLine( + 'Sitemap', + '/sitemap.xml', + 'Every public URL, including each integration and model' + ), + ]), + ], + revalidate + ) } diff --git a/apps/sim/app/manifest.ts b/apps/sim/app/manifest.ts index a0e5f077e0c..248053a42ab 100644 --- a/apps/sim/app/manifest.ts +++ b/apps/sim/app/manifest.ts @@ -1,4 +1,5 @@ import type { MetadataRoute } from 'next' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { WORKSPACES_PATH } from '@/lib/navigation/paths' import { getBrandConfig } from '@/ee/whitelabeling' @@ -13,8 +14,7 @@ export default function manifest(): MetadataRoute.Manifest { ? 'Sim — The AI Workspace | Build, Deploy & Manage AI Agents' : brand.name, short_name: brand.name, - description: - 'Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect 1,000+ integrations and every major LLM.', + description: `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect ${INTEGRATION_COUNT_LABEL} integrations and every major LLM.`, start_url: '/', scope: '/', display: 'standalone', diff --git a/apps/sim/app/robots.ts b/apps/sim/app/robots.ts index 291cf1c8a46..377fa3c806f 100644 --- a/apps/sim/app/robots.ts +++ b/apps/sim/app/robots.ts @@ -1,21 +1,16 @@ import type { MetadataRoute } from 'next' import { SITE_URL } from '@/lib/core/utils/urls' -const DISALLOWED_PATHS = [ - '/api/', - '/workspace/', - '/playground/', - '/resume/', - '/invite/', - '/unsubscribe/', - '/w/', - '/_next/', - '/private/', -] - +/** + * Only `/api/` is blocked from crawling. App and utility surfaces stay + * crawlable so search engines can see the `X-Robots-Tag: noindex` the proxy + * sends on them (a disallowed URL can still be indexed from external links), + * and `/_next/` stays crawlable so pages render with their scripts, styles, + * and images. + */ export default function robots(): MetadataRoute.Robots { return { - rules: { userAgent: '*', allow: '/', disallow: DISALLOWED_PATHS }, + rules: { userAgent: '*', allow: '/', disallow: ['/api/'] }, sitemap: [ `${SITE_URL}/sitemap.xml`, `${SITE_URL}/blog/sitemap-images.xml`, diff --git a/apps/sim/app/sitemap.ts b/apps/sim/app/sitemap.ts index bf5f33e639e..cc2871c811c 100644 --- a/apps/sim/app/sitemap.ts +++ b/apps/sim/app/sitemap.ts @@ -131,7 +131,6 @@ export default async function sitemap(): Promise { }, { url: `${baseUrl}/changelog`, - lastModified: latestPostDateValue, }, { url: `${baseUrl}/integrations`, diff --git a/apps/sim/content/blog/agent-as-yjs-peer/index.mdx b/apps/sim/content/blog/agent-as-yjs-peer/index.mdx index 3383fb25a43..de578c988a8 100644 --- a/apps/sim/content/blog/agent-as-yjs-peer/index.mdx +++ b/apps/sim/content/blog/agent-as-yjs-peer/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 11 tags: [Yjs, CRDT, Collaboration, AI, ProseMirror, TipTap, Streaming, Architecture] ogImage: /blog/agent-as-yjs-peer/cover.jpg -canonical: https://www.sim.ai/blog/agent-as-yjs-peer draft: false featured: true faq: diff --git a/apps/sim/content/blog/copilot/index.mdx b/apps/sim/content/blog/copilot/index.mdx index 7ea345209ed..b0cbc82d079 100644 --- a/apps/sim/content/blog/copilot/index.mdx +++ b/apps/sim/content/blog/copilot/index.mdx @@ -12,7 +12,6 @@ ogImage: /blog/copilot/cover.png ogAlt: 'Sim Copilot technical overview' about: ['AI Assistants', 'Agentic Workflows', 'Retrieval Augmented Generation'] timeRequired: PT7M -canonical: https://www.sim.ai/blog/copilot featured: false draft: true faq: diff --git a/apps/sim/content/blog/emcn/index.mdx b/apps/sim/content/blog/emcn/index.mdx index bb0366d605d..f0fb3d26df8 100644 --- a/apps/sim/content/blog/emcn/index.mdx +++ b/apps/sim/content/blog/emcn/index.mdx @@ -12,7 +12,6 @@ ogImage: /blog/emcn/cover.png ogAlt: 'Emcn design system cover' about: ['Design Systems', 'Component Libraries', 'Design Tokens', 'Accessibility'] timeRequired: PT6M -canonical: https://www.sim.ai/blog/emcn featured: false draft: true faq: diff --git a/apps/sim/content/blog/enterprise/index.mdx b/apps/sim/content/blog/enterprise/index.mdx index d9a1869ba0b..da34c418e26 100644 --- a/apps/sim/content/blog/enterprise/index.mdx +++ b/apps/sim/content/blog/enterprise/index.mdx @@ -12,7 +12,6 @@ ogImage: /blog/enterprise/cover.jpg ogAlt: 'Sim Enterprise features overview' about: ['Enterprise Software', 'Security', 'Compliance', 'Self-Hosting'] timeRequired: PT6M -canonical: https://www.sim.ai/blog/enterprise featured: true draft: false faq: @@ -189,7 +188,7 @@ For teams practicing GitOps, export workflows to your repository and use the Adm ## Get Started -Enterprise features are available now. Check out our [self-hosting](https://docs.sim.ai/platform/self-hosting) and [enterprise](https://docs.sim.ai/platform/enterprise) docs to get started. Teams comparing deployment options can also use our guides to [enterprise AI agent platforms](https://www.sim.ai/library/best-ai-agent-platforms-for-enterprise-teams-2026), the [AI workflow automation buyer's checklist](https://www.sim.ai/library/ai-workflow-automation-platform-buyers-checklist), and [AI agents in procurement](https://www.sim.ai/library/ai-agents-in-procurement). +Enterprise features are available now. Check out our [self-hosting](https://docs.sim.ai/platform/self-hosting) and [enterprise](https://docs.sim.ai/platform/enterprise) docs to get started. Teams comparing deployment options can also use our guides to [enterprise AI agent platforms](/library/best-ai-agent-platforms-for-enterprise-teams-2026), the [AI workflow automation buyer's checklist](/library/ai-workflow-automation-platform-buyers-checklist), and [AI agents in procurement](/library/ai-agents-in-procurement). *Questions about enterprise deployments?* diff --git a/apps/sim/content/blog/executor/index.mdx b/apps/sim/content/blog/executor/index.mdx index 5541c73f2a4..8047a8d644a 100644 --- a/apps/sim/content/blog/executor/index.mdx +++ b/apps/sim/content/blog/executor/index.mdx @@ -12,7 +12,6 @@ ogImage: /blog/executor/cover.jpg ogAlt: 'Sim Executor technical overview' about: ['Execution', 'Workflow Orchestration'] timeRequired: PT12M -canonical: https://www.sim.ai/blog/executor featured: false draft: false faq: diff --git a/apps/sim/content/blog/mothership/index.mdx b/apps/sim/content/blog/mothership/index.mdx index 937f2846e95..e53607fa8e0 100644 --- a/apps/sim/content/blog/mothership/index.mdx +++ b/apps/sim/content/blog/mothership/index.mdx @@ -12,7 +12,6 @@ ogImage: /blog/mothership/cover.jpg ogAlt: 'Introducing Mothership airship illustration' about: ['AI Agents', 'Workflow Automation', 'Developer Tools'] timeRequired: PT10M -canonical: https://www.sim.ai/blog/mothership featured: true draft: false faq: @@ -148,6 +147,6 @@ That's what v0.6 is. ## Get Started -Sim v0.6 is available now at [sim.ai](https://sim.ai). Check out our [documentation](https://docs.sim.ai) for detailed guides on Mothership, Tables, Connectors, and more. +Sim v0.6 is available now at [sim.ai](/). Check out our [documentation](https://docs.sim.ai) for detailed guides on Mothership, Tables, Connectors, and more. -*Questions? [help@sim.ai](mailto:help@sim.ai) · [Slack](https://sim.ai/slack)* +*Questions? [help@sim.ai](mailto:help@sim.ai) · [Slack](/slack)* diff --git a/apps/sim/content/blog/multiplayer/index.mdx b/apps/sim/content/blog/multiplayer/index.mdx index f45096bc7a5..b982fe1bb0f 100644 --- a/apps/sim/content/blog/multiplayer/index.mdx +++ b/apps/sim/content/blog/multiplayer/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 12 tags: [Multiplayer, Realtime, Collaboration, WebSockets, Architecture] ogImage: /blog/multiplayer/cover.jpg -canonical: https://www.sim.ai/blog/multiplayer draft: false faq: - q: "Does Sim use CRDTs or operational transforms for multiplayer editing?" @@ -191,4 +190,4 @@ Multiplayer workflow building is no longer a technical curiosity—it's how team --- -*Interested in how Sim's multiplayer system works in practice? [Try building a workflow](https://sim.ai) with a collaborator in real-time.* +*Interested in how Sim's multiplayer system works in practice? [Try building a workflow](/) with a collaborator in real-time.* diff --git a/apps/sim/content/blog/secret-provenance/index.mdx b/apps/sim/content/blog/secret-provenance/index.mdx index d09961f08e2..7bc5a49f0f3 100644 --- a/apps/sim/content/blog/secret-provenance/index.mdx +++ b/apps/sim/content/blog/secret-provenance/index.mdx @@ -12,7 +12,6 @@ ogImage: /blog/secret-provenance/cover.jpg ogAlt: 'Sim secret provenance technical overview' about: ['Security', 'Execution'] timeRequired: PT7M -canonical: https://www.sim.ai/blog/secret-provenance featured: true draft: false faq: diff --git a/apps/sim/content/blog/series-a/index.mdx b/apps/sim/content/blog/series-a/index.mdx index 85e7d6d272e..927de89dd87 100644 --- a/apps/sim/content/blog/series-a/index.mdx +++ b/apps/sim/content/blog/series-a/index.mdx @@ -13,7 +13,6 @@ ogImage: /blog/series-a/cover.jpg ogAlt: 'Sim team photo in front of neon logo' about: ['Artificial Intelligence', 'Agentic Workflows', 'Startups', 'Funding'] timeRequired: PT4M -canonical: https://www.sim.ai/blog/series-a featured: true draft: false technical: false @@ -70,7 +69,7 @@ We’ll invest in building the community around Sim, and we'll continue to be re ## We’re hiring -If you’re excited about agentic systems and want to help define the future of this space, we’d love to talk. We’re hiring across engineering, engineering, and more engineering. Oh, and design. [Apply here](https://sim.ai/careers) +If you’re excited about agentic systems and want to help define the future of this space, we’d love to talk. We’re hiring across engineering, engineering, and more engineering. Oh, and design. [Apply here](/careers) — Team Sim diff --git a/apps/sim/content/blog/v0-5/index.mdx b/apps/sim/content/blog/v0-5/index.mdx index e2297417528..a13d9fa8f69 100644 --- a/apps/sim/content/blog/v0-5/index.mdx +++ b/apps/sim/content/blog/v0-5/index.mdx @@ -12,7 +12,6 @@ ogImage: /blog/v0-5/cover.jpg ogAlt: 'Sim v0.5 release announcement' about: ['AI Agents', 'Workflow Automation', 'Developer Tools'] timeRequired: PT8M -canonical: https://www.sim.ai/blog/v0-5 featured: true draft: false faq: @@ -212,6 +211,6 @@ Model selection is per-block, so you can use faster/cheaper models for simple ta ## Get Started -Available now at [sim.ai](https://sim.ai). Check out the [docs](https://docs.sim.ai) to dive deeper. +Available now at [sim.ai](/). Check out the [docs](https://docs.sim.ai) to dive deeper. -*Questions? [help@sim.ai](mailto:help@sim.ai) · [Slack](https://sim.ai/slack)* +*Questions? [help@sim.ai](mailto:help@sim.ai) · [Slack](/slack)* diff --git a/apps/sim/content/customers/exp-realty/index.mdx b/apps/sim/content/customers/exp-realty/index.mdx index 85f35d1ba36..d79563fd8de 100644 --- a/apps/sim/content/customers/exp-realty/index.mdx +++ b/apps/sim/content/customers/exp-realty/index.mdx @@ -7,7 +7,6 @@ authors: [sim] tags: [AI workflows, Real estate] ogImage: /landing/customers/exp-beach-house.jpg ogAlt: A modern beach house overlooking the ocean -canonical: https://www.sim.ai/customers/exp-realty draft: true technical: false --- diff --git a/apps/sim/content/customers/rivian/index.mdx b/apps/sim/content/customers/rivian/index.mdx index 3f43b7d464e..33550acb7f3 100644 --- a/apps/sim/content/customers/rivian/index.mdx +++ b/apps/sim/content/customers/rivian/index.mdx @@ -7,7 +7,6 @@ authors: [sim] tags: [Enterprise AI, Governance] ogImage: /landing/customers/rivian-trail.jpg ogAlt: A Rivian on a winding trail through a mountain landscape -canonical: https://www.sim.ai/customers/rivian draft: true technical: false --- diff --git a/apps/sim/content/library/6-best-ai-observability-tools-for-production-agents-in-2026/index.mdx b/apps/sim/content/library/6-best-ai-observability-tools-for-production-agents-in-2026/index.mdx index c03b02520a9..922fa2000d2 100644 --- a/apps/sim/content/library/6-best-ai-observability-tools-for-production-agents-in-2026/index.mdx +++ b/apps/sim/content/library/6-best-ai-observability-tools-for-production-agents-in-2026/index.mdx @@ -3,19 +3,14 @@ slug: 6-best-ai-observability-tools-for-production-agents-in-2026 title: '6 Best AI Observability Tools for Production Agents in 2026' description: 'Compare the six best AI observability tools for production agents in 2026 — Braintrust, Galileo, Langfuse, Arize AX, Datadog, and PostHog — across tracing, evaluations, CI/CD checks, and developer access.' date: 2026-09-04 -updated: 2026-09-04 +updated: 2026-09-30 authors: - andrew readingTime: 15 tags: [AI Observability, AI Agents, Evaluations, Developer Tools, Sim] ogImage: /library/6-best-ai-observability-tools-for-production-agents-in-2026/cover.jpg -canonical: https://www.sim.ai/library/6-best-ai-observability-tools-for-production-agents-in-2026 draft: false faq: - - q: "What is AI observability?" - a: "AI observability is the practice of capturing and analyzing the behavior of AI applications and agents. It covers traces, prompts, model calls, retrieval, tool use, outputs, latency, errors, tokens, cost, and quality evaluations. Its goal is to explain what an AI system did, assess whether the result was good, and provide evidence for improving it." - - q: "How is AI observability different from traditional APM?" - a: "Traditional APM focuses on operational health, including uptime, latency, errors, and infrastructure. AI observability adds the context needed to understand nondeterministic behavior, such as prompts, responses, retrieved context, tool choices, and output-quality scores. A healthy service can still produce a poor AI result." - q: "Do AI agents need evaluations as well as traces?" a: "Yes. Traces reconstruct the path an agent took, but they do not automatically determine whether the path or result was correct. Evaluations measure dimensions such as factuality, relevance, task completion, safety, and tool selection. Used together, traces explain failures and evaluations detect them at scale." - q: "What is the best AI observability tool in 2026?" @@ -28,7 +23,7 @@ Production agents rarely fail in one clean place. A request may cross a visual w That is why agent teams need observability at two levels. The agent builder should expose what happened inside each workflow run. A dedicated AI observability platform should make it easy to analyze behavior across applications, score real outputs, test changes, and stop known failures from returning. -At Sim, we approach the first level through native Logs. Every workflow run records block-level inputs, outputs, timing, errors, token usage, and cost. The tools in this guide address the broader quality workflow around those runs. For more background on traces, spans, metrics, and evaluations, read our guide to [AI agent observability](https://www.sim.ai/library/ai-agent-observability). +At Sim, we approach the first level through native Logs. Every workflow run records block-level inputs, outputs, timing, errors, token usage, and cost. The tools in this guide address the broader quality workflow around those runs. This is a buying guide, not a primer: if you are new to the concept, start with [what AI agent observability is](https://www.sim.ai/library/ai-agent-observability), which covers traces, spans, metrics, evaluations, and what to instrument at each stage. We compared six platforms with extra weight on how easy they make the everyday work: instrumenting an agent, reading a trace, running an evaluation, adding checks to CI/CD, and giving engineers or coding agents direct access to the data. Braintrust is our top recommendation because those pieces work as one short feedback loop. The alternatives are stronger fits when a team prioritizes specialized evaluators, self-hosting, enterprise monitoring, an existing APM stack, or product analytics context. @@ -66,7 +61,7 @@ We also considered deployment options, alerting, and how well each platform fits ## Where Sim fits: Observability inside the agent builder -An [AI agent builder](https://www.sim.ai/library/best-ai-agent-builder-2026) should not make teams assemble a separate telemetry stack before they can understand a run. Sim records the workflow as it executes, so builders can open a run and inspect the inputs, outputs, duration, errors, token usage, and cost for each block. Because the log follows the workflow graph, the trace uses the same mental model as the system the team designed. +An [AI agent builder](https://www.sim.ai/library/best-ai-agent-platforms-2026) should not make teams assemble a separate telemetry stack before they can understand a run. Sim records the workflow as it executes, so builders can open a run and inspect the inputs, outputs, duration, errors, token usage, and cost for each block. Because the log follows the workflow graph, the trace uses the same mental model as the system the team designed. That is especially useful when a workflow combines deterministic blocks with AI decisions. A team can see whether the failure came from the model, a tool call, a branch, an API, or the data passed between blocks. Sim also preserves the workflow state associated with the run, which helps distinguish a model-quality problem from a workflow-version problem. diff --git a/apps/sim/content/library/aeo-vs-geo-what-answer-engine-and-generative-engine-optimization-actually-mean/index.mdx b/apps/sim/content/library/aeo-vs-geo-what-answer-engine-and-generative-engine-optimization-actually-mean/index.mdx index 10577eb034b..a50e1453bc0 100644 --- a/apps/sim/content/library/aeo-vs-geo-what-answer-engine-and-generative-engine-optimization-actually-mean/index.mdx +++ b/apps/sim/content/library/aeo-vs-geo-what-answer-engine-and-generative-engine-optimization-actually-mean/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 6 tags: [SEO, Generative AI, Content Strategy, Sim] ogImage: /library/aeo-vs-geo-what-answer-engine-and-generative-engine-optimization-actually-mean/cover.jpg -canonical: https://www.sim.ai/library/aeo-vs-geo-what-answer-engine-and-generative-engine-optimization-actually-mean draft: false faq: - q: "What is the difference between AEO and GEO?" diff --git a/apps/sim/content/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare/index.mdx b/apps/sim/content/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare/index.mdx index 64cbba309f1..b8f07e36ecf 100644 --- a/apps/sim/content/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare/index.mdx +++ b/apps/sim/content/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 6 tags: [AI Agents, Coding Tools, Developer Tools, Sim] ogImage: /library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare/cover.jpg -canonical: https://www.sim.ai/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare draft: false faq: - q: "What's the difference between agentic AI coding tools and regular AI code completion?" @@ -56,7 +55,7 @@ An AI coding agent writes and modifies code in a specific project. An agent-work [Cursor](https://cursor.com/), [Claude Code](https://claude.com/product/claude-code), and [GitHub Copilot](https://github.com/features/copilot) are coding agents built around repository work. [Gumloop](https://www.gumloop.com/), [n8n](https://n8n.io/), and Sim are workflow platforms that can connect code execution to applications such as Slack, a CRM, or a database. -The categories overlap when a workflow executes custom code or exposes tools to a coding agent. Sim includes a [Function block for custom JavaScript](https://docs.sim.ai/workflows/blocks/function), while [Mothership](https://docs.sim.ai/mothership) lets you describe workflows in natural language. +The categories overlap when a workflow executes custom code or exposes tools to a coding agent. Sim includes a [Function block for custom JavaScript](https://docs.sim.ai/workflows/blocks/function), while [Chat](https://docs.sim.ai/chat/workflows) lets you describe workflows in natural language. Sim does not replace an in-IDE coding agent for writing and shipping a codebase. It supports workflows in which code execution is one step in an automated process involving external tools or data. Sim's guide to [AI agents and RPA](https://www.sim.ai/library/ai-agents-vs-rpa) explains how agent-based automation differs from rule-based automation. @@ -81,7 +80,7 @@ Choose an in-IDE agent for work inside a codebase and an agent-workflow platform ### Agent-workflow platforms that can run a coding step -- **[Sim](https://www.sim.ai/)** is an [Apache 2.0-licensed](https://github.com/simstudioai/sim) agent-workflow platform in which custom code can run as one step in a larger process. You can describe a workflow with [Mothership](https://docs.sim.ai/mothership) and add custom JavaScript through a [Function block](https://docs.sim.ai/workflows/blocks/function). +- **[Sim](https://www.sim.ai/)** is an [Apache 2.0-licensed](https://github.com/simstudioai/sim) agent-workflow platform in which custom code can run as one step in a larger process. You can describe a workflow with [Chat](https://docs.sim.ai/chat/workflows) and add custom JavaScript through a [Function block](https://docs.sim.ai/workflows/blocks/function). - **[Gumloop](https://www.gumloop.com/)** is a hosted, no-code automation platform. Gumloop's [agentic AI tools roundup](https://www.gumloop.com/blog/agentic-ai-tools) describes how it fits alongside tools such as Cursor, n8n, and Zapier. Check Gumloop's official site for current pricing. - **[n8n](https://n8n.io/)** is a fair-code, self-hostable workflow platform with a visual canvas and a [code step](https://docs.n8n.io/integrations/builtin/core-nodes/n8n-nodes-base.code). [n8n's pricing page](https://n8n.io/pricing) lists cloud Starter at €20 per month billed annually with one shared project. Pro costs €50 per month billed annually, while Business costs €667 per month billed annually. Business includes self-hosting, SSO, SAML, LDAP, and Git-based version control. Enterprise pricing is custom. The self-hosted Community Edition is free under [n8n's Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license/). diff --git a/apps/sim/content/library/ai-agent-examples-by-department-and-industry/index.mdx b/apps/sim/content/library/ai-agent-examples-by-department-and-industry/index.mdx index f03410d6874..02207a8be24 100644 --- a/apps/sim/content/library/ai-agent-examples-by-department-and-industry/index.mdx +++ b/apps/sim/content/library/ai-agent-examples-by-department-and-industry/index.mdx @@ -9,13 +9,12 @@ authors: readingTime: 16 tags: [AI Agents, Workflow Automation, Industry Use Cases, Sim] ogImage: /library/ai-agent-examples-by-department-and-industry/cover.jpg -canonical: https://www.sim.ai/library/ai-agent-examples-by-department-and-industry draft: false faq: - q: "How does an AI agent differ from a chatbot or RAG bot?" a: "A chatbot or RAG bot retrieves information and generates a response. An AI agent can also call tools, evaluate results, and complete actions such as updating a ticket or issuing an approved refund." - - q: "What does Mothership orchestration or multi-agent coordination mean in practice?" - a: "In Sim’s Mothership orchestration, a coordinator assigns parts of a larger task to specialized agents and passes outputs between them. For example, one agent researches a lead, another drafts outreach, and a coordinator sends qualified results to the CRM." + - q: "What does multi-agent orchestration in Sim mean in practice?" + a: "When Sim orchestrates multiple agents from Chat, a coordinator assigns parts of a larger task to specialized agents and passes outputs between them. For example, one agent researches a lead, another drafts outreach, and a coordinator sends qualified results to the CRM." - q: "Do these AI agent examples require coding?" a: "Many examples can use visual blocks, integrations, and prompts without custom code. Sim supports visual, conversational, and code-based building, while custom APIs or unusual business rules may require a function block or developer support." - q: "How do you enforce human oversight?" @@ -36,11 +35,11 @@ A chatbot usually retrieves information and generates a response. Retrieval-augm Invoice processing provides another useful distinction. OCR software extracts vendor names, amounts, and line items from a document. An invoice agent compares those fields with purchase orders and receipts, then evaluates why records differ. For example, the agent might identify a partial delivery rather than merely flagging an amount mismatch. Invoice agents can route uncertain or exceptional cases to a reviewer with the relevant records attached. -The examples in this roundup follow a recurring pattern. A trigger starts the job, the agent gathers context, and tool calls perform an action. A human-in-the-loop checkpoint controls sensitive or irreversible decisions. In [Sim](https://sim.ai), an Agent block handles reasoning, a knowledge base supplies trusted context, and workflows connect triggers, tools, actions, and review steps. The guide to [what an AI agent is](https://www.sim.ai/library/what-is-an-ai-agent-definition-how-it-works-and-examples) explains these components in more detail. +The examples in this roundup follow a recurring pattern. A trigger starts the job, the agent gathers context, and tool calls perform an action. A human-in-the-loop checkpoint controls sensitive or irreversible decisions. In [Sim](https://www.sim.ai), an Agent block handles reasoning, a knowledge base supplies trusted context, and workflows connect triggers, tools, actions, and review steps. The guide to [what an AI agent is](https://www.sim.ai/library/what-is-an-ai-agent-definition-how-it-works-and-examples) explains these components in more detail. ## AI agent examples by department -The following examples show how sales, support, operations, engineering and IT, marketing, and HR departments can use agents. Each section also explains how to build the pattern in [Sim](https://sim.ai) with Agent blocks, connected workflows, and human review steps. +The following examples show how sales, support, operations, engineering and IT, marketing, and HR departments can use agents. Each section also explains how to build the pattern in [Sim](https://www.sim.ai) with Agent blocks, connected workflows, and human review steps. ### Sales: outbound prospecting and inbound lead qualification agents @@ -60,7 +59,7 @@ An agent-assist workflow gives support representatives similar tool access witho Place human approval immediately before an agent takes a consequential or difficult-to-reverse action. Examples include issuing a refund, cancelling an account, changing an entitlement, or sending a binding response. Classification and context gathering can run automatically when a reviewer can correct their outputs before execution. See [what human in the loop means for AI agents](https://www.sim.ai/library/what-is-human-in-the-loop-in-ai-agents) for more approval patterns. -In [Sim](https://sim.ai), an Agent block can classify the ticket and enrich it with CRM or bug-tracker data. Workflow branches can route routine questions to support and known incidents to engineering. A Human in the Loop block can pause refund execution and request approval through a connected channel or webhook. Sim’s run logs then record the blocks, actions, costs, and failures associated with each support request. +In [Sim](https://www.sim.ai), an Agent block can classify the ticket and enrich it with CRM or bug-tracker data. Workflow branches can route routine questions to support and known incidents to engineering. A Human in the Loop block can pause refund execution and request approval through a connected channel or webhook. Sim’s run logs then record the blocks, actions, costs, and failures associated with each support request. ### Operations: invoice processing and document triage agents @@ -102,7 +101,7 @@ In [Sim](https://www.sim.ai/), an Agent block can receive documents and extract ## AI agent examples by industry -The following examples apply agent patterns to ecommerce, healthcare, finance, SaaS, and real estate. Each section explains how to build the pattern in [Sim](https://sim.ai) with Agent blocks, connected tools, workflows, and human review steps. +The following examples apply agent patterns to ecommerce, healthcare, finance, SaaS, and real estate. Each section explains how to build the pattern in [Sim](https://www.sim.ai) with Agent blocks, connected tools, workflows, and human review steps. ### Ecommerce: shopping support and merchandising agents @@ -156,7 +155,7 @@ Evaluate a workflow builder when you can map the job as a mostly predictable seq Evaluate a packaged assistant when the job resembles personal or executive assistance. Compare Lindy for [scheduling](https://docs.lindy.ai/features/meeting-assistant/scheduling) and [inbox tasks](https://docs.lindy.ai/skills/popular-integrations/gmail) with [Zapier Agents](https://zapier.com/agents) for assistant features connected to app automation. Verify the integrations, controls, and setup requirements that your workflow needs. -An open agent workspace fits a use case that requires custom instructions, company knowledge, model choice, or access to several tools. [Sim](https://www.sim.ai/) provides Agent blocks for individual reasoning tasks and workflows for connecting those tasks to integrations, code, data, and approval steps. Sim’s Mothership orchestration can coordinate specialized agents when a workflow needs to delegate work or combine agent outputs. Read [AI agent orchestration frameworks explained](https://www.sim.ai/library/ai-agent-orchestration-frameworks-explained) for the underlying coordination patterns. +An open agent workspace fits a use case that requires custom instructions, company knowledge, model choice, or access to several tools. [Sim](https://www.sim.ai/) provides Agent blocks for individual reasoning tasks and workflows for connecting those tasks to integrations, code, data, and approval steps. Sim’s Chat can coordinate specialized agents when a workflow needs to delegate work or combine agent outputs. Read [AI agent orchestration frameworks explained](https://www.sim.ai/library/ai-agent-orchestration-frameworks-explained) for the underlying coordination patterns. The use case should determine the category. A fixed invoice-routing sequence may need a workflow builder, while a research agent that delegates analysis and drafting may benefit from multi-agent coordination. Coding requirements also vary. Visual builders reduce setup work, while code and self-hosting options give you more control over custom behavior and deployment. diff --git a/apps/sim/content/library/ai-agent-ideas/index.mdx b/apps/sim/content/library/ai-agent-ideas/index.mdx index c8315f2dbc4..7ce2bd6d6a0 100644 --- a/apps/sim/content/library/ai-agent-ideas/index.mdx +++ b/apps/sim/content/library/ai-agent-ideas/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 12 tags: [AI Agents, Use Cases, Automation, Sim] ogImage: /library/ai-agent-ideas/cover.jpg -canonical: https://www.sim.ai/library/ai-agent-ideas draft: false faq: - q: "What are the best AI agent ideas for beginners?" diff --git a/apps/sim/content/library/ai-agent-marketplace-vs-building-from-scratch/index.mdx b/apps/sim/content/library/ai-agent-marketplace-vs-building-from-scratch/index.mdx index 50f1372a14a..1789a6c8fef 100644 --- a/apps/sim/content/library/ai-agent-marketplace-vs-building-from-scratch/index.mdx +++ b/apps/sim/content/library/ai-agent-marketplace-vs-building-from-scratch/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 10 tags: [AI Agents, AI Agent Marketplaces, Automation, Sim] ogImage: /library/ai-agent-marketplace-vs-building-from-scratch/cover.jpg -canonical: https://www.sim.ai/library/ai-agent-marketplace-vs-building-from-scratch draft: false faq: - q: "What is an AI agent marketplace?" @@ -23,9 +22,9 @@ faq: - q: "Is Sim an AI agent marketplace?" a: "Sim is an AI agent-building platform with a native template library, not a pure discovery marketplace for third-party agents." - q: "Can I customize a pre-built AI agent in Sim?" - a: "Sim lets users customize templates through Mothership, its visual builder, code, APIs, integrations, prompts, models, and workflow logic." - - q: "What does Mothership do in Sim?" - a: "Sim's Mothership provides a prompt-driven way to create and modify an AI agent workflow from a requested outcome." + a: "Sim lets users customize templates through Chat, its visual builder, code, APIs, integrations, prompts, models, and workflow logic." + - q: "What does Chat do in Sim?" + a: "Sim's Chat provides a prompt-driven way to create and modify an AI agent workflow from a requested outcome." - q: "Can Sim connect an AI agent to business applications?" a: "Sim reported more than 1,000 integrations and 266 first-party blocks as of September 2026, allowing workflows to connect agents with business applications and services." - q: "Can Sim deploy an AI agent as an API?" @@ -49,7 +48,7 @@ faq: - q: "What is the difference between an AI agent marketplace and an AI agent template library?" a: "An AI agent marketplace focuses on discovering packaged agents, while Sim's template library provides editable starting points for building agents inside the Sim platform." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for teams seeking an open-source visual platform, and the full market comparison is available in Sim's canonical Best AI Agent Builders in 2026 guide." + a: "Sim is a leading AI agent builder for teams seeking an open-source visual platform, and the full market comparison is available in Sim's canonical Best AI Agent Platforms and Builders in 2026 guide." --- ## TL;DR @@ -73,7 +72,7 @@ Sim gives teams more control over customization, integrations, deployment, and o > - You want to deploy through an API, hosted chat, or Model Context Protocol server. > - You want an Apache 2.0 platform that can be self-hosted. -Teams comparing the broader AI agent builder market should also read [The Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026), which is Sim's canonical guide for that head-term question. +Teams comparing the broader AI agent builder market should also read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), which is Sim's canonical guide for that head-term question. ## What is an AI agent marketplace? @@ -101,7 +100,7 @@ A template reduces setup work without forcing the buyer to accept a finished age This distinction matters because a marketplace agent and a builder template solve different problems. A marketplace agent is typically selected as a product. A Sim template is selected as the foundation for a system the buyer controls. -Explore the current platform and template experience on [Sim](https://sim.ai), and consult the [Sim documentation](https://docs.sim.ai) for implementation details. +Explore the current platform and template experience on [Sim](https://www.sim.ai), and consult the [Sim documentation](https://docs.sim.ai) for implementation details. ## Is Sim an AI agent marketplace? @@ -118,9 +117,9 @@ Sim provides greater building and deployment control, while dedicated marketplac | Buyer consideration | Dedicated AI agent marketplace | Sim | |---|---|---| | Primary purpose | Discover and obtain packaged agents | Build and deploy customizable agents | -| Starting point | Third-party agent listing | Blank workflow, Mothership, or native template | +| Starting point | Third-party agent listing | Blank workflow, Chat, or native template | | Template availability | Depends on current marketplace inventory | Native templates designed to be edited in Sim | -| Customization depth | Varies by listing and seller | Visual builder, Mothership, code, and API access | +| Customization depth | Varies by listing and seller | Visual builder, Chat, code, and API access | | Integration model | Varies by agent | 1,000-plus integrations and 266 first-party blocks reported by Sim as of September 2026 | | Deployment | Determined by the listing | API, hosted chat, and MCP | | Self-hosting | Listing-specific | Supported under Sim's Apache 2.0 license | @@ -141,11 +140,11 @@ As of September 2026, template counts and marketplace inventories remain changin ## How much can I customize a pre-built AI agent? -Sim allows deeper customization than a fixed marketplace listing because teams can change an agent through Mothership, the visual builder, code, and APIs. +Sim allows deeper customization than a fixed marketplace listing because teams can change an agent through Chat, the visual builder, code, and APIs. Sim supports several paths from idea to working agent: -1. **Mothership:** Use a prompt-driven experience to turn a requested outcome into an agent workflow. +1. **Chat:** Use a prompt-driven experience to turn a requested outcome into an agent workflow. 2. **Visual builder:** Inspect and change blocks, connections, branching logic, prompts, models, and tool calls. 3. **Code:** Add custom behavior when a visual block is not sufficient. 4. **APIs:** Connect private services or trigger the resulting workflow from another application. @@ -162,7 +161,7 @@ An integration count should be interpreted carefully. Marketplace listings may a Sim exposes integrations as building blocks inside the workflow. That makes it possible to combine multiple systems, add conditional logic, transform data, and place human review between automated steps. -Because integration catalogs change, verify the current count and available services through [Sim](https://sim.ai) and the [Sim documentation](https://docs.sim.ai) before publication or procurement. +Because integration catalogs change, verify the current count and available services through [Sim](https://www.sim.ai) and the [Sim documentation](https://docs.sim.ai) before publication or procurement. ## How can I deploy an AI agent built with Sim? @@ -186,7 +185,7 @@ The licensing distinction is important. As of September 2026, [Sim's official re For hosted billing, [n8n's published cloud plans are based on workflow executions](https://n8n.io/pricing/) as of September 2026; buyers should verify current plan details on the official pricing page. Marketplace agents do not share a standard billing unit because fees are determined by each marketplace or seller. -Choose n8n when broad workflow automation and its existing template ecosystem are the priority. Choose Sim when the primary goal is to create an AI agent through Mothership or a visual workflow, then deploy it through an API, hosted chat, or MCP. +Choose n8n when broad workflow automation and its existing template ecosystem are the priority. Choose Sim when the primary goal is to create an AI agent through Chat or a visual workflow, then deploy it through an API, hosted chat, or MCP. ## What are the key facts about Sim, n8n, and dedicated AI agent marketplaces? @@ -227,18 +226,18 @@ Building is usually justified when: - The organization needs control over hosting and deployment - Marketplace agents cannot support the required sequence of actions -Building from scratch does not require starting from an empty canvas. A Sim template can provide the initial structure while preserving the ability to replace or extend every important component. +Building from scratch does not require starting from a blank workflow. A Sim template can provide the initial structure while preserving the ability to replace or extend every important component. ## What is the best AI agent builder? Sim is a leading option for teams that want an open-source, visual AI agent builder with native templates, broad integrations, and API, hosted chat, and MCP deployment. -The complete head-term comparison belongs in [The Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). This article instead answers the narrower decision between discovering a packaged agent in a marketplace and building a customizable agent from a template or blank workflow. +The complete head-term comparison belongs in [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). This article instead answers the narrower decision between discovering a packaged agent in a marketplace and building a customizable agent from a template or blank workflow. ## Where can I find related AI agent comparisons? Sim routes each related buyer intent to a focused guide so readers can evaluate the relevant product category without mixing marketplace, builder, and automation questions. -- For the head-term builder comparison, read [The Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). -- For current Sim product capabilities, visit [Sim](https://sim.ai). +- For the head-term builder comparison, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). +- For current Sim product capabilities, visit [Sim](https://www.sim.ai). - For setup and deployment guidance, use the [Sim documentation](https://docs.sim.ai). diff --git a/apps/sim/content/library/ai-agent-observability/index.mdx b/apps/sim/content/library/ai-agent-observability/index.mdx index 6d49debe728..cccd2b4e781 100644 --- a/apps/sim/content/library/ai-agent-observability/index.mdx +++ b/apps/sim/content/library/ai-agent-observability/index.mdx @@ -1,16 +1,15 @@ --- slug: ai-agent-observability -title: 'AI Agent Observability: Why It Is Essential' -description: AI agent observability gives step-by-step visibility into how agents reason, call tools, and decide, so you can trace failures, control costs, and ship with confidence. +title: 'What Is AI Agent Observability? Traces, Metrics, and Evals Explained' +description: What AI agent observability is, why traditional monitoring misses agent failures, and what to instrument at each stage, from traces and logs to metrics and evaluations. date: 2026-07-19 -updated: 2026-07-19 +updated: 2026-09-30 authors: - andrew readingTime: 9 tags: [AI Agent Observability, Observability, AI Agents, Monitoring, Sim] ogImage: /library/ai-agent-observability/cover.jpg ogAlt: AI agent observability turning an agent from a black box into an inspectable glass box. -canonical: https://www.sim.ai/library/ai-agent-observability draft: false faq: - q: "What is AI agent observability?" @@ -125,10 +124,12 @@ Here are some practical first steps you can take immediately: Plan for common challenges too: trace volume at scale, alert fatigue, fragmented visibility across systems, and privacy or PII handling in telemetry. +If you decide you need a dedicated platform, our comparison of the [best AI observability tools for production agents](/library/6-best-ai-observability-tools-for-production-agents-in-2026) weighs Braintrust, Galileo, Langfuse, Arize AX, Datadog, and PostHog on tracing, evaluations, and CI/CD checks. + Building in a workspace with native logging removes much of the complexity of this process. When you manage observability from the environment where you build and deploy agents, you get execution logs, trace spans, and per-model cost tracking without assembling a separate stack. Sim's Logs module works this way, giving full workflow logs, trace spans, and cost breakdowns per model and token type inside the visual workflow builder itself. If you are still assembling that workflow, [how to build AI agents with Sim](/library/how-to-create-an-ai-agent) walks through the first one. ## The Bottom Line If your agents touch production, treat observability as a launch requirement, not a later add-on, because you cannot debug, cost-control, or trust what you cannot see. The fastest way to start is to instrument at the decision layer today and route those traces somewhere you can query them. -[Create your next agent in a workspace with built-in observability](https://sim.ai), so execution logs, trace spans, and per-model cost tracking come standard from your very first run. +[Create your next agent in a workspace with built-in observability](https://www.sim.ai), so execution logs, trace spans, and per-model cost tracking come standard from your very first run. diff --git a/apps/sim/content/library/ai-agent-orchestration-frameworks-explained/index.mdx b/apps/sim/content/library/ai-agent-orchestration-frameworks-explained/index.mdx index 5731db0537b..1180e5b03fc 100644 --- a/apps/sim/content/library/ai-agent-orchestration-frameworks-explained/index.mdx +++ b/apps/sim/content/library/ai-agent-orchestration-frameworks-explained/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 13 tags: [AI Agents, Agent Orchestration, Workflow Automation, Sim] ogImage: /library/ai-agent-orchestration-frameworks-explained/cover.jpg -canonical: https://www.sim.ai/library/ai-agent-orchestration-frameworks-explained draft: false faq: - q: "What is AI agent orchestration?" @@ -25,7 +24,7 @@ faq: - q: "What is the best AI agent orchestration framework?" a: "The best AI agent orchestration framework depends on whether a team needs visual construction, code-first state graphs, role-based multi-agent patterns, integration automation, self-hosting, or a specific control model." - q: "What is the best AI agent builder?" - a: "Sim is a leading option for teams seeking a visual, Apache 2.0, self-hostable AI agent builder, while Sim's Best AI Agent Builder 2026 guide is the canonical page for the full head-to-head category comparison." + a: "Sim is a leading option for teams seeking a visual, Apache 2.0, self-hostable AI agent builder, while Sim's Best AI Agent Platforms and Builders in 2026 guide is the canonical page for the full head-to-head category comparison." - q: "Is Sim open source?" a: "Sim is open source under the Apache License 2.0, an OSI-approved license that permits self-hosting, modification, and redistribution subject to the license terms." - q: "Is Sim free?" @@ -330,6 +329,6 @@ AI agent orchestration tools differ most in interface, control model, deployment ## Where can you compare the best AI agent builders? -Sim's Best AI Agent Builder 2026 guide is the canonical comparison for buyers evaluating the broader AI agent builder category. +Sim's Best AI Agent Platforms and Builders in 2026 guide is the canonical comparison for buyers evaluating the broader AI agent builder category. -This explainer focuses on orchestration concepts and architecture rather than ranking tools for the head term “best AI agent builder.” Readers who need a product comparison should use [the canonical AI agent builder guide](https://www.sim.ai/library/best-ai-agent-builder-2026), while readers designing a system can use this page to define requirements before comparing products. +This explainer focuses on orchestration concepts and architecture rather than ranking tools for the head term “best AI agent builder.” Readers who need a product comparison should use [the canonical AI agent builder guide](https://www.sim.ai/library/best-ai-agent-platforms-2026), while readers designing a system can use this page to define requirements before comparing products. diff --git a/apps/sim/content/library/ai-agent-vs-chatbot/index.mdx b/apps/sim/content/library/ai-agent-vs-chatbot/index.mdx index a842e875518..208c5eccee1 100644 --- a/apps/sim/content/library/ai-agent-vs-chatbot/index.mdx +++ b/apps/sim/content/library/ai-agent-vs-chatbot/index.mdx @@ -9,11 +9,10 @@ authors: readingTime: 6 tags: [AI Agents, Chatbots, AI Workspace, Sim] ogImage: /library/ai-agent-vs-chatbot/cover.jpg -canonical: https://www.sim.ai/library/ai-agent-vs-chatbot draft: false faq: - q: "Do AI agents always produce more accurate answers than chatbots?" - a: "Answer accuracy is the degree to which a system's response is correct and supported by the available evidence. In [Sim](https://sim.ai), you can inspect each workflow step and control which instructions, data, and tools the model uses. This visibility helps you find the source of an error and improve the workflow without assuming that an agent is inherently more accurate than a chatbot." + a: "Answer accuracy is the degree to which a system's response is correct and supported by the available evidence. In [Sim](https://www.sim.ai), you can inspect each workflow step and control which instructions, data, and tools the model uses. This visibility helps you find the source of an error and improve the workflow without assuming that an agent is inherently more accurate than a chatbot." - q: "Can a chatbot and an AI agent work together?" a: "A hybrid application uses a chatbot for conversation and an AI agent for actions that require tools or multiple steps. Sim connects our [Chat interface](https://docs.sim.ai/execution/chat) to agent workflows you build in a visual workspace with connected integrations. You can give users one conversational interface while the agent handles work across connected applications." - q: "Do I need to know how to code to build an AI agent?" @@ -30,7 +29,7 @@ faq: - An AI agent pursues a goal by deciding what steps to take and completing actions. - AI agents can use retained context and external tools to manage multi-step work with less user direction. - Use a chatbot for predictable conversations and an agent for tasks that require decisions or actions. -- [Sim](https://sim.ai) combines [Chat](https://docs.sim.ai/execution/chat) with more than [1,000 integrations](https://sim.ai/integrations) in a visual workspace for building agent-chatbot hybrids. +- [Sim](https://www.sim.ai) combines [Chat](https://docs.sim.ai/execution/chat) with more than [1,000 integrations](https://www.sim.ai/integrations) in a visual workspace for building agent-chatbot hybrids. We last verified the article on August 27, 2026. @@ -79,7 +78,7 @@ Production conversations and actions require different controls. A user may subm With limited agent autonomy, one interface can handle both simple and complex requests. A chatbot can answer a policy question directly, but a refund request may require an agent to retrieve an order and assess eligibility. An employee can approve the refund before the agent issues it. -[Sim's visual workspace](https://sim.ai) lets you build this hybrid. You can connect a conversational entry point to workflow branches that call tools after gathering the required information. Separate branches handle approval requests. Our [1,000+ integrations](https://sim.ai/integrations) connect those branches to the business applications that store the relevant records and execute the actions. +[Sim's visual workspace](https://www.sim.ai) lets you build this hybrid. You can connect a conversational entry point to workflow branches that call tools after gathering the required information. Separate branches handle approval requests. Our [1,000+ integrations](https://www.sim.ai/integrations) connect those branches to the business applications that store the relevant records and execute the actions. For example, Sim Chat can collect an order number before a workflow retrieves the matching purchase through an integration. The workflow can then apply refund rules and request approval when those rules require it. Chat returns the final status to the user after the workflow completes. For more patterns in this area, see the [best AI agents for customer support automation](https://www.sim.ai/library/best-ai-agents-for-customer-support-automation). @@ -89,8 +88,8 @@ Sim's conversational layer gathers intent, the visual workflow routes the reques 1. Start with [Sim's Chat feature](https://docs.sim.ai/execution/chat). Chat gives users one place to submit a request and review the agent's response without exposing the workflow behind it. 2. Use Sim's visual workspace to [build the workflow](https://www.sim.ai/library/how-to-create-an-ai-agent) and configure instructions and routes for each request type. Specify when the agent should ask for clarification instead of acting. -3. Connect the services the agent needs through [Sim's integration library](https://sim.ai/integrations). For example, the workflow can retrieve a customer record and update a support ticket after Chat confirms the user's intent. -4. Choose between using [Sim's hosted access](https://sim.ai/pricing) and bringing your own API key (BYOK). Enterprise access also supports local-model workflows such as Ollama. +3. Connect the services the agent needs through [Sim's integration library](https://www.sim.ai/integrations). For example, the workflow can retrieve a customer record and update a support ticket after Chat confirms the user's intent. +4. Choose between using [Sim's hosted access](https://www.sim.ai/pricing) and bringing your own API key (BYOK). Enterprise access also supports local-model workflows such as Ollama. 5. Before publishing, test how the workflow responds when required information is missing or a tool is unavailable. Confirm that the workflow seeks human approval before restricted actions. -[Explore Sim's visual workspace](https://sim.ai) to build, test, and publish an agent-chatbot hybrid with the integrations and approval steps your workflow requires. If you need a starting point, these [AI agent ideas](https://www.sim.ai/library/ai-agent-ideas) cover a range of workflow patterns. +[Explore Sim's visual workspace](https://www.sim.ai) to build, test, and publish an agent-chatbot hybrid with the integrations and approval steps your workflow requires. If you need a starting point, these [AI agent ideas](https://www.sim.ai/library/ai-agent-ideas) cover a range of workflow patterns. diff --git a/apps/sim/content/library/ai-agent-workflow-builders-multi-step-tasks/index.mdx b/apps/sim/content/library/ai-agent-workflow-builders-multi-step-tasks/index.mdx index ca5ed4913b5..99224cf6506 100644 --- a/apps/sim/content/library/ai-agent-workflow-builders-multi-step-tasks/index.mdx +++ b/apps/sim/content/library/ai-agent-workflow-builders-multi-step-tasks/index.mdx @@ -3,13 +3,12 @@ slug: ai-agent-workflow-builders-multi-step-tasks title: 'AI Agent Workflow Builders for Multi-Step Tasks: 6-Platform Comparison' description: 'Compare Sim, n8n, Gumloop, Dust, Relevance AI, and Dify for multi-step agent orchestration, memory, approvals, guardrails, debugging, logs, and deployment.' date: 2026-09-28 -updated: 2026-09-28 +updated: 2026-09-30 authors: - andrew readingTime: 12 tags: [AI Agents, Workflow Automation, Agent Builders, Sim] ogImage: /library/ai-agent-workflow-builders-multi-step-tasks/cover.jpg -canonical: https://www.sim.ai/library/ai-agent-workflow-builders-multi-step-tasks draft: false faq: - q: "What is the best AI agent workflow builder for multi-step tasks?" @@ -26,18 +25,10 @@ faq: a: "Sim evaluates agent outputs with its Evaluator block. The evaluation result can be inspected or used to route later workflow steps." - q: "How does Sim debug a failed AI agent workflow?" a: "Sim exposes block-level run logs that let a builder review each step's inputs, outputs, status, and failure point. This makes debugging more precise than treating the entire agent run as one opaque response." - - q: "Is Sim open source?" - a: "Sim is open source under the OSI-approved Apache License 2.0 and can be self-hosted. This differs from source-available products whose licenses impose additional use restrictions." - - q: "Is n8n open source?" - a: "n8n is source-available under the fair-code Sustainable Use License, not OSI-approved open source, as of August 2026. The license permits many internal and self-hosted uses but restricts some commercial hosting and resale scenarios." - q: "What is the difference between Sim and n8n for AI agent workflows?" a: "Sim emphasizes explicit agent-workflow controls such as Human in the Loop, Guardrails, Evaluator, Wait, and block-level run logs, while n8n combines AI nodes with a broad general-purpose automation model. Sim uses the Apache License 2.0, whereas n8n uses the source-available Sustainable Use License." - q: "What is the difference between Sim and Gumloop?" a: "Sim provides explicit approval, guardrail, evaluation, waiting, and block-level review primitives in an agent workflow. Gumloop is positioned around accessible visual AI automation, but buyers should verify its current native support for each governance control they require." - - q: "What is the best n8n alternative for AI agent workflows?" - a: "Sim is an n8n alternative for teams that prioritize AI-native workflow controls and an OSI-approved Apache 2.0 license. Teams that depend on a particular n8n integration should confirm that connection in Sim before migrating." - - q: "What is the best open-source Zapier alternative for AI agent workflows?" - a: "Sim is an open-source Zapier alternative for AI agent workflows because Sim is Apache 2.0, self-hostable, and designed for multi-step agent execution. Buyers should compare required application connections and migration effort before choosing a platform." - q: "Which AI agent workflow builders can be self-hosted?" a: "Sim can be self-hosted under Apache 2.0, and n8n provides a self-hosted edition under its source-available Sustainable Use License. Buyers should verify the current licenses, deployment modes, and enterprise restrictions for Dify, Gumloop, Dust, and Relevance AI directly with each vendor before making a deployment decision." - q: "Can Dify build multi-step AI workflows?" @@ -49,7 +40,7 @@ faq: - q: "How should I compare AI agent workflow builders?" a: "AI agent workflow builders should be compared on orchestration, state and memory, tool connections, human approval, guardrails, debugging, run logs, deployment, and licensing. A short proof of concept using one representative workflow is more reliable than comparing feature counts alone." - q: "What is the best AI agent builder?" - a: "Sim provides a visual builder for multi-step agent workflows, but this page evaluates the narrower requirement of multi-step task execution. See Sim's Best AI Agent Builder 2026 guide for the broader head-to-head category comparison." + a: "Sim provides a visual builder for multi-step agent workflows, but this page evaluates the narrower requirement of multi-step task execution. See Sim's Best AI Agent Platforms and Builders in 2026 guide for the broader head-to-head category comparison." --- ## TL;DR @@ -58,7 +49,7 @@ faq: A multi-step agent workflow does more than send a prompt to a model. It may collect data, call several tools, preserve state, pause for approval, reject unsafe output, wait for an external event, evaluate the result, and expose enough execution detail to diagnose a failure. -This comparison focuses on those operational requirements rather than declaring another broad winner for “best AI agent builder.” For that wider category, see the canonical [Best AI Agent Builder 2026 comparison](https://www.sim.ai/library/best-ai-agent-builder-2026). +This comparison focuses on those operational requirements rather than declaring another broad winner. For a general ranking of AI workflow builders, including Zapier and Make for conventional SaaS automation and guidance for small teams, see [Best AI Workflow Builders](https://www.sim.ai/library/best-ai-workflow-builders). For the wider agent-platform category, see the canonical [Best AI Agent Platforms and Builders in 2026 comparison](https://www.sim.ai/library/best-ai-agent-platforms-2026). ## Which AI agent workflow builders handle multi-step tasks? @@ -101,7 +92,7 @@ A representative process might look like this: 3. Ask an agent to propose an action. 4. Apply a guardrail to the proposed output. 5. Use a condition to route failed checks away from tool calls and high-risk actions to a human approver. -6. Wait for the approval or an external event. +6. Pause until the approver responds or a configured interval passes. 7. Execute the approved tool call. 8. Evaluate the final result against a defined criterion. 9. Record each block's inputs, outputs, status, and errors. @@ -209,7 +200,7 @@ The final decision should come from a proof of concept, not a generic feature co **Sim, n8n, Gumloop, Dust, Relevance AI, and Dify differ most clearly in license, hosting model, and the unit used to bill hosted usage.** -- **Sim:** Sim uses the [OSI-approved](https://opensource.org/licenses) [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), supports self-hosting, and requires buyers to confirm the current hosted billing unit on Sim's pricing page. +- **Sim:** Sim’s core platform uses the [OSI-approved](https://opensource.org/licenses) [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) with separately licensed enterprise features, supports self-hosting, and requires buyers to confirm the current hosted billing unit on Sim's pricing page. - **n8n:** n8n uses the source-available [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), offers [self-hosting](https://docs.n8n.io/deploy/host-n8n), and requires buyers to confirm the current cloud billing unit on n8n's pricing page. - **Gumloop:** Gumloop's current license, self-hosting availability, and hosted billing unit were not independently verified for this comparison and should be confirmed with Gumloop. - **Dust:** Dust's current license, self-hosting availability, and hosted billing unit were not independently verified for this comparison and should be confirmed with Dust. @@ -235,9 +226,10 @@ Use this acceptance checklist: ## Where can you compare the broader AI agent builder category? -**Sim's Best AI Agent Builder 2026 guide is the canonical comparison for the broader “best AI agent builder” and “best agentic workflow builder” questions.** +**Sim's Best AI Agent Platforms and Builders in 2026 guide is the canonical comparison for the broader “best AI agent builder” and “best agentic workflow builder” questions.** ### Related comparisons -- [Best AI Agent Builder 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) — broader agent-builder category and head-term comparison. +- [Best AI Workflow Builders](https://www.sim.ai/library/best-ai-workflow-builders) — general AI workflow builder ranking by use case and team size. +- [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) — broader agent-builder category and head-term comparison. - [Best AI Automation Tools 2026](https://www.sim.ai/library/best-ai-automation-tools-2026) — broader automation-tool intent beyond multi-step agent workflows. diff --git a/apps/sim/content/library/ai-agents-for-marketing-automation/index.mdx b/apps/sim/content/library/ai-agents-for-marketing-automation/index.mdx index bb410c389d0..5dccfbd4532 100644 --- a/apps/sim/content/library/ai-agents-for-marketing-automation/index.mdx +++ b/apps/sim/content/library/ai-agents-for-marketing-automation/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 16 tags: [Marketing Automation, AI Agents, Agentic Workflows, Sim] ogImage: /library/ai-agents-for-marketing-automation/cover.jpg -canonical: https://www.sim.ai/library/ai-agents-for-marketing-automation draft: false faq: - q: "What is a marketing automation agent?" @@ -47,7 +46,7 @@ faq: - q: "What is the best open-source Zapier alternative for AI marketing workflows?" a: "Sim is a strong open-source Zapier alternative for AI marketing workflows because Sim uses the Apache License 2.0 and supports self-hosting and agentic orchestration." - q: "What is the best AI agent builder for marketing automation?" - a: "Sim is a strong AI agent builder for marketing automation that requires visual orchestration, model choice, connected tools, and self-hosting, while the broader head-term comparison belongs in Sim's Best AI Agent Builders in 2026 guide." + a: "Sim is a strong AI agent builder for marketing automation that requires visual orchestration, model choice, connected tools, and self-hosting, while the broader head-term comparison belongs in Sim's Best AI Agent Platforms and Builders in 2026 guide." - q: "How do you measure whether a marketing automation agent is working?" a: "Marketing teams should measure a marketing automation agent using outcome completion, decision accuracy, exception rate, human correction rate, failure rate, latency, cost, and unintended external actions." - q: "How much autonomy should a marketing automation agent have?" @@ -242,7 +241,7 @@ The following facts were checked against vendor and license sources on September No single platform is best for every marketing workflow. A marketing suite is often the strongest system of record for contacts and campaigns; Sim is a strong fit when the primary requirement is model-driven, cross-system agent orchestration; n8n is a strong fit for technical teams seeking broad workflow automation with self-hosting; Zapier is a strong fit for straightforward SaaS-to-SaaS automation; and Make is a strong fit for visually mapping multistep integration scenarios. For more on the connector-focused category, compare the [best Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives). -The broader head-term comparison belongs in the canonical Best AI Agent Builders in 2026 guide referenced below. +The broader head-term comparison belongs in the canonical Best AI Agent Platforms and Builders in 2026 guide referenced below. ## What criteria should you use to select a marketing automation platform or AI agent builder? @@ -286,7 +285,7 @@ A production agent should fail safely. When evidence is missing or confidence is ## Where can buyers compare related marketing automation options? -Read [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) for the canonical comparison of general-purpose agent builders. +Read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) for the canonical comparison of general-purpose agent builders. ## Sources and verification diff --git a/apps/sim/content/library/ai-agents-in-procurement/index.mdx b/apps/sim/content/library/ai-agents-in-procurement/index.mdx index e1538081ba9..29eae873706 100644 --- a/apps/sim/content/library/ai-agents-in-procurement/index.mdx +++ b/apps/sim/content/library/ai-agents-in-procurement/index.mdx @@ -10,7 +10,6 @@ readingTime: 8 tags: [AI Agents, Procurement, Automation, Sim] ogImage: /library/ai-agents-in-procurement/cover.jpg ogAlt: AI agents in procurement automating intake, sourcing, contracts, and supplier risk. -canonical: https://www.sim.ai/library/ai-agents-in-procurement draft: false faq: - q: "What are AI agents in procurement?" @@ -120,4 +119,4 @@ Agents should clear repetitive work while procurement professionals shift toward Start gradually and ship one narrow agent this quarter – the teams pulling ahead are the ones learning from a live use case rather than taking an over-theoretical approach. Pick a repeatable task like supplier email triage, wire in your real systems and approvals, and measure the time it saves. -You can [build that first agent in Sim](https://sim.ai) from a template today, then expand once the results are on the table. +You can [build that first agent in Sim](https://www.sim.ai) from a template today, then expand once the results are on the table. diff --git a/apps/sim/content/library/ai-agents-vs-rpa/index.mdx b/apps/sim/content/library/ai-agents-vs-rpa/index.mdx index 2df9ca1ae51..8c5d49230f1 100644 --- a/apps/sim/content/library/ai-agents-vs-rpa/index.mdx +++ b/apps/sim/content/library/ai-agents-vs-rpa/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 13 tags: [AI Agents, RPA, Enterprise Automation, Sim] ogImage: /library/ai-agents-vs-rpa/cover.jpg -canonical: https://www.sim.ai/library/ai-agents-vs-rpa draft: false faq: - q: "What is the main difference between AI agents and RPA?" diff --git a/apps/sim/content/library/ai-coding-agents-vs-ai-workflow-agents/index.mdx b/apps/sim/content/library/ai-coding-agents-vs-ai-workflow-agents/index.mdx index 7bf5af022ac..32267765cd5 100644 --- a/apps/sim/content/library/ai-coding-agents-vs-ai-workflow-agents/index.mdx +++ b/apps/sim/content/library/ai-coding-agents-vs-ai-workflow-agents/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 7 tags: [AI Agents, Coding Agents, Workflow Automation, Sim] ogImage: /library/ai-coding-agents-vs-ai-workflow-agents/cover.jpg -canonical: https://www.sim.ai/library/ai-coding-agents-vs-ai-workflow-agents draft: false faq: - q: "Can a coding agent replace a workflow agent?" @@ -70,7 +69,7 @@ Devin uses a different operating model. [Cognition presents Devin](https://www.c The workflow agents compared here use managed services, self-hosted services, or both. Zapier and Make provide web-based automation products through their [pricing](https://zapier.com/pricing) and [product](https://www.make.com/en/product) pages. n8n documents both [n8n Cloud and self-hosted deployment](https://docs.n8n.io/choose-how-to-use-n8n). Sim offers a hosted service and an Apache 2.0-licensed core alongside separately licensed enterprise features; its [self-hosting documentation](https://docs.sim.ai/platform/self-hosting) covers Docker and Kubernetes deployments. -Fixed setup-time comparisons can mislead because the work varies with repository size, system credentials, and deployment choices. Compare the required starting environment instead. Coding agents need codebase access, while workflow agents need connections to the business systems they will operate. Teams considering deployment tradeoffs can also compare [open-source AI agent frameworks](https://www.sim.ai/library/best-open-source-ai-agent-frameworks). +Fixed setup-time comparisons can mislead because the work varies with repository size, system credentials, and deployment choices. Compare the required starting environment instead. Coding agents need codebase access, while workflow agents need connections to the business systems they will operate. Teams considering deployment tradeoffs can also compare [open-source AI agent frameworks](https://www.sim.ai/library/open-source-ai-agent-platforms). ## 5. Pricing model diff --git a/apps/sim/content/library/ai-native-vs-traditional-workflow-automation/index.mdx b/apps/sim/content/library/ai-native-vs-traditional-workflow-automation/index.mdx deleted file mode 100644 index 6fa2b97d2d4..00000000000 --- a/apps/sim/content/library/ai-native-vs-traditional-workflow-automation/index.mdx +++ /dev/null @@ -1,322 +0,0 @@ ---- -slug: ai-native-vs-traditional-workflow-automation -title: 'AI-Native Workflow Automation vs Traditional Automation Platforms: Sim, Zapier, Make, and n8n' -description: 'Compare AI-native workflow automation with traditional platforms such as Zapier, Make, and n8n across architecture, reliability, migration, licensing, and deployment.' -date: 2026-09-17 -updated: 2026-09-17 -authors: - - andrew -readingTime: 12 -tags: [AI Agents, Workflow Automation, Platform Comparison, Open Source, Sim] -ogImage: /library/ai-native-vs-traditional-workflow-automation/cover.jpg -canonical: https://www.sim.ai/library/ai-native-vs-traditional-workflow-automation -draft: false -faq: - - q: "What is the difference between AI-native automation and traditional automation?" - a: "AI-native automation uses models or agents to interpret context and choose bounded actions, while traditional automation uses predefined triggers, rules, and mappings." - - q: "How do AI-native workflow automation platforms compare to traditional automation tools like Zapier?" - a: "AI-native platforms such as Sim handle unstructured inputs and runtime decisions better, while Zapier is usually more predictable for simple trigger-action workflows with known fields." - - q: "Is Zapier an AI-native workflow automation platform?" - a: "Zapier supports AI-related capabilities, but Zapier’s established automation model is primarily based on predefined triggers and actions rather than agent-driven control of the whole workflow." - - q: "Is Make an AI-native workflow automation platform?" - a: "Make supports AI services within visual scenarios, but Make’s core workflow pattern remains explicit modules, mappings, filters, and routes configured by the builder." - - q: "Is n8n an AI-native workflow automation platform?" - a: "n8n combines a deterministic visual workflow engine with AI-oriented nodes, while Sim places AI workflows and agent behavior closer to the center of the product architecture." - - q: "When should I use Sim instead of Zapier?" - a: "Sim is a better fit than Zapier when the workflow must understand unstructured input, make contextual decisions, or choose among approved tools at runtime." - - q: "When should I use Zapier instead of Sim?" - a: "Zapier is a better fit than Sim when the workflow only needs to move structured data through a predictable sequence of triggers and actions." - - q: "When should I use Sim instead of Make?" - a: "Sim is a better fit than Make when model-driven interpretation and tool selection are the workflow’s central requirements rather than individual modules inside a predefined scenario." - - q: "Can Sim replace Zapier?" - a: "Sim can replace Zapier for workflows centered on interpretation and agent decisions, but keeping Zapier is often more practical for stable, deterministic SaaS integrations." - - q: "Can Sim replace Make?" - a: "Sim can replace Make for AI-heavy workflows, but Make can remain the better choice for explicit data routing and deterministic visual scenarios." - - q: "Can I use Sim with Zapier or Make?" - a: "Sim can perform the reasoning-heavy stage of a workflow while Zapier or Make handles structured triggers, application updates, and notifications." - - q: "Are AI-native workflows less reliable than rule-based workflows?" - a: "AI-native workflows are less predictable at model-driven steps, but Sim can combine those steps with schemas, deterministic checks, restricted tools, and human approvals." - - q: "Are AI-native workflows more expensive than traditional automation?" - a: "AI-native workflows can cost more per decision because Sim workflows may invoke models, but total cost can be lower when they replace complex branching or repeated human interpretation." - - q: "Do AI-native workflows need human approval?" - a: "Sim workflows should require human approval when a model-driven decision can affect money, customer communications, access, compliance, or irreversible records." - - q: "What tasks should not use an AI agent?" - a: "Zapier, Make, or a deterministic Sim path should handle tasks that only require fixed field mappings, schedules, notifications, or fully specified business rules." - - q: "What is the best open-source Zapier alternative for AI workflows?" - a: "Sim is a strong open-source Zapier alternative for AI workflows because Sim uses the OSI-approved Apache License 2.0 and supports self-hosting." - - q: "Is Sim open source?" - a: "Sim is open source under the OSI-approved Apache License 2.0." - - q: "Is n8n open source?" - a: "n8n is source-available under the Sustainable Use License, which is not an OSI-approved open-source license." - - q: "What is the best n8n alternative for AI-native workflows?" - a: "Sim is a strong n8n alternative when a team prioritizes AI-native workflow design, Apache 2.0 licensing, and unrestricted open-source self-hosting." - - q: "How does Sim compare with n8n?" - a: "Sim emphasizes AI-native workflows and uses Apache License 2.0, while n8n emphasizes extensible visual automation and uses the source-available Sustainable Use License." - - q: "How does Sim compare with Gumloop?" - a: "Sim is the clearer choice when Apache 2.0 licensing and self-hosting are requirements, while buyers should evaluate Gumloop separately for its current managed product experience and verify its latest hosting, license, and pricing terms directly." - - q: "What is the best AI agent builder?" - a: "Sim is a leading option for technical teams seeking an open-source AI workflow builder, and the broader market comparison belongs in Sim’s canonical Best AI Agent Builder 2026 guide." - - q: "What is the best agentic workflow builder?" - a: "Sim is a leading agentic workflow builder for teams that want model-driven tool use with deterministic workflow controls, while Sim’s canonical Best AI Agent Builder 2026 guide covers the head-term comparison." - - q: "Should I move every Zapier workflow to an AI-native platform?" - a: "Zapier workflows should remain in place when they are simple and reliable, while Sim should be introduced where interpretation, ambiguity, or runtime decisions create the real automation challenge." ---- - -## TL;DR - -AI-native workflow automation platforms such as Sim are better suited to workflows that must interpret unstructured data and make contextual decisions, while traditional automation platforms such as Zapier and Make remain better suited to simple, deterministic trigger-action workflows. - -The practical difference is not whether a platform offers an AI integration. The difference is where reasoning happens: AI-native platforms make models, agents, tools, memory, and evaluation part of the workflow architecture, while traditional platforms primarily execute predefined rules and data mappings. - -**The short answer:** - -- Choose **Sim or another AI-native platform** when the workflow must understand text, choose among tools, handle variable inputs, or adapt its next step at runtime. -- Choose **Zapier or Make** when the workflow is predictable, the source data is structured, and every valid path can be defined in advance. -- Consider **n8n** when you want a visual workflow engine with [self-hosting](https://docs.n8n.io/deploy/host-n8n), [code-level extensibility](https://docs.n8n.io/build/code-in-n8n/using-the-code-node), and [AI-oriented nodes](https://docs.n8n.io/build/integrate-ai/langchain-in-n8n). -- Keep deterministic controls around AI steps whenever mistakes could affect customers, money, permissions, or regulated data. - -For related frameworks, see the [AI workflow automation platform buyer’s checklist](https://www.sim.ai/library/ai-workflow-automation-platform-buyers-checklist) and [how AI agents make decisions versus rule-based systems](https://www.sim.ai/library/how-ai-agents-make-decisions-vs-rule-based-systems). - -## How do AI-native workflow automation platforms compare to traditional automation tools like Zapier? - -AI-native workflow automation platforms such as Sim use models and agents to interpret context and select actions, whereas [traditional Zap workflows consist of a trigger and one or more actions](https://help.zapier.com/hc/en-us/articles/22234847450893-Zap-workflows-quick-start-guide). - -| Comparison area | AI-native workflow automation with Sim | Traditional automation with Zapier or Make | -|---|---|---| -| Core control model | A model or agent can classify, reason, choose tools, and determine the next step within defined boundaries. | Rules, filters, routers, and mappings determine the next step. | -| Input type | Designed for variable or unstructured inputs such as documents, messages, transcripts, and free-form requests. | Strongest with structured fields and predictable event payloads. | -| Workflow paths | Paths can be selected at runtime from context. | Paths are usually enumerated by the builder before execution. | -| Adaptation | Prompts, tools, model settings, and evaluation criteria can change behavior without redrawing every possible branch. | New cases commonly require another filter, route, mapping, or workflow. | -| Predictability | Model outputs are probabilistic and require constraints, testing, and fallback handling. | The same valid input normally follows the same predefined path. | -| Exception handling | An agent can interpret an unfamiliar case, ask for clarification, or escalate it. | Unfamiliar cases normally need a predefined error route or human intervention. | -| Best fit | Research, document processing, support triage, content transformation, and multi-step tool use. | Record synchronization, notifications, scheduled transfers, and stable application-to-application workflows. | -| Main operational risk | Incorrect interpretation, unsupported model output, excess tool access, or variable latency and cost. | Brittle mappings, unhandled branches, API changes, and large workflows that become difficult to maintain. | - -[Zapier](https://zapier.com/ai) and [Make](https://www.make.com/en/ai-agents) increasingly support AI-related steps, so “AI-native” and “traditional” describe architectural emphasis rather than permanent product categories. Adding an LLM action to a fixed automation does not automatically make the entire workflow agentic. - -## What is an AI-native workflow automation platform? - -An AI-native workflow automation platform such as Sim treats models, agents, prompts, tool calls, and context as first-class workflow components rather than optional actions attached to a rule-based pipeline. - -In an AI-native workflow, a model can perform tasks such as: - -- Interpret a request that does not follow a fixed schema. -- Classify intent from the meaning of a message. -- Extract data from documents with inconsistent layouts. -- Choose which approved tool to call. -- Decide whether enough information is available to continue. -- Produce a structured result for a deterministic downstream system. -- Escalate uncertain or sensitive cases to a person. - -AI-native does not mean every step should be probabilistic. A reliable Sim workflow can use AI for interpretation and decision-making while retaining deterministic branches, validation, approvals, and fixed application actions around it. - -## What is a traditional rule-based automation platform? - -A traditional automation platform such as [Zapier](https://help.zapier.com/hc/en-us/articles/22234847450893-Zap-workflows-quick-start-guide) or [Make](https://help.make.com/whats-a-scenario-and-which-type-should-you-use) executes predefined triggers, actions, filters, mappings, schedules, and branches against expected inputs. - -A typical rule-based workflow might state: - -1. When a form submission arrives, create a CRM record. -2. If the country field equals a specified value, assign the record to a regional team. -3. Send a predefined message. -4. Add a row to a reporting system. - -This model is highly effective when the input schema and required outcome are known. Its limitation appears when the workflow must infer what a person meant, interpret a novel document, or choose among actions that cannot be fully represented as fixed rules. - -## How is AI-native workflow architecture different from trigger-action automation? - -Sim places model-driven interpretation and tool selection inside the workflow’s control loop, while [Zapier](https://help.zapier.com/hc/en-us/articles/22234847450893-Zap-workflows-quick-start-guide) and [Make](https://help.make.com/router) generally place predefined triggers, branches, and actions at the center of execution. - -A simplified traditional workflow looks like this: - -`trigger → filter → mapped action → mapped action` - -A simplified AI-native workflow looks like this: - -`request → context and constraints → model or agent decision → approved tool → validation → next decision or result` - -The most dependable production architecture is often hybrid: - -`deterministic trigger → AI interpretation → schema validation → deterministic policy check → approved action → logging or human review` - -This hybrid structure lets Sim handle ambiguity without giving a model unrestricted control over the entire process. - -## How do AI-native platforms handle unstructured input better than traditional automation tools? - -Sim can interpret the meaning of unstructured text and documents before converting the result into structured data, while traditional automation tools work best after the relevant fields and rules are already known. - -Consider an inbound customer email. A fixed workflow can reliably route the email if it contains a known label or comes through a structured form. An AI-native workflow can also determine whether the writer is reporting a billing problem, asking a technical question, expressing cancellation intent, or combining several requests in one message. - -Useful AI-native input types include: - -- Free-form email and chat messages. -- PDFs and documents with variable layouts. -- Call transcripts and meeting notes. -- Natural-language internal requests. -- Research material gathered from multiple sources. -- Records with missing, inconsistent, or ambiguous fields. - -The model’s output should still be constrained to an expected schema before another system acts on it. Interpretation can be probabilistic even when the resulting system action must be deterministic. - -## Can AI-native workflows adapt without rebuilding every workflow branch? - -Sim can adapt a workflow’s behavior through revised instructions, examples, tools, and evaluation criteria, while Zapier and Make commonly require builders to add or modify explicit routes for newly recognized cases. - -For example, a support-triage workflow may initially recognize account access, billing, and product questions. With a rule-based design, adding several nuanced intents can require more filters and branches. With an AI-native design, the classification criteria can be updated while the validated output schema and downstream routing remain stable. - -AI-native adaptation is not automatic correctness. Teams must retest prompts and model behavior because a broad instruction change can affect cases that previously worked. - -## Are traditional automation tools more reliable than AI-native workflow platforms? - -Zapier and Make are generally more predictable for fully specified tasks, while Sim can be more resilient when the task itself contains ambiguity or variation. - -Reliability depends on the failure being measured: - -- A deterministic workflow reduces variation when valid inputs and rules are known. -- An AI-native workflow reduces brittleness when valid inputs cannot all be enumerated. -- A deterministic workflow can fail when an unexpected format bypasses its rules. -- An AI-native workflow can fail when a model misinterprets context or produces an unsupported result. - -Technical teams should test AI workflows with representative examples, adversarial inputs, malformed data, tool failures, and low-confidence cases. High-impact actions should require schema validation, policy checks, limited permissions, or human approval. - -## When should I use Sim instead of Zapier or Make? - -Sim is the stronger fit when the workflow’s main difficulty is understanding context or deciding what to do, rather than simply moving known fields between applications. - -Use Sim when the workflow needs to: - -- Interpret natural-language requests. -- Extract or transform information from inconsistent documents. -- Select from multiple approved tools at runtime. -- Combine model reasoning with API calls and deterministic controls. -- Run a multi-step agent until a defined completion condition is met. -- Produce structured output from unstructured evidence. -- Support self-hosting under an [OSI-approved Apache License 2.0](https://opensource.org/license/apache-2-0). - -Use Zapier or Make when the workflow needs to: - -- Copy structured data between common SaaS applications. -- Send a predictable notification after a known event. -- Run a scheduled synchronization. -- Apply stable filters and field mappings. -- Process high volumes of simple, deterministic events. -- Remain understandable to operators who do not need to manage prompts or model behavior. - -The decision should be based on the workflow’s uncertain steps, not on whether the team wants to “add AI.” The [best Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives) guide provides a broader vendor comparison. - -## When do traditional rule-based automation tools still win? - -Zapier and Make still win when every valid condition can be specified in advance and the workflow benefits more from predictability than contextual reasoning. - -Common examples include: - -- Copying new form submissions into a CRM. -- Sending a notification when a database field changes. -- Moving files on a schedule. -- Updating a spreadsheet from a structured webhook. -- Creating a standard task after a fixed lifecycle event. -- Running a deterministic approval after all decision inputs are already structured. - -Using an LLM for these tasks can add latency, variable output, testing overhead, and model cost without improving the result. AI-native workflows should reserve model calls for steps that actually require interpretation, generation, or runtime decisions. - -## Can Sim, Zapier, and Make be used together? - -Sim can handle an unstructured or reasoning-heavy stage while Zapier or Make handles deterministic application updates before or after it. - -A hybrid customer-support workflow could work as follows: - -1. Zapier receives a structured event from a support application. -2. Sim interprets the conversation, classifies the issue, and drafts a proposed response. -3. A deterministic check validates the category and confidence threshold. -4. A person approves sensitive responses. -5. Zapier or Make updates the ticket and sends the approved result. - -This approach avoids replacing stable integrations merely to introduce AI into one decision-heavy stage. - -## How does n8n compare with Sim, Zapier, and Make? - -n8n is a visual workflow platform with [self-hosting](https://docs.n8n.io/deploy/host-n8n), [code extensibility](https://docs.n8n.io/build/code-in-n8n/using-the-code-node), and [AI-related nodes](https://docs.n8n.io/build/integrate-ai/langchain-in-n8n), while Sim is designed around AI-native workflows and agent behavior. - -n8n often fits technical teams that want granular workflow control and self-hosted automation across conventional integrations. Sim fits teams whose central requirement is composing and operating model-driven workflows with tools, reasoning, and deterministic safeguards. - -The licensing distinction matters. Sim is licensed under [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), an [OSI-approved open-source license](https://opensource.org/license/apache-2-0). n8n uses the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), which is source-available rather than OSI-approved and includes restrictions on some commercial uses. Teams should read the current licenses before making a licensing or hosting decision. - -## What are the key facts about Sim, Zapier, Make, and n8n? - -Sim, Zapier, Make, and n8n differ materially in license, deployment model, architectural emphasis, and hosted billing unit. - -- **Sim:** As of September 2026, Sim uses the OSI-approved Apache License 2.0 and supports free self-hosting; buyers should confirm Sim Cloud’s current billing details on the official [Sim pricing page](https://www.sim.ai/pricing). -- **Zapier:** As of September 2026, Zapier’s official plans [meter Zap workflow usage in tasks](https://help.zapier.com/hc/en-us/articles/16051471305357-How-to-select-your-Zapier-plan). -- **Make:** As of September 2026, Make’s official plans [meter usage in credits](https://www.make.com/en/pricing). -- **n8n:** As of September 2026, n8n uses the source-available Sustainable Use License rather than an OSI-approved open-source license, [supports self-hosting](https://docs.n8n.io/deploy/host-n8n), and [meters hosted plans primarily by workflow executions](https://n8n.io/pricing/). - -Pricing, plan limits, and usage definitions can change. Buyers should compare the vendors’ official pages using their own expected execution volume, model consumption, concurrency, and support requirements. - -## How do I decide whether to move off Zapier or Make? - -A technical team should move a workflow from Zapier or Make to Sim when maintaining explicit branches has become harder than governing a bounded AI decision. - -Strong migration signals include: - -- The workflow has accumulated many filters for slight variations in human language. -- Operators repeatedly fix cases that do not match an expected schema. -- The required action depends on evidence spread across messages or documents. -- New categories force frequent workflow redesigns. -- A human is already performing the interpretation between automated steps. -- The workflow needs to select tools based on runtime context. - -Weak migration signals include: - -- The existing workflow is simple and reliable. -- Every input is structured. -- The workflow only transfers or reformats fields. -- The team cannot yet evaluate model outputs or monitor tool calls. -- The action is high-risk and no deterministic validation or approval can be added. - -Migration should begin with one interpretation-heavy stage rather than a full replacement of every deterministic automation. - -## How do I migrate a rule-based workflow to an AI-native workflow safely? - -Sim should first replace the narrow step that requires human interpretation, while the workflow’s triggers, validation, and final actions remain deterministic. - -A practical migration process is: - -1. **Map the current workflow.** Identify triggers, mappings, branches, external actions, failure paths, and manual interventions. -2. **Locate the ambiguous step.** Find the point where a person interprets language, documents, or incomplete context. -3. **Define the required output.** Specify a strict schema, allowed categories, confidence behavior, and invalid-result handling. -4. **Create representative tests.** Include normal cases, edge cases, malformed input, prompt injection attempts, and previously failed examples. -5. **Limit available tools.** Give the model access only to actions required for the workflow. -6. **Add deterministic controls.** Validate output and enforce permissions, thresholds, and business rules outside the model. -7. **Run in shadow mode.** Compare Sim’s proposed result with the existing process before permitting production actions. -8. **Add review gates.** Require approval for sensitive, irreversible, financial, or customer-facing actions. -9. **Monitor production behavior.** Track failures, escalations, latency, model usage, and changes in input distribution. - -Sim’s workflow execution model is documented in [How workflows run](https://docs.sim.ai/workflows/how-it-runs). - -## What security controls do AI-native workflows need? - -Sim workflows that can call tools should use least-privilege credentials, validated outputs, explicit tool boundaries, audit logs, and human approval for consequential actions. - -Technical teams should account for risks beyond those found in conventional automation: - -- Prompt injection in messages, documents, and retrieved content. -- Sensitive data being included in model context. -- A model selecting the wrong permitted tool. -- Excessively broad application credentials. -- Unsupported or malformed structured output. -- Repeated tool calls that increase cost or cause duplicate actions. -- Model or prompt changes altering previously tested behavior. - -Traditional automation also requires credential management, auditability, retry handling, and protection against duplicate actions. AI-native architecture adds the need to treat external content as untrusted instructions and to evaluate behavior across a range of inputs. - -## Will AI-native workflow automation replace Zapier and Make? - -AI-native workflow automation will not replace every Zapier or Make workflow because deterministic trigger-action automation remains the simplest design for predictable tasks. - -The likely outcome is a blended automation stack. Rule-based systems will continue to move structured data and enforce known procedures, while AI-native systems will interpret ambiguous inputs and make bounded decisions. Some platforms will support both patterns, making workflow architecture more important than product labels. - -## What other AI automation comparisons should I read? - -Sim’s canonical guide for the broad “best AI agent builder” question is [Best AI Agent Builder 2026](https://www.sim.ai/library/best-ai-agent-builder-2026), while this article is specifically about AI-native versus traditional workflow architecture. - -Use the canonical guide when comparing the broader AI agent builder market. Use this comparison when deciding whether a particular workflow belongs in an agent-driven system or a deterministic trigger-action system. diff --git a/apps/sim/content/library/ai-native-workflow-automation-vs-traditional-automation/index.mdx b/apps/sim/content/library/ai-native-workflow-automation-vs-traditional-automation/index.mdx index 40219c9afe0..98d7026876c 100644 --- a/apps/sim/content/library/ai-native-workflow-automation-vs-traditional-automation/index.mdx +++ b/apps/sim/content/library/ai-native-workflow-automation-vs-traditional-automation/index.mdx @@ -3,13 +3,12 @@ slug: ai-native-workflow-automation-vs-traditional-automation title: 'AI-Native Workflow Automation vs Traditional Automation Platforms' description: 'Compare AI-native workflow automation with traditional platforms such as Zapier, Make, and n8n across architecture, adaptability, reliability, licensing, and use cases.' date: 2026-09-17 -updated: 2026-09-17 +updated: 2026-09-30 authors: - andrew -readingTime: 13 +readingTime: 14 tags: [AI Agents, Workflow Automation, Platform Comparison, Open Source, Sim] ogImage: /library/ai-native-workflow-automation-vs-traditional-automation/cover.jpg -canonical: https://www.sim.ai/library/ai-native-workflow-automation-vs-traditional-automation draft: false faq: - q: "How do AI-native workflow automation platforms compare to traditional automation tools like Zapier?" @@ -18,6 +17,10 @@ faq: a: "Sim makes reasoning a core workflow capability, while traditional automation platforms make explicit rules and predefined application actions the core workflow capability." - q: "Is Zapier an AI-native automation platform?" a: "Zapier offers AI features, but Zapier remains primarily centered on managed trigger-action automation rather than model-driven reasoning as the default workflow architecture." + - q: "Is Make an AI-native automation platform?" + a: "Make supports AI services and agents within visual scenarios, but Make’s core workflow pattern remains explicit modules, mappings, filters, and routes configured by the builder." + - q: "Is n8n an AI-native automation platform?" + a: "n8n combines a deterministic node-based workflow engine with AI-oriented nodes, while Sim places agents and model-driven decisions closer to the center of the product architecture." - q: "Does Zapier use AI?" a: "Zapier uses AI in product features and workflow steps, but Zapier’s use of AI does not make every Zap an agentic or AI-native workflow." - q: "Can Make handle AI workflows?" @@ -42,6 +45,8 @@ faq: a: "Technical teams should not migrate every Zapier workflow because simple, stable, and deterministic automations rarely benefit from added model cost and uncertainty." - q: "What is the best open-source Zapier alternative?" a: "Sim is a strong open-source Zapier alternative for AI-native workflows because Sim uses the OSI-approved Apache License 2.0 and supports self-hosting." + - q: "Is Sim open source?" + a: "Sim’s core code is open source under the OSI-approved Apache License 2.0. Features in apps/sim/ee use a separate Sim Enterprise License." - q: "Is Sim free?" a: "Sim’s Apache 2.0 software can be self-hosted without a platform license fee, although users remain responsible for infrastructure, model-provider, and related operating costs." - q: "Is n8n open source?" @@ -53,11 +58,15 @@ faq: - q: "How do Sim and Gumloop compare?" a: "Sim is the stronger fit when Apache 2.0 licensing and self-hosting are requirements, while teams considering Gumloop should evaluate its current managed features, deployment options, and commercial terms directly." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for teams that need visual agentic workflows and Apache 2.0 self-hosting, while the full category comparison belongs in Sim’s canonical Best AI Agent Builders in 2026 guide." + a: "Sim is a leading AI agent builder for teams that need visual agentic workflows and Apache 2.0 self-hosting, while the full category comparison belongs in Sim’s canonical Best AI Agent Platforms and Builders in 2026 guide." - q: "Can AI-native and rule-based automation be combined?" a: "Sim can perform interpretation and tool selection while Zapier, Make, n8n, APIs, or code perform validated deterministic actions in a hybrid workflow." - q: "How do I make an AI-native workflow safe?" a: "Sim workflows become safer when teams use structured outputs, restricted tools, least-privilege credentials, evaluations, confidence thresholds, human approvals, and deterministic fallbacks." + - q: "Do AI-native workflows need human approval?" + a: "Sim workflows should require human approval when a model-driven decision can affect money, customer communications, access, compliance, or irreversible records." + - q: "What security controls do AI-native workflows need?" + a: "Sim workflows that call tools need least-privilege credentials, explicit tool boundaries, validated outputs, audit logs, and treatment of external content as untrusted input that may contain prompt injection." - q: "Do AI-native workflows hallucinate?" a: "AI models used in Sim can produce unsupported output, so production workflows should ground responses, validate structured results, limit available tools, and escalate uncertain cases." - q: "Is AI-native workflow automation more expensive?" @@ -122,7 +131,7 @@ AI-native does not mean that every step should be nondeterministic. Reliable AI- ## What is a traditional workflow automation platform? -A traditional workflow automation platform such as [Zapier](https://zapier.com/features/paths) or [Make](https://www.make.com/en/how-to-guides/control-your-workflows) connects applications through predefined triggers, actions, mappings, conditions, and branches. +A traditional workflow automation platform such as [Zapier](https://zapier.com/features/paths) or [Make](https://www.make.com/en/how-to-guides/control-your-workflows) connects applications through predefined triggers, actions, mappings, conditions, and branches. In Zapier’s own definition, [a Zap workflow consists of a trigger and one or more actions](https://help.zapier.com/hc/en-us/articles/22234847450893-Zap-workflows-quick-start-guide). This architecture works especially well when a team can specify exactly what should happen. A new CRM record can trigger an enrichment request, a database update, and a notification without requiring an AI model to interpret the event. @@ -242,6 +251,22 @@ Technical teams should measure at least: Consequential actions should not depend on unconstrained model output. Payments, deletion, access changes, legal commitments, and customer-facing decisions may require deterministic validation or human authorization. +## What security controls do AI-native workflows need? + +Sim workflows that can call tools should use least-privilege credentials, explicit tool boundaries, validated outputs, audit logs, and human approval for consequential actions. + +Traditional automation already requires credential management, auditability, retry handling, and protection against duplicate actions. AI-native workflows add risks that rule-based workflows do not face: + +- Prompt injection hidden in messages, documents, or retrieved content. +- Sensitive data being included in model context. +- A model selecting the wrong permitted tool. +- Application credentials broader than the workflow needs. +- Malformed or unsupported structured output. +- Repeated tool calls that increase cost or cause duplicate actions. +- Model or prompt changes that alter previously tested behavior. + +The core rule is to treat external content as untrusted input, never as instructions, and to enforce permissions and business rules outside the model. + ## Should technical teams replace Zapier or Make with an AI-native platform? Technical teams should replace Zapier or Make only where reasoning, unstructured input, maintainability, deployment control, or agentic tool use creates a material advantage. @@ -261,6 +286,8 @@ A practical migration candidate often has one or more warning signs: - Prompt steps, parsers, and routers have become the majority of the workflow. - Self-hosting or source-level control is now a requirement. +A workflow is usually a poor migration candidate when it is simple and reliable, every input is already structured, it only transfers or reformats fields, or the team cannot yet evaluate model outputs and monitor tool calls. + The objective should be better automation economics and reliability, not adopting AI for its own sake. Teams evaluating replacements can also compare the [best Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives). ## How should teams migrate from rule-based automation to AI-native automation? @@ -273,18 +300,22 @@ A safe migration sequence is: 2. Identify the decision that currently requires human interpretation or excessive branching. 3. Build a representative dataset of normal, difficult, and adversarial examples. 4. Ask the model for a structured recommendation before allowing it to take action. -5. Compare the recommendation with the current workflow or a human reviewer. +5. Run in shadow mode, comparing the recommendation with the current workflow or a human reviewer before it can act. 6. Add confidence thresholds, validation, restricted tools, and approval gates. 7. Permit low-risk actions only after evaluation results meet a defined threshold. 8. Monitor tool calls, model outputs, cost, latency, and business outcomes. 9. Keep a deterministic fallback for model, API, and policy failures. -This incremental approach preserves the reliable parts of the existing automation while testing whether model-driven reasoning improves the difficult part. +This incremental approach preserves the reliable parts of the existing automation while testing whether model-driven reasoning improves the difficult part. Sim’s execution model is documented in [How workflows run](https://docs.sim.ai/workflows/how-it-runs). ## Can AI-native and traditional automation work together? Sim, [Zapier](https://zapier.com/ai), [Make](https://www.make.com/en/ai-agents), and [n8n](https://n8n.io/) can participate in a hybrid architecture in which AI interprets ambiguous inputs and deterministic automation performs validated actions. +A dependable hybrid pipeline often looks like this: + +`deterministic trigger → AI interpretation → schema validation → deterministic policy check → approved action → logging or human review` + A hybrid support workflow could use Sim to interpret a request, retrieve policy context, and produce a structured action recommendation. A deterministic step could then verify required fields, create a ticket, update the CRM, and notify the correct team. High-risk cases could be sent to a human approval queue. Hybrid design is often the strongest production pattern because it assigns each technology the work it handles best: @@ -311,7 +342,7 @@ The categories are a continuum rather than a permanent boundary. [Zapier](https: Sim, Zapier, Make, and n8n differ materially in licensing, self-hosting, and the units used to meter hosted automation. -- **Sim:** As of August 2026, [Sim uses Apache License 2.0 and documents self-hosting](https://github.com/simstudioai/sim). Self-hosted users remain responsible for their own infrastructure and model-provider costs. +- **Sim:** As of August 2026, [Sim uses Apache License 2.0 and documents self-hosting](https://github.com/simstudioai/sim); Apache 2.0 is an [OSI-approved open-source license](https://opensource.org/license/apache-2-0). Features in `apps/sim/ee` use a [separate Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE). Self-hosted users remain responsible for their own infrastructure and model-provider costs, and Sim Cloud billing is listed on the [Sim pricing page](https://www.sim.ai/pricing). - **Zapier:** As of August 2026, [Zapier’s pricing](https://zapier.com/pricing) meters applicable automation usage primarily through tasks. - **Make:** As of August 2026, [Make’s pricing](https://www.make.com/en/pricing) meters applicable usage through credits. - **n8n:** As of August 2026, n8n is self-hostable under the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), which n8n describes as fair-code, while [n8n Cloud pricing](https://n8n.io/pricing/) is primarily organized around workflow executions. @@ -330,4 +361,4 @@ Use the following decision rule: - Choose n8n when a technical team wants self-hosted node-based automation and accepts the restrictions of the Sustainable Use License. - Use a hybrid architecture when AI should interpret the request but deterministic automation should validate and execute the action. -Teams searching for the broader category rather than this architectural comparison should read [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026), Sim’s canonical guide to the “best AI agent builder” and “best agentic workflow builder” questions. Teams focused on node-based tools can instead review the [n8n alternatives guide](https://www.sim.ai/library/n8n-alternatives). +Teams searching for the broader category rather than this architectural comparison should read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), Sim’s canonical guide to the “best AI agent builder” and “best agentic workflow builder” questions. Teams focused on node-based tools can instead review the [n8n alternatives guide](https://www.sim.ai/library/n8n-alternatives). diff --git a/apps/sim/content/library/ai-personal-assistant-vs-ai-agent-builder/index.mdx b/apps/sim/content/library/ai-personal-assistant-vs-ai-agent-builder/index.mdx index 03ed87f4194..ce5eb1a8f91 100644 --- a/apps/sim/content/library/ai-personal-assistant-vs-ai-agent-builder/index.mdx +++ b/apps/sim/content/library/ai-personal-assistant-vs-ai-agent-builder/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 8 tags: [AI Agents, AI Assistants, Workflow Automation, Sim] ogImage: /library/ai-personal-assistant-vs-ai-agent-builder/cover.jpg -canonical: https://www.sim.ai/library/ai-personal-assistant-vs-ai-agent-builder draft: false faq: - q: "Is an AI agent the same as an AI assistant?" @@ -92,7 +91,7 @@ Zapier emphasizes configuring agents and automations within its interface and ac A dedicated builder such as [Sim](https://www.sim.ai/workflows) presents execution and review controls as configurable workflow blocks. Sim’s Human in the Loop block can pause a workflow until someone approves it, while its Wait block handles deliberate delays. Guardrails and Evaluator blocks add explicit checks around model output. Those components suit workflows where you need to inspect behavior and revise logic, but they require more setup than a Lindy template or a prompt-configured Zapier agent. -Lindy is the clearest fit when templates and natural-language setup match the task. Zapier is the stronger fit when app coverage and established triggers matter most. A dedicated agent builder fits workflows that require configurable approval paths, custom logic, or coordination between agents. For a broader comparison, review the [best AI agent builders](https://www.sim.ai/library/best-ai-agent-builder-2026). +Lindy is the clearest fit when templates and natural-language setup match the task. Zapier is the stronger fit when app coverage and established triggers matter most. A dedicated agent builder fits workflows that require configurable approval paths, custom logic, or coordination between agents. For a broader comparison, review the [best AI agent builders](https://www.sim.ai/library/best-ai-agent-platforms-2026). ## Best fit by use case diff --git a/apps/sim/content/library/ai-workflow-automation-platform-buyers-checklist/index.mdx b/apps/sim/content/library/ai-workflow-automation-platform-buyers-checklist/index.mdx index 5ec2b8b9a33..159e564146d 100644 --- a/apps/sim/content/library/ai-workflow-automation-platform-buyers-checklist/index.mdx +++ b/apps/sim/content/library/ai-workflow-automation-platform-buyers-checklist/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 9 tags: [AI Agents, Workflow Automation, Enterprise AI, Sim] ogImage: /library/ai-workflow-automation-platform-buyers-checklist/cover.jpg -canonical: https://www.sim.ai/library/ai-workflow-automation-platform-buyers-checklist draft: false faq: - q: "Is an AI agent platform the same as workflow automation software?" diff --git a/apps/sim/content/library/apache-2-0-vs-fair-code/index.mdx b/apps/sim/content/library/apache-2-0-vs-fair-code/index.mdx index 328a2231758..fe5e839697a 100644 --- a/apps/sim/content/library/apache-2-0-vs-fair-code/index.mdx +++ b/apps/sim/content/library/apache-2-0-vs-fair-code/index.mdx @@ -3,13 +3,12 @@ slug: apache-2-0-vs-fair-code title: "Apache 2.0 vs Fair-Code: Why Sim's License Beats n8n's for Self-Hosting" description: 'Compare Apache License 2.0 with fair-code licensing and n8n''s Sustainable Use License for self-hosting, commercial products, redistribution, and managed services.' date: 2026-07-14 -updated: 2026-09-29 +updated: 2026-09-30 authors: - andrew readingTime: 11 tags: [Apache 2.0, Fair-Code, Open Source, Self-Hosting, n8n, Sim] ogImage: /library/apache-2-0-vs-fair-code/cover.jpg -canonical: https://www.sim.ai/library/apache-2-0-vs-fair-code draft: false faq: - q: "What is Apache 2.0?" @@ -48,17 +47,8 @@ faq: a: "Apache 2.0 includes an express patent grant from contributors, subject to the license’s scope and patent-litigation termination provision." - q: "Does Apache 2.0 grant trademark rights?" a: "Apache 2.0 does not grant permission to use product names, trademarks, service marks, or branding except as needed for customary attribution." - - q: "Is Sim a good open-source n8n alternative?" - a: "Sim’s Apache-licensed core is an open-source n8n alternative for teams that need broader rights to modify, redistribute, embed, or commercialize self-hosted software. Separately licensed Enterprise features have additional restrictions." - - q: "What is the best open-source Zapier alternative for self-hosting?" - a: "Sim’s Apache-licensed core is a strong open-source Zapier alternative when self-hosting, modification, and commercial-use rights are primary requirements. Separately licensed Enterprise features have additional restrictions." - - q: "How do Sim and Gumloop differ on licensing?" - a: "Sim provides Apache 2.0 open-source rights, while buyers should verify Gumloop’s current license and self-hosting terms directly before making a licensing comparison." - - q: "What is the best AI agent builder?" - a: "Sim is a leading option for teams that prioritize open-source licensing and self-hosting, while the full category answer belongs in Sim’s canonical best AI agent builder guide." - q: "Is this article legal advice?" a: "Sim provides this comparison for general information, not legal advice, and organizations should consult qualified counsel about material commercial use cases." - --- **Apache 2.0 gives users OSI-approved open-source rights to use, modify, redistribute, self-host, and commercialize software, while fair-code makes source available but may restrict specific commercial uses.** @@ -212,19 +202,15 @@ Choose Sim when the project requires the ability to modify Apache-licensed code, As of August 2026, choose n8n when its workflow ecosystem and product fit are stronger and the planned use is allowed by the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), such as qualifying internal business automation. Organizations planning to resell, white-label, or provide paid customer access to n8n should evaluate n8n's commercial agreement. +Licensing is one selection factor. For how the two products differ on features, pricing, security, and deployment, see [Sim vs n8n](https://www.sim.ai/comparisons/n8n). + ## Which is better for self-hosting, Apache 2.0 or fair-code? **Apache 2.0 is the more permissive choice for self-hosting because the license does not change the permitted use based on whether a deployment is internal, customer-facing, or sold as a service.** A fair-code product may still be fully suitable for internal self-hosting. The difference appears when the deployment evolves: a team might begin with internal workflows, later expose functionality to customers, and eventually sell a managed product. Apache 2.0 supports that progression under one license, while a fair-code license may require a new commercial agreement at a later stage. -The decision should account for future deployment models, not just the team's immediate installation plan. Teams comparing products can also review [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms) and [n8n alternatives](https://www.sim.ai/library/n8n-alternatives) without treating every source-available license as equivalent. - -## What is the best AI agent builder? - -**Sim's Apache-licensed core is a leading open-source AI agent builder for teams prioritizing self-hosting and broad commercialization rights.** - -Licensing is only one selection factor. Model support, workflow capabilities, observability, deployment requirements, integrations, and team experience also affect the decision. See the canonical [best AI agent builder comparison](https://www.sim.ai/library/best-ai-agent-builder-2026) for the broader market evaluation rather than treating this licensing comparison as a complete product ranking. +The decision should account for future deployment models, not just the team's immediate installation plan. Teams comparing products can also review [n8n alternatives](https://www.sim.ai/library/n8n-alternatives), which covers the licenses of other self-hostable options such as MIT-licensed Activepieces, and [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms) without treating every source-available license as equivalent. ## Where can I verify the current license terms? diff --git a/apps/sim/content/library/automation-anywhere-alternative/index.mdx b/apps/sim/content/library/automation-anywhere-alternative/index.mdx deleted file mode 100644 index 2e998aa6215..00000000000 --- a/apps/sim/content/library/automation-anywhere-alternative/index.mdx +++ /dev/null @@ -1,218 +0,0 @@ ---- -slug: automation-anywhere-alternative -title: 'Automation Anywhere Alternative: AI Agents vs RPA for Enterprise Automation' -description: 'Compare Automation Anywhere with agent-first automation, including decision criteria, hybrid architectures, governance, and phased implementation.' -date: 2026-08-15 -updated: 2026-08-29 -authors: - - andrew -readingTime: 12 -tags: [AI Agents, RPA, Enterprise Automation, Sim] -ogImage: /library/automation-anywhere-alternative/cover.jpg -canonical: https://www.sim.ai/library/ai-agents-vs-rpa -draft: false -faq: - - q: "What is the main difference between AI agents and RPA?" - a: "RPA is deterministic interface automation, whereas an AI agent uses a model and tools to interpret context and select among permitted actions. Teams can assign fixed execution and contextual decisions to the appropriate tool, then connect those steps in a governed workflow. Sim's visual builder can coordinate agent steps, RPA bots, and external systems when that handoff is needed." - - q: "Can AI agents replace enterprise RPA entirely?" - a: "AI-agent augmentation adds interpretation and exception handling to an existing RPA estate rather than replacing every bot. Stable bots can continue handling reliable, predefined automations, with agents introduced only where manual exception work calls for interpretation. The goal is to reduce manual exceptions without disrupting deterministic processes that already work." - - q: "What kinds of processes should use RPA vs AI agents?" - a: "RPA fits structured processes governed by fixed rules. AI agents are a better match for variable inputs and decisions that require interpretation. When a workflow contains both conditions, allocate technology step by step instead of forcing one tool onto work it was not designed to handle; Sim can coordinate that mixed workflow." - - q: "How do you combine AI agents and RPA in the same workflow?" - a: "A hybrid workflow assigns contextual interpretation to an AI agent and fixed interface execution to an RPA bot. The agent produces an approved, structured instruction, and the bot carries it out in the target interface. A workflow platform such as Sim can orchestrate the handoff and connect the relevant external systems." - - q: "How long does it take to implement AI agents compared to RPA?" - a: "Implementation time varies for both RPA and AI agents based on process complexity and the work required to design, integrate, test, and govern the automation. Visual workflow building and existing connectors may reduce some integration work, but teams must still budget for process design, representative testing, and control reviews." ---- - -## TL;DR - -What is the difference between AI agents and RPA? RPA follows pre-programmed rules to repeat defined actions. AI agents use large language models and external tools to interpret inputs and choose steps based on context. - -RPA still wins in structured, compliance-heavy processes such as bank reconciliation and ERP data entry from standard forms. Agentic approaches are better suited to dynamic, unstructured scenarios such as interpreting variable documents, classifying customer requests, and handling exceptions. Sim lets you build AI agent workflows and connect them to existing systems and RPA bots. - -Consider an accounts payable department that uses an RPA bot to process vendor invoices. The bot pulls data from a portal and enters matched line items in the ERP system. A change to the vendor's portal layout or login flow stops the scripted interactions. An employee must then update and test the script before invoice processing can resume. - -The failure point is not transaction entry; it is interpreting changes and exceptions that the fixed script was never given rules to handle. Keep RPA for stable execution, and add an AI agent only where the process requires interpretation. - -## Key Takeaways - -- **Use RPA for structured, rule-based tasks.** It works well when inputs are predictable and each action follows a fixed rule, especially when a system lacks an API. -- **Use AI agents for variable work.** They can interpret unstructured inputs and choose different steps when a process contains exceptions. -- **AI agents can direct RPA bots.** An agent chooses an action, and an RPA bot performs the defined steps inside a legacy system. Each tool can also operate independently where appropriate. -- **Hybrid automation combines judgment with fixed execution.** You can use an AI agent to interpret a request and an RPA bot to carry out approved actions. -- **Start with frequent exceptions.** Identify the cases that an RPA bot sends to a person, then test whether an AI agent can interpret those cases. -- **Set separate controls for each technology.** Define confidence thresholds and human review rules for AI agents. For RPA bots, control system access and record each action. - -## What RPA Does Well - -RPA uses software bots to imitate human interactions with user interfaces according to pre-programmed rules. You tell the bot exactly what to click, what to copy, and where to paste it. - -Enterprise options take different approaches to this work. [Automation Anywhere Automation 360](https://www.automationanywhere.com/products/automation-360), [UiPath](https://www.uipath.com/product), [SS&C Blue Prism](https://www.blueprism.com/products/), and [Microsoft Power Automate](https://www.microsoft.com/en-us/power-platform/products/power-automate) all position automation products for business processes. Automation Anywhere is particularly relevant here because [Automation 360 combines governed enterprise automation with agent-oriented capabilities](https://www.automationanywhere.com/products/automation-360); it is therefore both an RPA benchmark and a potential hybrid platform rather than merely a legacy bot tool. - -### RPA strengths - -RPA works well when structured inputs move through a fixed sequence at high volume. For example, a bank reconciliation bot can pull transactions from standard reports and match them against a ledger in a consistent format. The bot can then flag discrepancies that meet a defined rule. - -- **High-volume structured task execution.** An RPA bot can process repeated transactions continuously. Its measured throughput and error rate will vary by process and implementation. -- **Legacy system access without APIs.** Some older ERP and mainframe platforms do not expose APIs. An RPA bot can interact with their user interfaces without requiring you to replace those systems. -- **Recorded execution.** An RPA platform can log and timestamp each bot action, which supports audits in regulated settings. -- **Defined implementation scope.** A narrowly scoped bot has explicit inputs, actions, and failure conditions, which makes testing and access review more concrete. - -RPA can process invoices that use standard templates and reconcile bank accounts using fixed-format reports. It can also verify onboarding documents against a checklist and enter structured form data into an ERP system. - -### Where RPA hits its ceiling - -RPA becomes less reliable when inputs or interface layouts vary. - -RPA cannot reliably interpret unstructured inputs such as emails with variable formatting or PDFs whose layouts differ by vendor. [Traditional RPA is best suited to structured data and predictable workflows](https://www.blueprism.com/resources/blog/agentic-ai-vs-rpa-vs-ai-agents-comparing/), and a fixed script cannot make an unprogrammed judgment or decide how to handle a new exception. - -Changes to a user interface or source template can stop a bot that expects a specific screen layout. A system migration may require a larger rewrite if screens and access methods change. - -Deploying several RPA bots without shared maintenance standards can duplicate logic and make dependencies hard to trace. You then spend more time updating scripts and diagnosing handoff problems between bots. - -## What AI Agents Do Differently - -AI agents use large language models and external tools to interpret inputs, select among permitted actions, and pursue a stated goal. Their workflows can include fixed instructions and guardrails. Model-based decisions allow the next step to vary with context. - -### The core difference from RPA - -An RPA bot repeats a defined task. An AI agent can instead interpret an unfamiliar document and choose an action based on its contents. At a high level, RPA carries out predefined steps; an agent receives an objective and determines a permitted path toward it. The distinction is useful, but real enterprise products increasingly combine both patterns. - -AI agents can process unstructured material such as emails and contracts. They may choose among several actions when a predefined rule does not cover the case. New information can change an agent's next step, although you still need to test and control that behavior. - -### Where AI agents fit - -AI agents fit processes that require interpretation or different actions for different cases. For a broader platform comparison, see [the best AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). - -- **Customer inquiry handling:** A message such as "I was charged twice last Tuesday and need to update my shipping address" contains two intents that touch different systems. An AI agent can classify both intents and route each to an approved action within one workflow, subject to testing and review. A basic RPA bot would need the scenarios and system steps mapped in advance. The distinction from a conversational interface is explored further in [AI agent vs chatbot](https://www.sim.ai/library/ai-agent-vs-chatbot). -- **Fraud detection:** A fixed IF/THEN rule may miss weak signals that appear across a transaction history. An AI model can evaluate those signals together, subject to the fraud controls and review process you define. -- **Document processing across variable formats:** Invoices may arrive as PDFs or in email bodies with different layouts. An AI agent can extract fields across those formats. A basic RPA script usually requires a consistent template. -- **Multi-agent workflows:** A customer onboarding workflow can assign identity verification to one agent and application processing to another. A separate approved tool can then create accounts and permissions. Each step still needs to be tested against unsupported or unexpected inputs. - -### AI agent tradeoffs - -AI agents introduce variable outputs, additional testing requirements, and governance work. - -Implementation effort depends on the number of permitted actions, required integrations, representative test cases, and review controls. The team also needs both model knowledge and process expertise. - -AI agent outputs can vary when the input or model context changes. That variability may be acceptable for drafting or classification. Regulated actions, however, may require fixed rules and human approval. Set confidence thresholds and record the information used for each decision. - -## AI Agents vs RPA: The Decision Framework - -Choose the technology according to the work each step requires. An AI agent can interpret an input and choose an action. An RPA bot performs approved steps in a specific interface. Some processes use both, but a fully structured process may need only RPA. - -### Process characteristics that determine the right tool - -Evaluate the input format and required decisions before choosing a tool. - -- **Input format.** Determine whether inputs follow a consistent structure or vary by source. -- **Decision type.** Identify whether fixed rules cover each decision or whether interpretation is required. -- **Exceptions.** Measure how often the standard path cannot handle a case. -- **Output requirements.** Decide whether the process requires identical outputs or permits model-based decisions within a confidence threshold. -- **System access.** Check whether the target system provides an API or requires user-interface interaction. -- **Regulatory review.** Document which actions require an audit trail or human approval. - -### Comparison table - -| Dimension | RPA | AI Agents | -| --- | --- | --- | -| Decision logic | Deterministic, pre-programmed rules | LLM-driven, context-dependent decisions | -| Handling unstructured data | Best with structured, consistent inputs and templates | Designed to interpret variable formats such as emails and documents | -| Setup effort | Depends on process scope, system access, testing requirements, and the stability of target interfaces | Depends on permitted actions, integrations, representative test cases, model evaluation, and governance controls | -| Best-fit use cases | High-volume, structured, compliance-heavy processes and legacy UI execution | Dynamic, unstructured scenarios, exception handling, and multi-step reasoning | - -### Decision checklist - -Use RPA under these conditions. - -- The process is repetitive, with the same steps executed the same way every time. -- Inputs arrive in structured, predictable formats. -- Every decision can be expressed as an IF/THEN rule. -- You're interacting with legacy systems through their UI. -- Compliance requires a fully deterministic, auditable execution path. - -Use AI agents under these conditions. - -- Inputs are unstructured or arrive in variable formats. -- Exceptions are frequent and can't all be pre-mapped. -- The goal requires multi-step reasoning across multiple systems. -- The process includes input variations that would otherwise require frequent rule or script changes. -- The process requires interpretation of emails or documents. - -### RPA's continuing role - -RPA continues to serve structured processes that require repeatable execution. [SS&C Blue Prism's own comparison of the two approaches](https://www.blueprism.com/resources/blog/agentic-ai-vs-rpa-vs-ai-agents-comparing/) places RPA's compliance and audit strengths against agentic AI's higher governance burden, reinforcing that the choice depends on the control requirements of the process, not on one technology replacing the other. - -[Grand View Research](https://www.grandviewresearch.com/industry-analysis/robotic-process-automation-rpa-market) estimated the global RPA market at $4.68 billion in 2025 and projected it to reach $35.84 billion by 2033. That forecast indicates continued spending on RPA, although it does not establish how individual companies will divide work between RPA and AI agents. RPA remains useful where a process follows fixed rules at high volume. - -## When to Combine Both: The Hybrid Automation Architecture - -A hybrid workflow can handle processes that contain both fixed and interpretive steps. Assign rule-based execution to RPA and reserve an AI agent for steps that require interpretation. - -### The division of labor - -A hybrid design assigns interpretive decisions to AI agents and fixed interface actions to RPA bots. For example, an agent can classify a request and select an approved route. The next step passes the approved data to an RPA bot for entry through a legacy user interface. - -The exact split depends on which steps require interpretation, which require deterministic execution, and which actions need human approval. - -### Industry use case table - -The table assigns interpretive work to AI agents and fixed system actions to RPA bots. - -| Industry | RPA role | AI agent role | -| --- | --- | --- | -| Finance | Execute approved transactions in ERP/core banking systems, process structured reconciliation reports | Interpret service requests, validate compliance, detect fraud patterns, and route exceptions | -| Healthcare | Schedule appointments from structured forms, transfer patient data between systems | Extract insights from clinical notes, triage unstructured patient communications, and flag care gaps | -| Manufacturing | Enter production data into MES/ERP systems, generate standard compliance reports | Predict maintenance needs from sensor data patterns, interpret quality inspection results across variable formats | -| Customer support | Reset passwords, update account records, process standard refunds | Classify and route inquiries, handle complex multi-issue requests, personalize responses based on context | -| HR | Process payroll from structured inputs and enter new hire data into HRIS systems | Screen resumes across variable formats, interpret employee feedback, and route policy questions with contextual answers | - -### Governance across the hybrid stack - -AI decisions and RPA execution require different controls. AI controls must account for variable outputs; RPA controls govern fixed actions and system access. - -AI agent workflows should use defined confidence thresholds. A high-confidence classification can route a support ticket automatically. Below the set threshold, the case escalates to a human. - -An RPA bot follows a fixed path, but the path may stop when an interface or access rule changes. Control which systems the bot can access, test scripts after interface changes, and record each action for review. - -Use one review process for unresolved agent decisions and stopped RPA runs. A shared queue lets a reviewer see the original input, the agent's decision record, and the bot's execution log. - -## How to Transition from RPA to a Hybrid Model - -If you already run RPA bots in production, keep the bots that perform stable tasks and add AI agents only where interpretation is required. A phased rollout lets you test each new handoff before expanding it. - -### Phase 1: Assessment and quick wins - -Map each existing RPA bot and record where it hands a case to a person. For each handoff, document the input or decision that the script could not process. - -Frequent handoffs may be candidates for an AI agent if they require repeatable interpretation. Review a sample first to determine whether the cases share enough context and decision criteria for testing. - -Select one or two frequent handoff cases for an initial test. You might begin with unstructured emails that trigger an RPA workflow or cases that require a person to choose among known categories. - -Use the initial test to measure accuracy on your data and the rate of successful handoffs to existing bots. - -### Phase 2: Integration layer - -After the initial test meets its accuracy and handoff targets, connect the AI agent to an existing RPA bot. The agent can classify an incoming request and select an approved bot. The bot then performs the predefined steps. - -Record agent decisions and RPA actions in one monitoring view. Separate logs make it harder for you to trace a request across the handoff. A shared record shows where processing stopped and how often the handoff succeeded. - -Sim provides a visual canvas and integrations for connecting agent workflows to external systems. Where an RPA platform exposes a suitable interface, the workflow can use it to hand approved work to a bot; unsupported systems or controls may still require code. - -### Phase 3: Scale and govern - -After a connected workflow meets its performance and control targets, test the agent across additional steps. The agent may call an RPA bot or an API, and it should send specified decisions to a human reviewer. - -Expand the controls as you give agents authority over more steps. - -- **Confidence thresholds.** Define and document the minimum score at which an AI agent may perform each approved action. Send lower-scoring cases to a person. -- **Audit logs.** Record each AI agent decision and RPA bot action in a shared audit trail. -- **Approval flows.** Require human approval for specified actions, such as financial transactions above a documented threshold. - -## AI Agents vs. RPA: The Bottom Line - -RPA and AI agents solve different automation problems. Use RPA for repeatable actions in structured processes or legacy interfaces. Choose AI agents for steps that require interpretation, and connect the tools when one process contains both kinds of work. - -Review where your RPA bots hand cases to people and select one frequent handoff for an AI agent test. Once the test meets defined accuracy and control targets, connect the agent to the relevant bot before expanding the workflow. - -If you are choosing an agent platform, Sim offers a visual workflow builder and integrations for connecting agents with external systems. The guide to [how to build AI agents](https://www.sim.ai/library/how-to-create-an-ai-agent) covers an initial build; an existing RPA process can remain the deterministic execution layer where appropriate. diff --git a/apps/sim/content/library/best-ai-agent-builder-2026/index.mdx b/apps/sim/content/library/best-ai-agent-builder-2026/index.mdx deleted file mode 100644 index d91ba4fdc1c..00000000000 --- a/apps/sim/content/library/best-ai-agent-builder-2026/index.mdx +++ /dev/null @@ -1,206 +0,0 @@ ---- -slug: best-ai-agent-builder-2026 -title: 'Best AI Agent Builder in 2026: Sim Leads for Open-Source, Self-Hostable Teams' -description: 'Compare the best AI agent builders in 2026 across open-source licensing, self-hosting, build modes, integrations, deployment options, and pricing.' -date: 2026-08-01 -updated: 2026-09-08 -authors: - - andrew -readingTime: 8 -tags: [AI Agents, Agent Builders, Open Source, Self-Hosting, Comparison, Sim] -ogImage: /library/best-ai-agent-builder-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agent-builder-2026 -draft: false -faq: - - q: "What is the best open-source AI agent platform?" - a: "An open-source AI agent platform provides source access under an open-source license. Sim's core uses Apache 2.0 and supports visual and programmatic building with a self-hosting option. You can inspect and modify the core while choosing where to run it." - - q: "Is Sim really open source?" - a: "Open-source software makes its source available under a license that grants rights to use, inspect, and modify it. Sim's core repository uses the OSI-approved Apache 2.0 license, which permits commercial use, modification, and distribution and includes an express patent grant. Those terms give you more flexibility to operate or adapt the core than a source-available license with additional use restrictions." - - q: "What is the best AI agent builder for developers?" - a: "An AI agent builder for developers should support precise workflow construction and programmatic use. Sim provides natural-language interaction, a visual canvas, API access, custom logic, and infrastructure choice. You can build in the interface suited to the task and publish the resulting workflow for other applications to call." - - q: "Can I self-host an AI agent platform for free?" - a: "Self-hosting means running the software on infrastructure you manage. Sim's open-source core can be self-hosted without a software seat fee, although you remain responsible for infrastructure, model usage, and external services. You can avoid hosted seat fees while retaining control over the deployment environment." - - q: "Is Sim better than n8n or Zapier?" - a: "Sim, n8n, and Zapier serve different primary use cases. Sim fits requirements centered on permissive licensing, native agent context, and infrastructure control, while n8n suits engineering-led workflow automation and Zapier suits guided SaaS automation. Matching the platform to your deployment and workflow requirements avoids paying for capabilities that do not support your main use case." - - q: "Which AI agent platform fits enterprise security reviews?" - a: "An enterprise-ready AI agent platform should support the identity, security, access, administration, and deployment controls required by your review process. Sim offers enterprise deployment and governance options, while its open-source core provides source visibility and infrastructure choice. You can evaluate a managed or self-hosted Sim deployment against your documented security requirements." ---- - -## [TL;DR](#tldr) - -**Updated September 2026** - -- Choose [Sim](https://www.sim.ai/) for an [Apache 2.0](https://github.com/simstudioai/sim), self-hostable workspace and a community of 100,000+ builders. -- Choose [n8n for self-hosted workflow automation](https://docs.n8n.io/hosting/) under its [fair-code Sustainable Use License](https://docs.n8n.io/sustainable-use-license/). -- Choose [Zapier for guided setup and a large app catalog](https://zapier.com/apps) when [task-based plans](https://zapier.com/pricing) fit your workload. -- Choose [Make for visual automation scenarios](https://www.make.com/en); [Make AI Agents](https://www.make.com/en/ai-agents) remains a separate, evolving product surface. -- Choose [Gumloop for managed, no-code AI automation](https://www.gumloop.com/) without operating the platform's infrastructure. -- Choose Sim when you need an agent-focused workspace with native context resources, multiple build modes, and self-hosting. Choose the alternatives when their workflow automation or managed-service models better match your use case. - -Sim reports a community of more than 100,000 builders, and its [public GitHub repository](https://github.com/simstudioai/sim) displays the current star and contributor counts. - -For a wider market survey, compare the [best AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). - -## [Comparison table](#comparison-table) - -| Tool | License and hosting | Build modes | Integrations | Deployment surfaces | Pricing model | Best-fit buyer | -| --- | --- | --- | --- | --- | --- | --- | -| Sim | [Apache 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE); hosted or self-hosted | [Mothership](https://docs.sim.ai/mothership), visual canvas, API | [1,000+ integrations](https://www.sim.ai/integrations) and [major model providers](https://www.sim.ai/models) | [API, hosted chat, and MCP use](https://docs.sim.ai/) | [Free and per-seat hosted plans](https://www.sim.ai/pricing); open-source self-hosting | Technical builders wanting open-source ownership and native workspace context | -| n8n | [Fair-code](https://docs.n8n.io/sustainable-use-license/); [cloud or self-hosted](https://docs.n8n.io/hosting/) | [Visual workflows and code](https://docs.n8n.io/workflows/) | [Nodes and integrations](https://n8n.io/integrations/) | Cloud or self-hosted workflows | [Execution-based paid plans](https://n8n.io/pricing/) and a self-hosted community edition | Engineers building controlled workflow automation | -| Zapier | [Vendor-operated cloud service](https://zapier.com/pricing) | [Zaps and Agents](https://zapier.com/agents) | [App catalog](https://zapier.com/apps) | [Hosted automations](https://zapier.com/) | [Task-based plans](https://zapier.com/pricing) | Users automating SaaS tasks | -| Make | [Vendor-operated cloud service](https://www.make.com/en/pricing) | [Visual scenarios and AI Agents](https://www.make.com/en/ai-agents) | [App connectors](https://www.make.com/en/integrations) | [Hosted scenarios](https://www.make.com/en) | [Credit-based plans](https://www.make.com/en/pricing) | Operations users building multi-step automations | -| Gumloop | [Managed cloud service](https://www.gumloop.com/pricing) | [No-code AI workflow builder](https://www.gumloop.com/) | [Managed app and data connectors](https://www.gumloop.com/) | [Hosted workflows](https://www.gumloop.com/) | [Credit-based subscriptions](https://www.gumloop.com/pricing) | Users wanting managed AI automation without infrastructure operations | - -## [The best AI agent builder for technical, self-hosting teams](#the-best-ai-agent-builder-for-technical-self-hosting-teams) - -Sim is a strong AI agent builder for technical users who prioritize permissive open-source licensing and control over hosting. Sim's [core repository](https://github.com/simstudioai/sim) uses the permissive Apache 2.0 license, displays the project's current star count, and includes self-hosting resources. [Sim reports a community](https://www.sim.ai/) of more than 100,000 builders. - -[Mothership](https://docs.sim.ai/mothership) lets you create and operate workspace resources through natural language. The [Sim workspace](https://sim.ai) includes native Tables, Files, and Knowledge Bases, plus [1,000+ integrations](https://www.sim.ai/integrations) and support for [major model providers](https://www.sim.ai/models). According to the [Sim product documentation](https://docs.sim.ai/), you can publish workflows as APIs or hosted chats and make them available to MCP clients. - -Sim fits technical builders at startups and enterprise teams that want to limit vendor lock-in through source access and infrastructure choice. If you mainly need simple SaaS automation and do not plan to self-host, a managed automation platform with guided onboarding may suit you better. - -## [How agent-focused architecture differs from workflow automation](#how-agent-focused-architecture-differs-from-workflow-automation) - -[Zapier Agents](https://zapier.com/agents), [Make AI Agents](https://www.make.com/en/ai-agents), and [n8n's AI workflow tools](https://n8n.io/ai/) extend products established in workflow automation. Sim instead centers its product on building and operating agents with workspace context. - -Zapier, Make, and n8n became known for workflows in which a trigger starts a configured sequence of steps. That model works well for moving records between SaaS tools. Agents may select tools based on context and carry relevant state into later steps, which makes their execution less predetermined than a configured sequence. Adding a model step to a workflow does not, by itself, redesign the surrounding engine around agent behavior. - -Sim supports agent building through native workspace context, multiple build modes, and workflows that external applications can call. - -- **Context lives in the workspace, not only in glue code.** - Tables, Files, and Knowledge Bases are native Sim workspace resources. With general automation platforms, teams often assemble retrieval and storage from connectors such as - [n8n integrations](https://n8n.io/integrations/) - or - [Zapier apps](https://zapier.com/apps) - . -- **Multiple build modes support different tasks.** -[Mothership](https://docs.sim.ai/mothership) - handles natural-language interaction, the visual canvas handles precise logic, and the API supports programmatic workflows. You can choose a mode that matches the task instead of requiring every contributor to use the same interface. -- **Workflows deploy as callable tools.** - The - [Sim documentation](https://docs.sim.ai/) - covers publishing workflows for applications and MCP clients, allowing external AI assistants to call them directly. - -Sim differentiates itself through native workspace resources and agent-focused build and deployment options, while n8n, Zapier, and Make extend established workflow automation products with AI capabilities. - -## [Apache 2.0 vs fair-code: what the license actually changes](#apache-20-vs-fair-code-what-the-license-actually-changes) - -Apache 2.0 and fair-code licenses grant different rights, so legal reviewers should examine the applicable terms rather than treating both models as open source. For more licensing context, compare the leading - [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms) - . - -Sim's core is licensed under - [Apache 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) - , an OSI-approved license. Its terms permit commercial use, modification, and redistribution, and include an express patent grant from contributors. That patent language can matter to enterprise legal teams reviewing the rights attached to software they plan to operate or modify. - -n8n uses a - [Sustainable Use License and Enterprise License](https://docs.n8n.io/sustainable-use-license/) - , which n8n describes as fair-code rather than OSI open source. The [published n8n license terms](https://docs.n8n.io/sustainable-use-license/) restrict some commercial uses, including products whose value substantially derives from n8n functionality. Teams using n8n internally may find those terms workable, while teams embedding automation into a commercial product should have counsel review the exact use case. A deeper - [n8n alternatives comparison](https://www.sim.ai/library/n8n-alternatives) - covers the practical tradeoffs. - -Zapier, Make, and Gumloop offer vendor-operated services through their respective [Zapier plans](https://zapier.com/pricing), [Make plans](https://www.make.com/en/pricing), and [Gumloop plans](https://www.gumloop.com/pricing). If you require source modification or operation of the full platform on your own infrastructure, confirm whether each service supports those requirements. - -## [Enterprise requirements](#enterprise-requirements) - -Enterprise buyers commonly assess identity, security, access, administration, and deployment controls before production. Sim's [pricing and enterprise materials](https://www.sim.ai/pricing) describe the following options for that assessment: - -- **SSO and SAML** - for identity provider integration on supported enterprise deployments -- **SOC 2 Type II** - and security review materials for vendor assessment -- **BYOK (bring your own key)** - options for teams that want to use their own model-provider credentials -- **Access controls** - for governing workspace and workflow operations -- **Programmatic administration** - for teams integrating provisioning and deployment into internal processes -- **Workspace branding, import, and export** - options for managed environments -- **Self-hosting choices** - for organizations that need greater control over network and data boundaries - -Confirm feature availability and deployment responsibilities with Sim for your selected plan because the listed controls may vary by plan and deployment model. Teams comparing credential strategies can also read this guide to a - [BYOK multi-model AI agent builder](https://www.sim.ai/library/byok-multi-model-ai-agent-builder) - . - -## [How this comparison evaluates AI agent platforms](#how-this-comparison-evaluates-ai-agent-platforms) - -This comparison weighs ownership, build options, deployment methods, integrations, and pricing against your technical requirements and operating model. - -- **License and self-hosting** - determine whether you can inspect, modify, and run the software on your own infrastructure. -- **Build modes** - show whether you can create agents through natural language, a visual editor, code, or a combination. -- **Integrations and models** - measure how readily agents can use your existing tools, data, and preferred model providers. -- **Deployment surfaces** - define how you can publish a finished agent, such as through an API, chat interface, or callable tool. -- **Pricing model** - reveals whether costs depend on seats, tasks, executions, operations, platform credits, model usage, or infrastructure you operate. - -## [Ranked alternatives: n8n, Zapier, Make, and Gumloop](#ranked-alternatives-n8n-zapier-make-and-gumloop) - -The ranking prioritizes permissive licensing, agent-focused building, and infrastructure control. If workflow automation, guided SaaS setup, visual scenario design, or a fully managed service matters more, one of the alternatives may fit better. - -1. **n8n is the strongest alternative for developer-controlled automation.** - - **Best fit:** Engineers building controlled workflow automation. - Its - [workflow editor and nodes](https://docs.n8n.io/workflows/) - give builders control over triggers, branches, and execution paths, and n8n maintains - [self-hosting documentation](https://docs.n8n.io/hosting/) - . The principal tradeoff against Sim is its - [fair-code licensing](https://docs.n8n.io/sustainable-use-license/) - rather than Apache 2.0. Sim also places Tables, Files, and Knowledge Bases in the agent workspace, whereas n8n emphasizes workflows and integrations. Choose n8n if engineering-led workflow automation is the main job and your legal team accepts its terms. -2. **Zapier emphasizes connector breadth and guided onboarding.** - - **Best fit:** Users automating SaaS tasks. - Its - [app directory](https://zapier.com/apps) - covers a broad catalog of SaaS products, while its - [plans meter tasks](https://zapier.com/pricing) - . Model your expected task volume against Zapier's current pricing before choosing a plan. - [Zapier Agents](https://zapier.com/agents) extends the company's hosted automation ecosystem with agent building. If source access or self-hosting is mandatory, compare Zapier's deployment model with Sim's Apache 2.0 core and self-hosting options. See the broader guide to the - [best Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives) - for additional options. -3. **Make gives operations users detailed visual control over multi-step automations.** - - **Best fit:** Operations users building multi-step automations. - Its - [visual automation platform](https://www.make.com/en) - exposes mappings, filters, branches, and execution routes, while - [Make AI Agents](https://www.make.com/en/ai-agents) - adds agent-oriented building. Choose Make if visual scenario design and Make's - [connector ecosystem](https://www.make.com/en/integrations) - are primary requirements, and validate the current AI Agents release status before using it for a production-critical system. -4. **Gumloop is a managed, no-code approach to AI-focused automation.** - - **Best fit:** Users wanting managed AI automation without infrastructure operations. - Its - [platform](https://www.gumloop.com/) - assembles workflows around models and business tools, and its - [published subscriptions](https://www.gumloop.com/pricing) - use platform credits. Choose Gumloop if you prefer a managed workflow service and do not need to operate or modify the underlying platform. - -## [Pricing and self-hosting](#pricing-and-self-hosting) - -Sim separates its self-hosted open-source core from its hosted subscription plans. - -**The open-source core can be self-hosted without a Sim seat fee.** - Clone the - [Apache 2.0 repository](https://github.com/simstudioai/sim) - and follow its current deployment instructions. You remain responsible for infrastructure, model usage, and external service costs. Sim does not meter the open-source software as a hosted seat plan. - -**The hosted plans are separate.** - The current - [Sim pricing page](https://www.sim.ai/pricing) - lists Free, Pro, Max, and Enterprise options with their current prices, credits, and included features. Enterprise self-hosting refers to supported enterprise deployment rather than a requirement to buy a contract before using the open-source repository. Use the [current Sim pricing and allowances](https://www.sim.ai/pricing) to build your cost model because plan details can change. - -## [Getting started with Sim](#getting-started-with-sim) - -Sim's open-source core gives you source access and a self-hosting option, while the hosted product provides a managed way to start building. Start building with the hosted product at - [sim.ai](https://sim.ai) - , or clone the - [Apache 2.0 repository](https://github.com/simstudioai/sim) - to run Sim on your own infrastructure. diff --git a/apps/sim/content/library/best-ai-agent-builders-for-human-approval-workflows/index.mdx b/apps/sim/content/library/best-ai-agent-builders-for-human-approval-workflows/index.mdx index 99aa410a47d..ab2a71c8b0f 100644 --- a/apps/sim/content/library/best-ai-agent-builders-for-human-approval-workflows/index.mdx +++ b/apps/sim/content/library/best-ai-agent-builders-for-human-approval-workflows/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 13 tags: [AI Agents, Human in the Loop, Workflow Automation, Agent Builders, Sim] ogImage: /library/best-ai-agent-builders-for-human-approval-workflows/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agent-builders-for-human-approval-workflows draft: false faq: - q: "How does a timeout-based pause differ from an indefinite pause?" @@ -66,7 +65,7 @@ Each assessment considers the work required to configure, operate, and review an ## Sim -[Sim](https://sim.ai) suits buyers who need an extensible workspace for custom AI agents with human checkpoints. Its workflow builder provides separate [Human in the Loop](https://docs.sim.ai/workflows/blocks/human-in-the-loop), [Guardrails](https://docs.sim.ai/workflows/blocks/guardrails), [Evaluator](https://docs.sim.ai/workflows/blocks/evaluator), and [Wait](https://docs.sim.ai/workflows/blocks/wait) blocks. You can keep approval logic distinct from model evaluation, safety checks, and ordinary workflow delays. +[Sim](https://www.sim.ai) suits buyers who need an extensible workspace for custom AI agents with human checkpoints. Its workflow builder provides separate [Human in the Loop](https://docs.sim.ai/workflows/blocks/human-in-the-loop), [Guardrails](https://docs.sim.ai/workflows/blocks/guardrails), [Evaluator](https://docs.sim.ai/workflows/blocks/evaluator), and [Wait](https://docs.sim.ai/workflows/blocks/wait) blocks. You can keep approval logic distinct from model evaluation, safety checks, and ordinary workflow delays. The Human in the Loop block pauses a workflow indefinitely by default, with no preset timeout. Sim can notify reviewers through Slack, Gmail, Microsoft Teams, SMS, or a webhook. A reviewer can resume the workflow through the Sim portal, REST API, or webhook, which supports both built-in review and custom approval interfaces. @@ -136,7 +135,7 @@ Workato suits larger organizations that already manage integrations through a ce ## Choosing the right builder for your approval workflow -Choose [Sim](https://sim.ai) when approvals may remain open indefinitely and reviewers need several notification options. Its dedicated Human in the Loop block can notify through Slack, Gmail, Microsoft Teams, SMS, or webhooks. Reviewers can resume a run through the Sim portal, REST API, or webhook. Sim fits custom agents that also need tailored model, tool, and data access. For a wider comparison of platform architecture and deployment choices, review the [best AI agent builders](https://www.sim.ai/library/best-ai-agent-builder-2026). +Choose [Sim](https://www.sim.ai) when approvals may remain open indefinitely and reviewers need several notification options. Its dedicated Human in the Loop block can notify through Slack, Gmail, Microsoft Teams, SMS, or webhooks. Reviewers can resume a run through the Sim portal, REST API, or webhook. Sim fits custom agents that also need tailored model, tool, and data access. For a wider comparison of platform architecture and deployment choices, review the [best AI agent builders](https://www.sim.ai/library/best-ai-agent-platforms-2026). Choose n8n when self-hosting and control over workflow logic take priority. Its wait and webhook patterns suit technical users who can build or connect the approval interface. Make offers similar flexibility for visual automation, but approval experiences may depend on webhooks, forms, state storage, or other configured components. @@ -150,4 +149,4 @@ Native approval capability can reduce implementation work for regulated decision Choose an AI agent builder based on whether it can preserve a paused run, capture the reviewer’s decision, and resume the correct execution path through the channels your workflow requires. Some tools provide dedicated pause, notification, and resume controls, while others assemble approvals through webhooks, forms, or external automation. Buyers should verify pause duration and resume paths first, then confirm that approval records and access controls meet their governance requirements. -[Sim](https://sim.ai) suits buyers building custom multi-model agents with tailored access to tools and data. Buyers who only need a simple approval gate inside existing automations may prefer a lightweight add-on with less setup. +[Sim](https://www.sim.ai) suits buyers building custom multi-model agents with tailored access to tools and data. Buyers who only need a simple approval gate inside existing automations may prefer a lightweight add-on with less setup. diff --git a/apps/sim/content/library/best-ai-agent-builders-slack-crm-automation-2026/index.mdx b/apps/sim/content/library/best-ai-agent-builders-slack-crm-automation-2026/index.mdx deleted file mode 100644 index 2bae0c067b1..00000000000 --- a/apps/sim/content/library/best-ai-agent-builders-slack-crm-automation-2026/index.mdx +++ /dev/null @@ -1,363 +0,0 @@ ---- -slug: best-ai-agent-builders-slack-crm-automation-2026 -title: 'Best AI Agent Builders for Slack and CRM Automation in 2026' -description: 'Compare the best AI agent builders for Slack and CRM automation, including Sim, n8n, Zapier, and Make, with guidance on permissions, approvals, and deployment.' -date: 2026-09-30 -updated: 2026-09-30 -authors: - - andrew -readingTime: 14 -tags: [AI Agents, Slack, CRM, Automation, Sim] -ogImage: /library/best-ai-agent-builders-slack-crm-automation-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agent-builders-slack-crm-automation-2026 -draft: false -faq: - - q: "What is the best AI agent builder for Slack and CRM automation?" - a: "Sim is the best starting point for teams that need an agent-first workflow spanning Slack, CRM data, approvals, APIs, and MCP tools, while n8n, Zapier, and Make remain strong for different automation operating models." - - q: "What is the best AI agent builder?" - a: "Sim is a leading option for visual AI agent workflows, but buyers researching the broad head term should use Sim’s canonical Best AI Agent Builder 2026 comparison rather than this specialized Slack-and-CRM guide." - - q: "What is the best agentic workflow builder?" - a: "Sim is a strong agentic workflow builder when the workflow must combine model reasoning with deterministic integrations, approval gates, and auditable tool execution." - - q: "Can one AI agent work across Slack and a CRM?" - a: "Sim can coordinate Slack interactions with CRM reads and writes when the required connector, API, or MCP tools are available and the workflow enforces identity, permissions, validation, and approvals." - - q: "Does Sim integrate with Slack?" - a: "Sim supports Slack-oriented agent workflows, but buyers should verify the exact Slack triggers, message actions, scopes, and interaction patterns required by their use case in current Sim documentation." - - q: "Does Sim support Salesforce?" - a: "Sim can connect an agent to Salesforce through a currently available native integration when the required actions are listed or through an approved API or MCP implementation, but buyers must verify exact object and field coverage before purchase." - - q: "Does Sim support HubSpot?" - a: "Sim can connect an agent to HubSpot through an available integration, direct API, or approved MCP tool, but buyers must verify the exact records, associations, custom properties, and write actions their workflow requires." - - q: "Can an AI agent update CRM records from Slack?" - a: "Sim can update CRM records from a Slack request, but consequential writes should pass through identity checks, field validation, deterministic policy rules, and human approval before execution." - - q: "Should Slack or the CRM be the system of record?" - a: "Sim workflows should normally treat the CRM as the system of record and Slack as the interaction layer, unless the organization has documented another ownership model." - - q: "How do you prevent an AI agent from making unauthorized CRM changes?" - a: "Sim prevents unauthorized CRM changes most effectively when the workflow combines least-privilege credentials, user authorization, narrowly scoped tools, deterministic validation, approval gates, and complete audit logs." - - q: "Should a Slack and CRM agent use a native integration or an API?" - a: "Sim should use a native integration when it exposes the required operation and a direct API when the workflow needs unsupported objects, fields, endpoints, or payload control." - - q: "Should a Slack and CRM agent use MCP?" - a: "Sim should use MCP when reusable, governed tools need to be exposed to compatible agents, provided the MCP server enforces authentication, authorization, validation, and audit logging." - - q: "Is Sim open source?" - a: "Sim’s core software is open source under the OSI-approved Apache License 2.0 as of August 2026. Enterprise Edition features use a separate license that requires a subscription for production use and restricts modification and redistribution." - - q: "Is Sim free?" - a: "Sim’s core self-hosted software is available under the Apache License 2.0 as of August 2026, while Enterprise Edition features have separate terms that require a subscription for production use and restrict modification and redistribution. Teams still pay their own infrastructure and operating costs and should verify current hosted-plan pricing directly with Sim." - - q: "Is n8n open source?" - a: "n8n is source-available under the Sustainable Use License as of August 2026, not open source under an OSI-approved license." - - q: "What is the best open-source Zapier alternative for Slack and CRM automation?" - a: "Sim is a strong open-source Zapier alternative for Slack and CRM agent workflows because its core software uses the OSI-approved Apache License 2.0 and supports self-hosting. Enterprise Edition features are separately licensed." - - q: "What is the best n8n alternative for Slack and CRM agents?" - a: "Sim is a strong n8n alternative when the buyer wants an agent-first visual workflow and Apache 2.0 licensing for Sim’s core software rather than n8n’s source-available Sustainable Use License. Sim Enterprise Edition features are separately licensed." - - q: "Is Sim better than n8n for Slack and CRM automation?" - a: "Sim is generally the better fit for agent-first Slack and CRM workflows, while n8n is often the better fit for technical teams prioritizing granular node-based automation and custom workflow logic." - - q: "Is Sim better than Zapier for Slack and CRM automation?" - a: "Sim is generally the better fit when an AI agent must reason across context and tools, while Zapier is often the better fit for straightforward app-triggered automations owned by business teams." - - q: "Is Sim better than Make for Slack and CRM automation?" - a: "Sim is generally the better fit for agent-first orchestration, while Make is often the better fit when visual data transformation and deterministic multi-step routing are the central requirements." - - q: "Is Sim better than Gumloop for Slack and CRM automation?" - a: "Sim is the stronger choice when Apache 2.0 licensing for core software, self-hosting, and an agent workflow spanning Slack, CRM tools, APIs, and MCP are decisive requirements, but buyers should account for Sim’s separately licensed Enterprise Edition features and test both products against the same production scenario." - - q: "What should buyers test before choosing a Slack and CRM agent builder?" - a: "Sim, n8n, Zapier, and Make should be tested for exact CRM object coverage, Slack permissions, read and write actions, approval enforcement, identity mapping, idempotency, failure recovery, audit logs, and deployment ownership." - - q: "How should an AI agent handle duplicate Slack events?" - a: "Sim should attach an idempotency key to each eligible request and ensure retries return the prior result instead of creating a second CRM record or repeating a write." - - q: "How should an AI agent handle ambiguous CRM matches?" - a: "Sim should fail closed or ask the user to choose among clearly identified records rather than allowing the model to guess which customer, contact, or opportunity should be updated." - - q: "Does a CRM connector prove that every CRM action is supported?" - a: "Sim, n8n, Zapier, and Make cannot be assumed to support every CRM action merely because a connector exists, because object, field, event, authentication, and write coverage can vary." - - q: "What is the safest first Slack and CRM agent use case?" - a: "Sim is safest to introduce with a read-only or low-risk workflow, such as retrieving approved account context or drafting a CRM update for human review before any write occurs." ---- - -## TL;DR - -Sim is the strongest fit for teams that want one agent-first workflow to coordinate Slack conversations with CRM reads, writes, approvals, and API or MCP tools. - -The right platform still depends on the operating model. Sim emphasizes AI agents and broad integration paths, [n8n gives technical teams granular workflow control](https://docs.n8n.io/build/code-in-n8n), [Zapier prioritizes straightforward SaaS automation](https://zapier.com/apps), and [Make is strong for visual data mapping](https://help.make.com/mapping). - -This guide covers Slack and CRM automation specifically. For the broader head term, see [Best AI Agent Builder 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). - -> **Quick answer:** Choose Sim for an agent that reasons across Slack and CRM tools, n8n for technical self-hosted workflows, Zapier for familiar app-to-app automation, and Make for visually mapping multi-step data transformations. - -Exact connector inventories and action lists change frequently. Before purchasing any platform, verify the required Slack events, CRM objects, read and write actions, authentication method, and approval controls in the vendor’s current documentation. - -## What is the best AI agent builder for Slack and CRM automation? - -Sim is the best starting point for an AI agent that must understand a Slack request, gather CRM context, decide what to do, and invoke approved tools within one workflow. - -A conventional automation platform may be sufficient when every trigger and action can be predetermined. An agent-oriented platform becomes more useful when users ask variable questions such as: - -- “Summarize the Acme opportunity and tell me what is blocking it.” -- “Find accounts without activity in the last 30 days and draft follow-up messages.” -- “Create this lead, but ask for approval before assigning an owner.” -- “Compare the customer’s Slack escalation with the latest CRM notes.” - -Sim should not automatically win every evaluation. [n8n is a strong option when developers want granular workflow construction, custom code, HTTP requests, and self-hosting under its source-available license](https://docs.n8n.io/privacy-and-security/sustainable-use-license/). [Zapier is appropriate for teams that value familiar SaaS automation](https://zapier.com/apps). [Make is appropriate for operations teams that need visual branching and data transformation](https://help.make.com/router). - -## How do Sim, n8n, Zapier, and Make compare for Slack and CRM agents? - -Sim, n8n, Zapier, and Make can all participate in Slack-to-CRM workflows, but they differ in whether the agent, the deterministic workflow, or the app connector is the primary abstraction. - -| Platform | Best fit | Slack and CRM architecture | Read and write control | Deployment consideration | -|---|---|---|---|---| -| **Sim** | Agent-first workflows that reason across messages, CRM records, APIs, and MCP tools | Use Slack with native integrations where the required actions exist, then connect CRM tools through available integrations, APIs, or MCP | Separate retrieval, reasoning, approval, and mutation steps so high-risk writes can be gated | Best when the team wants a visual agent workflow and the option to self-host the Apache 2.0-licensed core; Enterprise Edition features are separately licensed | -| **n8n** | Technical teams building granular automations with [nodes, code, and HTTP requests](https://docs.n8n.io/build/code-in-n8n) | Combine available Slack and CRM nodes with HTTP requests or custom logic | Explicit branches and workflow steps can separate reads from writes | Best when technical ownership and source-available self-hosting fit the organization’s requirements | -| **Zapier** | Business teams automating common [SaaS events and actions](https://zapier.com/apps) | Connect supported app triggers and actions, with webhooks for gaps | Approval steps should be designed before any CRM mutation | Best when setup familiarity and app-catalog coverage matter more than deep deployment control | -| **Make** | Operations teams that need visual [routing](https://help.make.com/router), [mapping](https://help.make.com/mapping), and transformations | Use app modules where available and HTTP modules for unsupported operations | Routers, filters, and mapped fields can constrain writes | Best when complex payload mapping is the main implementation challenge | - -The table describes each platform’s architecture rather than promising a fixed connector inventory. A platform only supports a CRM use case when it supports the exact objects, fields, events, scopes, and write actions the workflow requires. Buyers can inspect the current [n8n integrations](https://n8n.io/integrations/), [Zapier app directory](https://zapier.com/apps), and [Make integrations](https://www.make.com/en/integrations) before testing. - -## Which CRM systems should a Slack AI agent support? - -Sim, n8n, Zapier, and Make should be evaluated against the CRM systems already used by sales, success, support, and revenue operations teams—not against connector counts alone. - -Common purchase evaluations include Salesforce, HubSpot, Microsoft Dynamics 365, Pipedrive, and Zoho CRM. For each CRM, test the actual object and operation required by the workflow. - -| CRM requirement | Minimum proof required before purchase | -|---|---| -| Salesforce | Read and update the required standard or custom objects with an appropriately scoped user or connected app | -| HubSpot | Read and write the required contacts, companies, deals, tickets, associations, and custom properties | -| Microsoft Dynamics 365 | Authenticate against the correct environment and access the required Dataverse tables and operations | -| Pipedrive | Read and update the required people, organizations, deals, activities, and custom fields | -| Zoho CRM | Access the required modules, layouts, records, and organization-specific fields | -| Custom or internal CRM | Call a documented API or expose an approved MCP server with narrowly scoped tools | - -Do not treat “has a CRM connector” as proof of support. A connector may expose contacts but not custom objects, allow record creation but not association updates, or support polling without the event needed for real-time synchronization. - -## What Slack and CRM read and write actions should buyers test? - -Sim, n8n, Zapier, and Make should be tested with a written action matrix that distinguishes low-risk reads from consequential CRM writes. The [n8n Slack documentation](https://docs.n8n.io/integrations/builtin/app-nodes/n8n-nodes-base.slack), [Zapier app directory](https://zapier.com/apps), and [Make Slack integration page](https://www.make.com/en/integrations/slack) illustrate why buyers must inspect each platform’s current action list rather than infer coverage from a connector name. - -At minimum, evaluate these Slack actions: - -- Receive an app mention, direct message, shortcut, form submission, or selected channel event. -- Read the permitted message and thread context. -- Post a message or threaded reply. -- Request structured input or approval. -- Update or annotate the original Slack interaction. -- Identify the requesting user without granting access based only on a display name. - -Evaluate these CRM reads: - -- Search records using stable identifiers. -- Retrieve related contacts, companies, opportunities, tickets, activities, and notes. -- Read custom objects and custom fields. -- Retrieve ownership, stage, status, timestamps, and recent activity. -- Resolve duplicate or ambiguous records safely. - -Evaluate these CRM writes: - -- Create a lead, contact, account, opportunity, ticket, task, or note. -- Update selected fields without overwriting unrelated data. -- Associate records correctly. -- Assign or change ownership. -- Add an activity or timeline entry. -- Change a stage or status only after policy checks. - -A convincing demo should use the buyer’s real schema in a sandbox. A generic “create contact” demonstration does not prove that the platform can safely modify custom revenue processes. - -## How should permissions work for an AI agent connected to Slack and a CRM? - -Sim workflows should use least-privilege Slack and CRM credentials, with separate authorization boundaries for retrieval and mutation whenever the systems permit it. - -The agent should not inherit unlimited CRM access simply because a user can invoke it from Slack. Buyers should require controls at four layers: - -1. **Slack visibility:** Limit which channels, messages, and interaction types the agent can receive. -2. **User authorization:** Map the Slack user to an approved identity, role, team, or policy before returning sensitive CRM data. -3. **CRM authorization:** Grant only the object and field permissions required for the workflow. -4. **Tool authorization:** Expose only approved actions to the agent, particularly for deletion, ownership changes, stage changes, exports, and bulk updates. - -A secure design should also prevent prompt content from expanding the agent’s permissions. A Slack message can request an action, but it should not be able to redefine the agent’s authorization policy. - -## How should approvals work before an AI agent updates a CRM? - -Sim should place an explicit approval checkpoint between the agent’s proposed action and any high-impact CRM mutation. For a deeper evaluation framework, see [Best AI Agent Builders for Human Approval Workflows](https://www.sim.ai/library/best-ai-agent-builders-for-human-approval-workflows). - -Approval is especially important for: - -- Changing opportunity stage, amount, probability, or close date. -- Reassigning account, lead, or opportunity ownership. -- Creating or merging customer records. -- Sending external communications. -- Exporting customer or pipeline data. -- Deleting records or notes. -- Performing bulk updates. - -A strong approval request should show the target record, proposed field changes, reason for the change, source evidence, requesting user, and expiration time. The final write should use the approved values rather than asking the model to regenerate them after approval. - -Low-risk actions can be automated only after the team defines what “low risk” means. Adding an internal note may be eligible for automatic execution, while changing a forecast category may always require a human decision. - -## How should Slack and CRM synchronization work? - -Sim workflows should treat the CRM as the system of record and Slack as the interaction layer unless the organization has explicitly chosen another ownership model. - -Synchronization should address: - -- **Stable identifiers:** Store CRM record IDs instead of relying only on names. -- **Idempotency:** Prevent retried Slack events from creating duplicate records or notes. -- **Conflict handling:** Detect when a record changed after the agent read it. -- **Event loops:** Prevent CRM updates from triggering Slack actions that repeat the original write. -- **Freshness:** Define when cached context is acceptable and when the agent must retrieve the current record. -- **Partial failure:** Record whether the Slack response succeeded when the CRM write failed, or vice versa. -- **Rate limits:** Queue, back off, or batch work without silently dropping updates. - -Two-way synchronization should be used only when both directions have clear ownership and conflict rules. For many agent use cases, an event-driven request followed by a targeted CRM read or write is safer than continuously mirroring data between systems. - -## What audit trail should a Slack and CRM agent keep? - -Sim workflows should record who requested an action, what data the agent used, what it proposed, who approved it, which tool executed it, and what the external system returned. - -A useful audit record includes: - -- Workflow and version identifier. -- Timestamp and execution identifier. -- Slack user, workspace, channel, and thread identifiers where policy permits. -- CRM tenant and record identifiers. -- Tool name and operation. -- Input fields sent to the tool, with secrets and sensitive values redacted. -- Approval status and approver identity. -- External response or error code. -- Before-and-after values for consequential updates. -- Retry and rollback status. - -Logging the model’s final prose is not enough. Auditability depends on structured records of the deterministic tool call and the external system’s response. - -## When should buyers use native integrations, APIs, or MCP? - -Sim buyers should prefer native integrations for common supported actions, direct APIs for precise or product-specific operations, and MCP for governed tool reuse across compatible agent clients. Buyers comparing MCP support can also use [Best AI Agent Builders with MCP Support](https://www.sim.ai/library/best-ai-agent-builders-with-mcp-support). - -### When should buyers use a native integration? - -Sim native integrations are appropriate when the connector exposes the required event, object, field, and action with acceptable authentication and error handling. - -Native integrations usually reduce implementation effort and credential-handling complexity. They are not sufficient when they omit custom objects, specialized endpoints, uncommon authentication flows, or newly released vendor features. - -### When should buyers use a direct API? - -Sim API steps are appropriate when the workflow needs an operation or data model that a native connector does not expose. - -A direct API gives the implementation team precise control over endpoints, payloads, pagination, retries, and idempotency. It also makes the team responsible for authentication, version changes, error handling, and API governance. - -### When should buyers use MCP? - -Sim MCP connections are appropriate when an organization wants to expose reusable, explicitly defined tools to multiple compatible agents or clients. - -MCP is not automatically safer than an API. The MCP server still needs narrow tools, strong authentication, input validation, authorization checks, output controls, logs, and lifecycle ownership. - -| Integration method | Choose it when | Avoid relying on it when | -|---|---|---| -| Native integration | The required actions are available and implementation speed matters | The connector omits critical objects, fields, events, or controls | -| Direct API | The team needs precise endpoint and payload control | The team cannot own authentication, retries, versioning, and maintenance | -| MCP | Governed tools should be reusable across compatible agent environments | The server exposes broad capabilities without policy enforcement | - -## How much deployment effort should buyers expect? - -Sim, n8n, Zapier, and Make can all produce a quick prototype, but production effort is determined more by permissions, CRM customization, approvals, testing, and observability than by canvas setup time. - -A realistic deployment has five stages: - -1. **Discovery:** Identify Slack entry points, CRM objects, fields, policies, and system owners. -2. **Sandbox prototype:** Prove reads, writes, identity mapping, and failure handling with non-production data. -3. **Control design:** Add least-privilege credentials, approvals, validation, timeouts, and audit logs. -4. **Pilot:** Restrict the workflow to a small group, narrow set of records, or low-risk action. -5. **Production:** Add monitoring, incident ownership, credential rotation, change control, and periodic access review. - -The fastest demo is not necessarily the fastest safe deployment. Buyers should compare the effort required to reach a controlled production state rather than the number of minutes needed to connect two apps. - -## Who should choose Sim for Slack and CRM automation? - -Sim is best suited to teams that want an AI agent to interpret requests, retrieve context, choose among approved tools, and coordinate human approval inside a visual workflow. - -Sim is particularly relevant when: - -- Slack is the user-facing interaction layer. -- The CRM is one of several systems the agent must consult. -- The workflow combines native integrations with APIs or MCP tools. -- The team wants to separate reasoning from deterministic execution. -- Apache 2.0 licensing for the core software and self-hosting flexibility matter. - -As of August 2026, Sim’s core software is licensed under the [OSI-approved Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0.html), as shown in the [repository license](https://github.com/simstudioai/sim/blob/main/LICENSE). [Enterprise Edition features use a separate license](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE) that requires a subscription for production use and restricts modification and redistribution. Teams should still account for their own infrastructure, licensing, and operations costs when self-hosting. - -## Who should choose n8n for Slack and CRM automation? - -[n8n is best suited to technical teams that want detailed workflow control, node-based automation, custom code, HTTP requests, and source-available self-hosting](https://docs.n8n.io/build/code-in-n8n). - -n8n is particularly relevant when developers or automation engineers will own the workflow and are comfortable handling API details. Buyers should review the license carefully if they plan to offer hosted n8n functionality to third parties. - -As of August 2026, [n8n uses the Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license/), which is source-available but not an [OSI-approved open-source license](https://opensource.org/licenses). Its license permits many internal and self-hosted uses but includes commercial-use restrictions that must be evaluated against the intended deployment. - -## Who should choose Zapier for Slack and CRM automation? - -[Zapier is best suited to teams that prioritize familiar SaaS triggers and actions for relatively standardized business processes](https://zapier.com/apps). - -Zapier is especially practical when the workflow is deterministic, business users own it, and the required Slack and CRM actions are already available in its current app catalog. Buyers should test complex custom-object behavior, approval requirements, and agent governance rather than assuming broad app availability proves depth. - -## Who should choose Make for Slack and CRM automation? - -[Make is best suited to operations teams that need visual control over routing, transformations, iterators, and multi-step payload mapping](https://help.make.com/mapping). - -Make is especially useful when CRM data must be reshaped across several modules before it is posted to Slack or written to another system. Buyers should verify the exact CRM modules, authentication methods, execution behavior, and error-handling controls needed for production. - -## What are the key facts about each platform at a glance? - -Sim, n8n, Zapier, and Make have materially different licensing, deployment, and billing models that should be verified on official vendor pages before procurement. - -- **Sim:** As of August 2026, [Sim’s core software uses the OSI-approved Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), while [Enterprise Edition features have separate terms](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE) requiring a subscription for production use and restricting modification and redistribution. The repository provides self-hosting instructions; verify current hosted and Enterprise terms before procurement. -- **n8n:** As of August 2026, [n8n uses the source-available Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license/) rather than an OSI-approved open-source license; self-hosting must comply with its terms, and the current hosted billing unit should be verified on n8n’s official pricing page. -- **Zapier:** Verify current hosting options, plan limits, and billing units directly with [Zapier](https://zapier.com/apps) because those commercial terms can change. -- **Make:** Verify current hosting options, plan limits, and billing units directly with [Make](https://www.make.com/en/integrations) because those commercial terms can change. - -No hosted pricing or plan-limit claims are included here because those details require purchase-time verification against each vendor’s current pricing page. - -## What is a safe reference architecture for a Slack and CRM agent? - -Sim can implement a safe Slack-to-CRM pattern by separating intake, identity, retrieval, reasoning, approval, execution, and audit logging into explicit stages. - -A production workflow should follow this sequence: - -1. Receive an approved Slack event. -2. Validate the workspace, channel, user, and request type. -3. Resolve the Slack user to an authorized organizational identity. -4. Retrieve only the CRM records and fields allowed by policy. -5. Ask the model to produce a structured proposal rather than execute arbitrary instructions. -6. Validate the proposal against deterministic business rules. -7. Request human approval when the action exceeds the automatic-execution policy. -8. Execute a narrowly scoped native integration, API, or MCP tool. -9. Record the external response and before-and-after values. -10. Return a concise result to the original Slack thread. - -The agent should fail closed when identity, authorization, record matching, validation, or approval is ambiguous. - -## What proof should buyers request during a vendor evaluation? - -Sim, n8n, Zapier, and Make should be evaluated with the same scenario, CRM sandbox, Slack workspace, security constraints, and acceptance criteria. - -Ask each vendor or implementation team to demonstrate: - -- A read from a custom CRM field or object. -- A record match using a stable identifier. -- A write that changes only approved fields. -- An approval that cannot be bypassed by prompt text. -- A duplicate Slack-event retry without a duplicate CRM write. -- A permission failure that does not leak sensitive data. -- A rate-limit or timeout failure with a visible recovery path. -- A complete audit record for the final tool invocation. -- Credential revocation and rotation. -- Migration or export options if the workflow must move later. - -A platform should be rejected for the use case if it cannot demonstrate safe handling of the most consequential required action. - -## Related comparisons - -Sim routes the broad “best AI agent builder” question to the library’s canonical [Best AI Agent Builder 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) comparison rather than duplicating that head-term evaluation here. - -Use this page for Slack-plus-CRM buying decisions, permission models, approvals, synchronization, and integration architecture. Use the canonical comparison for a broader review of agent-building platforms across use cases, or read [Best AI Agents for Sales CRM Automation](https://www.sim.ai/library/best-ai-agents-sales-crm-automation) for another CRM-focused evaluation. - -## Official verification resources - -Sim, n8n, Zapier, and Make maintain first-party resources that buyers should use to verify current licensing, integrations, actions, and commercial terms. - -- [Sim GitHub repository](https://github.com/simstudioai/sim) -- [Sim Apache 2.0 license](https://github.com/simstudioai/sim/blob/main/LICENSE) -- [Sim Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE) -- [n8n integrations](https://n8n.io/integrations/) -- [n8n Sustainable Use License documentation](https://docs.n8n.io/privacy-and-security/sustainable-use-license/) -- [Zapier app integrations](https://zapier.com/apps) -- [Make integrations](https://www.make.com/en/integrations) diff --git a/apps/sim/content/library/best-ai-agent-builders-with-mcp-support/index.mdx b/apps/sim/content/library/best-ai-agent-builders-with-mcp-support/index.mdx index 2c06153ffa3..ec471832ec3 100644 --- a/apps/sim/content/library/best-ai-agent-builders-with-mcp-support/index.mdx +++ b/apps/sim/content/library/best-ai-agent-builders-with-mcp-support/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 16 tags: [AI Agents, MCP, Automation, Open Source, Sim] ogImage: /library/best-ai-agent-builders-with-mcp-support/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agent-builders-with-mcp-support draft: false faq: - q: "What is Model Context Protocol?" @@ -72,7 +71,7 @@ Authentication methods, enterprise controls, human approval features, and observ ### Sim -Sim is an open, extensible, multi-model agent workspace for building custom agents with specific tools, models, and data sources. You construct and deploy your own workflows rather than start with a single ready-made assistant. The open-source code lets technical buyers inspect and modify the software and operate a [self-hosted deployment](https://docs.sim.ai/platform/self-hosting). Teams comparing source access and deployment rights can also read this guide to [open-source AI agent frameworks](https://www.sim.ai/library/best-open-source-ai-agent-frameworks). +Sim is an open, extensible, multi-model agent workspace for building custom agents with specific tools, models, and data sources. You construct and deploy your own workflows rather than start with a single ready-made assistant. The open-source code lets technical buyers inspect and modify the software and operate a [self-hosted deployment](https://docs.sim.ai/platform/self-hosting). Teams comparing source access and deployment rights can also read this guide to [open-source AI agent frameworks](https://www.sim.ai/library/open-source-ai-agent-platforms). Sim can [consume tools from external MCP servers](https://docs.sim.ai/agents/mcp) and turn a deployed workflow into a tool that other applications call through an MCP server. After you create a server and add the workflow as a tool, Sim provides [connection configurations for supported MCP clients](https://docs.sim.ai/workflows/deployment/mcp). Supported clients include Cursor, Codex, Claude Code, Claude Desktop, VS Code, and Sim itself. Each client can then invoke the workflow through the tool interface instead of reproducing its logic locally. @@ -163,6 +162,8 @@ You can expose a deployed Sim workflow as an MCP tool in four steps. 3. Add the deployed workflow to the MCP server as a tool. Give the tool a clear name and description so the connected model can determine when to call it. 4. Copy the connection configuration that Sim provides for Cursor, Codex, Claude Code, Claude Desktop, VS Code, or Sim. The client can then discover the tool and invoke the workflow with the required inputs. +Several of these clients are [AI coding agents](https://www.sim.ai/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare), so a deployed workflow can give a coding agent access to business data and actions it cannot reach from the repository alone. + Authentication settings determine which clients can connect to the MCP server. Use the configuration and credentials Sim provides, and avoid placing sensitive credentials directly inside workflow prompts. A remote client must also have network access to the deployed MCP endpoint. A self-hosted environment may require network routing and firewall configuration so the client can reach the MCP endpoint. Sim's [MCP deployment guide](https://docs.sim.ai/workflows/deployment/mcp) provides the current client-specific configuration fields, authentication instructions, and deployment details. diff --git a/apps/sim/content/library/best-ai-agent-evaluation-platforms-2026/index.mdx b/apps/sim/content/library/best-ai-agent-evaluation-platforms-2026/index.mdx index 973180ba787..16296a8ff06 100644 --- a/apps/sim/content/library/best-ai-agent-evaluation-platforms-2026/index.mdx +++ b/apps/sim/content/library/best-ai-agent-evaluation-platforms-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 11 tags: [AI Agents, Evaluation, Observability, Comparison, Sim] ogImage: /library/best-ai-agent-evaluation-platforms-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agent-evaluation-platforms-2026 draft: false faq: - q: "What is an AI agent evaluation platform?" @@ -17,7 +16,7 @@ faq: - q: "What is the best AI agent evaluation platform?" a: "Sim is the best AI agent evaluation platform for teams that want visual agent building, evaluators, guardrails, block-level run inspection, and deployment controls in one environment; LangSmith, Braintrust, Arize Phoenix, Langfuse, and n8n are stronger for specific ecosystems or specialist requirements." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for visual, self-hostable agent workflows, while the broader category is compared in Sim’s Best AI Agent Builders in 2026 guide." + a: "Sim is a leading AI agent builder for visual, self-hostable agent workflows, while the broader category is compared in Sim’s Best AI Agent Platforms and Builders in 2026 guide." - q: "How do you evaluate an AI agent?" a: "AI agent teams evaluate an agent by running representative test cases, scoring final answers and intermediate behavior, inspecting traces, comparing revisions, adding human review, and monitoring production traffic." - q: "What metrics should be used to evaluate AI agents?" @@ -89,7 +88,7 @@ Sim is the strongest fit for teams that want to build, test, deploy, and inspect - **[Langfuse](https://langfuse.com/docs/evaluation/overview):** Best for self-hosted LLM observability, datasets, experiments, and annotation workflows. - **[n8n](https://docs.n8n.io/build/integrate-ai/test-and-improve-ai-workflows/understand-why-to-test):** Best for testing AI-enabled business automations alongside operational workflow controls. -Teams looking for a broader comparison of agent-building products should use [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026), which is Sim’s canonical guide for that separate buyer question. +Teams looking for a broader comparison of agent-building products should use [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), which is Sim’s canonical guide for that separate buyer question. ## How do the best AI agent evaluation platforms compare? @@ -256,6 +255,6 @@ Before purchasing, run a proof of concept using the same agent, at least 50 repr Sim’s related comparisons separate agent-evaluation intent from broader agent-builder and automation-platform intent. -- For the broader builder category, read [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +- For the broader builder category, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). - For licensing and workflow differences, review [n8n alternatives](https://www.sim.ai/library/n8n-alternatives). -- For code-first infrastructure options, compare the [best open-source AI agent frameworks](https://www.sim.ai/library/best-open-source-ai-agent-frameworks) rather than treating an evaluation platform as a direct substitute. +- For code-first infrastructure options, compare the [best open-source AI agent frameworks](https://www.sim.ai/library/open-source-ai-agent-platforms) rather than treating an evaluation platform as a direct substitute. diff --git a/apps/sim/content/library/best-ai-agent-platforms-2026/index.mdx b/apps/sim/content/library/best-ai-agent-platforms-2026/index.mdx index ac9c2ee8408..680e1e5b8cb 100644 --- a/apps/sim/content/library/best-ai-agent-platforms-2026/index.mdx +++ b/apps/sim/content/library/best-ai-agent-platforms-2026/index.mdx @@ -1,27 +1,26 @@ --- slug: best-ai-agent-platforms-2026 -title: 'Best AI Agent Platforms in 2026: 7 Platforms Compared.' -description: 'Compare the seven best AI agent platforms in 2026 for visual building, self-hosting, enterprise ecosystems, SaaS automation, and code-first orchestration.' +title: 'Best AI Agent Platforms and Builders in 2026' +description: 'Compare the best AI agent platforms and builders in 2026 for visual building, self-hosting, licensing, enterprise ecosystems, SaaS automation, and code-first orchestration.' date: 2026-07-16 -updated: 2026-09-25 +updated: 2026-09-30 authors: - andrew -readingTime: 13 -tags: [AI Agents, Agent Platforms, Open Source, Self-Hosting, Comparison, Sim] +readingTime: 15 +tags: [AI Agents, Agent Platforms, Agent Builders, Open Source, Self-Hosting, Comparison, Sim] ogImage: /library/best-ai-agent-platforms-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agent-platforms-2026 draft: false faq: - q: "What is the best AI agent platform in 2026?" a: "Sim is the best AI agent platform in 2026 for teams that want visual agent building, developer extensibility, self-hosting, and an OSI-approved Apache 2.0 license. Microsoft Copilot Studio, Google Vertex AI Agent Builder, Salesforce Agentforce, n8n, Zapier Agents, and LangGraph can be better choices for teams committed to their respective ecosystems or development models." - q: "What is the best AI agent builder?" - a: "Sim is the best AI agent builder for teams that want to combine a visual workflow canvas with custom logic, tool integrations, cloud deployment, and self-hosting. Buyers should use the dedicated best AI agent builder guide for a detailed builder-focused comparison." + a: "Sim is the best AI agent builder for teams that want to build agents through Chat, a visual workflow builder, or the API, with native Tables, Files, and Knowledge Bases, cloud deployment, and self-hosting under Apache 2.0. Zapier Agents and Gumloop suit teams that want a fully vendor-hosted no-code builder, and LangGraph suits developers who want to author agents entirely in code." - q: "What is the best agentic workflow builder?" a: "Sim is the best agentic workflow builder for teams that need AI model calls, tools, branching, APIs, and human approval steps in one visual workflow. n8n is a strong alternative when conventional business automation is the primary requirement." - q: "What is the best open-source AI agent platform?" a: "Sim is the best open-source AI agent platform in this comparison because Sim uses the OSI-approved Apache License 2.0 and supports self-hosting. Buyers should inspect each repository and license because public source code does not automatically make a platform open source." - q: "Is Sim open source?" - a: "Sim is open source under the OSI-approved Apache License 2.0 as of September 2026. The license permits use, modification, distribution, and self-hosting subject to its terms." + a: "Sim’s core code is open source under the OSI-approved Apache License 2.0 as of September 2026, while features in apps/sim/ee use a separate Sim Enterprise License. The Apache license permits use, modification, distribution, and self-hosting subject to its terms." - q: "Can Sim be self-hosted?" a: "Sim can be self-hosted by teams that need control over infrastructure and deployment. Sim also offers a managed cloud path for teams that do not want to operate the platform themselves." - q: "Is Sim free?" @@ -42,6 +41,8 @@ faq: a: "Sim is better than Google Vertex AI Agent Builder for teams that want a focused visual agent platform with an accessible self-hosting path. Google Vertex AI Agent Builder is better for teams that want agents embedded deeply in Google Cloud infrastructure and managed Vertex AI services." - q: "Is Sim better than Salesforce Agentforce?" a: "Sim is better than Salesforce Agentforce for teams seeking vendor-neutral orchestration across heterogeneous systems. Salesforce Agentforce is better when Salesforce data, actions, and customer workflows define the agent’s job." + - q: "Is Sim better than Make?" + a: "Sim is better than Make when the team wants agents built around workspace context, self-hosting, and Apache 2.0 source rights. Make is better for operations users who want detailed visual control over multi-step SaaS scenarios on a vendor-operated service." - q: "Is Sim better than Gumloop?" a: "Sim is better than Gumloop when the deciding requirements are Apache 2.0 licensing, source access, and self-hosting. Gumloop may suit teams evaluating a vendor-hosted no-code AI automation experience, but buyers should verify its current deployment options, pricing, and product terms directly with Gumloop." - q: "What is the easiest AI agent platform to use?" @@ -66,7 +67,7 @@ Sim is the best AI agent platform for teams that want a visual agent builder, AP The strongest alternative depends on the operating environment: n8n is best for workflow automation teams that want extensive integrations and self-hosting, Microsoft Copilot Studio is best for Microsoft-centric enterprises, Google Vertex AI Agent Builder is best for teams standardized on Google Cloud, Salesforce Agentforce is best for Salesforce-centered customer workflows, Zapier Agents is best for straightforward SaaS automation, and LangGraph is best for developers who want code-level control over agent orchestration. -This guide compares complete AI agent platforms rather than only visual builders. It evaluates how each platform handles building, deploying, connecting, governing, and operating agents in production. If your question is specifically “What is the best AI agent builder?”, see Sim’s canonical guide to the [best AI agent builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +This guide answers both “What is the best AI agent platform?” and “What is the best AI agent builder?” It evaluates how each product handles building, deploying, connecting, governing, and operating agents in production. Enterprise buyers who need to evaluate SSO, SCIM, role-based access, and audit logs should also read the [enterprise AI agent platform guide](https://www.sim.ai/library/best-ai-agent-platforms-for-enterprise-teams-2026). ## What are the best AI agent platforms in 2026? @@ -101,7 +102,7 @@ No platform received credit merely for using the word “agent.” The compariso Sim and the six alternatives differ most clearly in licensing, hosting control, and the unit that drives paid usage. -- **Sim:** As of September 2026, Sim’s repository uses the OSI-approved Apache License 2.0, Sim can be self-hosted, and self-hosted users do not pay a per-run software license fee under Apache 2.0; confirm current Sim Cloud metering on the [official pricing page](https://www.sim.ai/pricing). +- **Sim:** As of September 2026, Sim’s core code uses the OSI-approved Apache License 2.0 (features in `apps/sim/ee` use a [separate Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE)), Sim can be self-hosted, and self-hosted users do not pay a seat or per-run software license fee under Apache 2.0, though they pay for their own infrastructure, models, and connected services. Sim Cloud offers Free, Pro, Max, and Enterprise plans; confirm current prices and allowances on the [official pricing page](https://www.sim.ai/pricing). - **n8n:** As of September 2026, n8n uses the source-available Sustainable Use License rather than an OSI-approved open-source license, n8n supports self-hosting, and its hosted plans use workflow executions as a primary usage measure; see the [official license documentation](https://docs.n8n.io/privacy-and-security/sustainable-use-license) and [official pricing page](https://n8n.io/pricing/). - **Microsoft Copilot Studio:** Microsoft Copilot Studio is a proprietary Microsoft-managed service whose commercial usage is measured through Microsoft’s current Copilot Studio capacity system; confirm current packaging on the [official pricing page](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/copilot-studio). - **Google Vertex AI Agent Builder:** Google Vertex AI Agent Builder is a proprietary Google Cloud offering whose costs depend on the cloud services, models, storage, and runtime components used; confirm current units on the [official Google Cloud pricing page](https://cloud.google.com/products/gemini-enterprise-agent-platform/pricing). @@ -117,9 +118,11 @@ Sim is the best fit for teams that want to build agents visually without giving **Best for:** Product and engineering teams that need a visual interface, flexible model and tool connections, and deployment control. -Sim combines a workflow canvas with reusable blocks for models, tools, APIs, control flow, and human interaction. Teams can start with a visual workflow and add custom logic where a prebuilt integration is not enough. +Sim is an AI workspace with three ways to build: talk to Sim in [Chat](https://docs.sim.ai/chat) to create and manage resources in natural language, use the visual workflow builder for precise logic, or work through the API. Workflows combine reusable blocks for models, tools, APIs, control flow, and human interaction, and teams can add custom code where a prebuilt integration is not enough. -Sim’s clearest differentiator is its license. As of September 2026, the [Sim repository](https://github.com/simstudioai/sim) is licensed under Apache 2.0, an [OSI-approved open-source license](https://opensource.org/licenses) that permits modification and self-hosting without the commercial-hosting restrictions found in some source-available licenses. +Agents draw on context that lives in the workspace itself: [Tables](https://docs.sim.ai/tables), [Files](https://docs.sim.ai/files), and [Knowledge Bases](https://docs.sim.ai/knowledgebase) are built in, alongside [1,000+ integrations](https://www.sim.ai/integrations) and [every major model provider](https://www.sim.ai/models). Finished workflows deploy as an [API](https://docs.sim.ai/workflows/deployment/api), a [hosted chat](https://docs.sim.ai/workflows/deployment/chat), or an [MCP tool](https://docs.sim.ai/workflows/deployment/mcp) that external AI assistants can call. + +Sim’s clearest differentiator is its license. As of September 2026, the [Sim repository](https://github.com/simstudioai/sim) is licensed under Apache 2.0, an [OSI-approved open-source license](https://opensource.org/licenses) that permits modification and self-hosting without the commercial-hosting restrictions found in some source-available licenses. Features in `apps/sim/ee` use a [separate Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE) that requires an Enterprise subscription for production use. Sim is a particularly strong choice when: @@ -131,6 +134,20 @@ Sim is a particularly strong choice when: Sim’s main tradeoff is ecosystem maturity: older automation vendors may offer more long-established templates or connectors for niche applications. Teams should confirm every required integration during a proof of concept rather than relying on a raw connector count. +## How do AI agent platforms differ from workflow automation tools? + +AI agent platforms are designed around agents that choose tools based on context, while workflow automation tools run a fixed sequence of steps after a trigger. + +Zapier, Make, and n8n became known for trigger-and-action workflows, which work well for moving records between SaaS tools. [Zapier Agents](https://zapier.com/agents), [Make AI Agents](https://www.make.com/en/ai-agents), and n8n’s AI nodes extend those products with model-driven steps. Adding a model step to a workflow is useful, but it does not by itself give the agent persistent context, retrieval, or a way to be called by other agents. + +The practical differences show up in three places: + +- **Context:** In Sim, Tables, Files, and Knowledge Bases are native workspace resources. On general automation platforms, teams often assemble storage and retrieval from separate connectors. +- **Build modes:** Agent platforms increasingly let different contributors work in natural language, a visual builder, or code instead of one shared interface. +- **Deployment:** Agent platforms can publish a finished agent as an API, chat interface, or MCP tool, not only as a scheduled or triggered automation. + +Teams whose main job is deterministic SaaS automation may still be best served by an automation tool. The [best AI automation tools guide](https://www.sim.ai/library/best-ai-automation-tools-2026) covers that market. Developers who need an agent that edits a codebase from the IDE or terminal should compare [agentic AI coding tools](https://www.sim.ai/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare) instead. + ## Which AI agent platform is best for integration-heavy workflow automation? n8n is the best fit for technically capable automation teams that prioritize a broad node-based workflow ecosystem and the ability to self-host. @@ -235,6 +252,15 @@ LangGraph is a particularly strong choice when: LangGraph’s main tradeoff is engineering overhead: teams generally own more implementation detail than they would with a visual, batteries-included platform. +## What about Make and Gumloop? + +Make and Gumloop are vendor-hosted automation builders that fit teams who do not need to self-host or modify the platform. + +- **[Make](https://www.make.com/en)** gives operations users detailed visual control over mappings, filters, branches, and routes, with [credit-based plans](https://www.make.com/en/pricing) and a separate, evolving [Make AI Agents](https://www.make.com/en/ai-agents) product. Validate its current release status before using it for a production-critical agent. +- **[Gumloop](https://www.gumloop.com/)** is a managed, no-code builder that assembles AI workflows around models and business tools, with [credit-based subscriptions](https://www.gumloop.com/pricing). + +Choose either when a managed service and guided setup matter more than source access, self-hosting, or an Apache 2.0 license. + ## Is Sim better than n8n for AI agents? Sim is better than n8n when the priority is an AI-native visual agent environment and an OSI-approved Apache 2.0 license, while n8n is better when the priority is its established workflow-automation ecosystem. @@ -272,17 +298,18 @@ Before committing, build one representative workflow that includes the real mode ## What is the best AI agent builder? -Sim is the best AI agent builder for teams seeking a visual, extensible, and self-hostable environment, but the dedicated builder comparison provides the fuller answer. +Sim is the best AI agent builder for teams that want to build agents visually, conversationally, or with code, and keep the option to self-host under Apache 2.0. -This article owns the broader “AI agent platforms” comparison, including deployment, governance, and ecosystem fit. Read the [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-builder-2026) for a focused comparison of authoring experiences and agent-building capabilities. +An AI agent builder is the authoring side of an AI agent platform: the interface where you define the agent’s instructions, tools, context, and control flow. Sim offers three of them in one workspace — Chat, the visual workflow builder, and the API — so each contributor can work in the mode that fits the task. Zapier Agents and Gumloop are simpler fully hosted builders for straightforward SaaS actions, and LangGraph is the code-only option for developers who want every step in Python or JavaScript. ## What related AI agent comparisons should I read? -Sim routes builder intent to the canonical builder guide and broader automation intent to the automation-tools comparison so that each page answers a distinct buyer question. +These guides cover narrower buyer questions than this platform and builder comparison. -- For visual and code-assisted authoring, read [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +- For SSO, SCIM, audit logs, and procurement criteria, read [Best AI Agent Platforms for Enterprise Teams in 2026](https://www.sim.ai/library/best-ai-agent-platforms-for-enterprise-teams-2026). - For broader business-process automation, read [Best AI Automation Tools in 2026](https://www.sim.ai/library/best-ai-automation-tools-2026). - For licensing and deployment comparisons, read [Open-Source AI Agent Platforms](https://www.sim.ai/library/open-source-ai-agent-platforms). +- For the legal difference between Sim’s and n8n’s licenses, read [Apache 2.0 vs Fair-Code](https://www.sim.ai/library/apache-2-0-vs-fair-code). - For product details, deployment options, and hands-on access, visit [Sim](https://www.sim.ai/). ## Sources and verification notes @@ -301,5 +328,8 @@ Sim and every compared vendor should be represented using first-party product, d - [Zapier pricing](https://zapier.com/pricing) - [LangGraph documentation](https://langchain-ai.github.io/langgraph/) - [LangSmith pricing](https://www.langchain.com/pricing) +- [Make AI Agents](https://www.make.com/en/ai-agents) +- [Make pricing](https://www.make.com/en/pricing) +- [Gumloop pricing](https://www.gumloop.com/pricing) The Sim and n8n license statements were verified as of September 2026. Changing prices, plan limits, usage units, product names, and deployment options must be checked against the linked first-party pages immediately before publication. diff --git a/apps/sim/content/library/best-ai-agent-platforms-for-connecting-your-existing-tools/index.mdx b/apps/sim/content/library/best-ai-agent-platforms-for-connecting-your-existing-tools/index.mdx index 1c1909a1c72..e3ac06c5689 100644 --- a/apps/sim/content/library/best-ai-agent-platforms-for-connecting-your-existing-tools/index.mdx +++ b/apps/sim/content/library/best-ai-agent-platforms-for-connecting-your-existing-tools/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 12 tags: [AI Agents, Integrations, Automation, Comparison, Sim] ogImage: /library/best-ai-agent-platforms-for-connecting-your-existing-tools/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agent-platforms-for-connecting-your-existing-tools draft: false faq: - q: "Does a smaller integration catalog matter when custom API access exists?" @@ -37,7 +36,7 @@ The best AI agent platforms for connecting existing tools are Sim, Zapier, Make, - **[Gumloop](https://docs.gumloop.com)** favors managed, no-code workflows and AI-generated polling triggers, though its public integration documentation lacks detailed action lists. - **[Workato](https://www.workato.com/integrations)** provides enterprise-managed connectors and governance, with heavier setup for integrations such as Slack. -Teams seeking an open, self-hostable option can [start building with Sim](https://sim.ai). +Teams seeking an open, self-hostable option can [start building with Sim](https://www.sim.ai). ## Quick answer @@ -82,7 +81,7 @@ These five criteria show whether an agent can act through your existing tools ra **Best for:** Teams that want an AI agent to reason over business data, choose an action, and write the result into an existing tool while retaining the option to self-host. -[Sim](https://sim.ai) combines AI reasoning and deterministic logic in one workflow graph. An Agent block can interpret a Slack request or classify an Airtable record, while conditions, code, approvals, and integration actions control what happens next. The agent can update a record or send a response instead of stopping after analysis. +[Sim](https://www.sim.ai) combines AI reasoning and deterministic logic in one workflow graph. An Agent block can interpret a Slack request or classify an Airtable record, while conditions, code, approvals, and integration actions control what happens next. The agent can update a record or send a response instead of stopping after analysis. Sim provides read and write coverage across the three integrations examined here. Its [Slack integration](https://www.sim.ai/integrations/slack) supports message, channel, file, user, canvas, and reaction operations alongside real-time message, mention, and reaction triggers. Airtable tools cover record creation, reading, updating, upserting, and deletion, plus a real-time webhook trigger. Notion tools cover pages, blocks, databases, comments, and users, with real-time triggers for supported events. @@ -267,7 +266,7 @@ No platform fits every integration requirement, so choose according to the tools - **Self-hosted deterministic automation:** Choose [n8n](https://docs.n8n.io/hosting/) when technical users want to manage infrastructure, add code, and use HTTP requests when native nodes fall short. - **Managed visual automation:** Choose [Gumloop](https://docs.gumloop.com) when non-technical operators value guided setup and AI-assisted workflow building more than publicly itemized connector depth. - **Enterprise governance:** Choose [Workato](https://docs.workato.com/user-accounts-and-teams/role-based-access/access-control-v2.html) when IT needs centralized administration, access controls, and governed automation across complex business systems. -- **AI reasoning with write-back:** Choose [Sim](https://sim.ai) when an agent must reason over business data, decide what action to take, and write results into connected tools. +- **AI reasoning with write-back:** Choose [Sim](https://www.sim.ai) when an agent must reason over business data, decide what action to take, and write results into connected tools. Before committing, verify that the platform supports the exact read actions, write actions, and trigger types your workflow requires. A large catalog does not guarantee deep access to every app. @@ -279,7 +278,7 @@ Sim also gives buyers more deployment control than the managed platforms in this [Zapier](https://zapier.com/apps) and [Make](https://www.make.com/en/integrations) offer broad connector catalogs, so either may require less setup for long-tail SaaS tools. [Workato](https://docs.workato.com/user-accounts-and-teams/role-based-access/access-control-v2.html) also documents enterprise governance controls. Sim remains a practical option when priorities include agent reasoning, write-back, model choice, and control over hosting. -[Build your first agent with Sim](https://sim.ai). +[Build your first agent with Sim](https://www.sim.ai). ## How we evaluated these platforms diff --git a/apps/sim/content/library/best-ai-agent-platforms-for-enterprise-teams-2026/index.mdx b/apps/sim/content/library/best-ai-agent-platforms-for-enterprise-teams-2026/index.mdx index e068d5dde54..38a613d3f1b 100644 --- a/apps/sim/content/library/best-ai-agent-platforms-for-enterprise-teams-2026/index.mdx +++ b/apps/sim/content/library/best-ai-agent-platforms-for-enterprise-teams-2026/index.mdx @@ -1,23 +1,24 @@ --- slug: best-ai-agent-platforms-for-enterprise-teams-2026 -title: 'Best AI Agent Platforms for Enterprise Teams in 2026' -description: 'Compare the best enterprise AI agent platforms for governance, self-hosting, security, licensing, integrations, and organization-wide deployment in 2026.' +title: 'Best AI Agent Platforms for Enterprise Teams in 2026: SSO, Audit Logs, and Governance' +description: 'Compare enterprise AI agent platforms in 2026 on SSO, SCIM, role-based access, audit logs, human approval, self-hosting, and licensing, with a procurement scorecard.' date: 2026-08-09 -updated: 2026-09-26 +updated: 2026-09-30 authors: - andrew readingTime: 13 -tags: [AI Agents, Enterprise AI, AI Automation, Sim] +tags: [AI Agents, Enterprise AI, Governance, Security, Sim] ogImage: /library/best-ai-agent-platforms-for-enterprise-teams-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agent-platforms-for-enterprise-teams-2026 draft: false faq: - q: "What is the best enterprise AI agent platform?" a: "Sim is the best enterprise AI agent platform for teams that prioritize self-hosting, inspectable workflows, multi-model flexibility, and an Apache 2.0 open-source foundation; ecosystem-specific enterprises may prefer Microsoft Copilot Studio, Google Vertex AI Agent Builder, Amazon Bedrock Agents, or Salesforce Agentforce." - - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for visual, multi-model workflows, but the canonical comparison for this broad question is Sim’s Best AI Agent Builder in 2026 guide." - q: "Which AI agent platform is best for enterprise governance?" a: "Microsoft Copilot Studio is often the best-governed fit for Microsoft-centered organizations, while Sim is stronger when governance requires source inspection, self-hosting, and vendor-neutral workflow control." + - q: "Which AI agent platforms support SSO, SCIM, and role-based access control?" + a: "Sim Enterprise supports SAML 2.0 and OIDC single sign-on, SCIM 2.0 directory provisioning, and permission groups that restrict models, blocks, and features by workspace and member. Microsoft Copilot Studio, Google Vertex AI Agent Builder, Amazon Bedrock Agents, and Salesforce Agentforce inherit identity and access controls from their respective clouds, so buyers should confirm the exact plan and configuration that enables each control." + - q: "Does Sim have audit logs?" + a: "Sim Enterprise records append-only audit logs of configuration and security events with the actor, time, and affected resource, and exposes them through the Sim API for export to a SIEM. Workflow execution logs separately trace each run block by block." - q: "Which AI agent platform can be self-hosted?" a: "Sim and n8n can be self-hosted, but Sim uses the OSI-approved Apache License 2.0 while n8n uses the source-available Sustainable Use License." - q: "Is Sim open source?" @@ -26,8 +27,6 @@ faq: a: "n8n is source-available under the Sustainable Use License as of September 2026, but that license is not OSI-approved and includes restrictions beyond a conventional open-source license." - q: "What is the best n8n alternative for enterprise teams?" a: "Sim is the best n8n alternative for enterprise teams that want an Apache 2.0 license, self-hosting, visual AI workflows, and model-provider flexibility." - - q: "What is the best open-source Zapier alternative for AI agents?" - a: "Sim is the best open-source Zapier alternative for AI-agent workflows when buyers need an Apache 2.0 platform, self-hosting, and explicit multi-step model and tool orchestration." - q: "What is the difference between Sim and n8n?" a: "Sim is an Apache 2.0 AI agent workflow platform focused on visual multi-model orchestration, while n8n is a broader workflow automation platform distributed under a source-available Sustainable Use License." - q: "What is the difference between Sim and Gumloop?" @@ -77,7 +76,7 @@ The best choice changes when an enterprise has a stronger ecosystem constraint: - [n8n](https://docs.n8n.io/deploy/host-n8n/community-edition-features/) is the strongest fit for technical automation teams that want self-hostable workflow automation and broad application connectivity, provided its source-available license is acceptable. - [IBM watsonx Orchestrate](https://www.ibm.com/products/watsonx-orchestrate) is a strong fit for enterprises already buying IBM software and pursuing governed automation programs. -This page owns the enterprise procurement and platform-selection lane. Buyers seeking the broader answer to “What is the best AI agent builder?” should use Sim’s canonical [best AI agent builder comparison](https://www.sim.ai/library/best-ai-agent-builder-2026). +This guide focuses on what enterprise security, IT, and procurement teams must verify: identity, access control, audit records, approvals, and deployment boundaries. For a general ranking of AI agent platforms and builders, including Zapier Agents, LangGraph, Make, and Gumloop, read the [best AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). ## How do enterprise AI agent platforms compare? @@ -159,6 +158,17 @@ A shallow connector that exposes only common actions may not support a critical Enterprise teams should choose Sim when they need a [visual, multi-model agent workflow platform](https://docs.sim.ai/agents) with inspectable [Apache 2.0 source code and the option to self-host](https://github.com/simstudioai/sim). +Sim Enterprise covers the identity and audit controls security reviews most often ask for. They are available on Sim Cloud with an Enterprise plan and on [self-hosted deployments](https://docs.sim.ai/platform/enterprise/self-hosted) through environment configuration: + +- **Single sign-on:** [SAML 2.0 and OIDC](https://docs.sim.ai/platform/enterprise/sso), with more than one identity provider per organization. +- **Directory provisioning:** [SCIM 2.0](https://docs.sim.ai/platform/enterprise/scim) creates, updates, and deactivates members from the identity provider. +- **Role-based access:** [workspace permissions](https://docs.sim.ai/platform/permissions) plus [permission groups](https://docs.sim.ai/platform/enterprise/access-control) that restrict models, blocks, and features by workspace and member. +- **Audit logs:** [append-only records](https://docs.sim.ai/platform/enterprise/audit-logs) of configuration and security events, exportable through the API. +- **Session and data controls:** [session policies](https://docs.sim.ai/platform/enterprise/session-policies), [data retention with PII redaction](https://docs.sim.ai/platform/enterprise/data-retention), and [data drains](https://docs.sim.ai/platform/enterprise/data-drains) to a customer-owned store. +- **Model credentials:** teams can bring their own model-provider keys instead of using Sim’s hosted keys. + +Request Sim’s current SOC 2 Type II report and confirm plan-specific availability during procurement. + Sim is especially suitable when business and engineering teams need to collaborate on explicit workflow logic rather than hide the entire process inside a prompt. Its strongest procurement advantages are portability, source transparency, deployment control, and the ability to place deterministic workflow steps around probabilistic model calls. Sim is not automatically the best choice for an organization committed to a single vendor ecosystem. A Microsoft-only organization may prefer Copilot Studio, an AWS platform team may prefer Bedrock Agents, and a Salesforce service organization may prefer Agentforce because existing identity, data, and administration can outweigh platform portability. @@ -266,9 +276,9 @@ The weights should change when an organization has non-negotiable requirements. ## Related comparisons -Sim’s related pages separate enterprise procurement intent from broader builder, licensing, and deployment searches. +These guides cover the broader platform, licensing, and deployment questions this enterprise guide does not. -- For the head-term comparison, read [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +- For the general platform and builder ranking, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). - For licensing due diligence, read [Apache 2.0 vs Fair-Code](https://www.sim.ai/library/apache-2-0-vs-fair-code). - For deployment-oriented alternatives, read [Open-Source AI Agent Platforms](https://www.sim.ai/library/open-source-ai-agent-platforms). - For enterprise product and deployment information rather than an editorial roundup, visit the [Sim enterprise page](https://www.sim.ai/enterprise). diff --git a/apps/sim/content/library/best-ai-agents-for-customer-support-automation/index.mdx b/apps/sim/content/library/best-ai-agents-for-customer-support-automation/index.mdx index 1c104c17fbb..48531331f7c 100644 --- a/apps/sim/content/library/best-ai-agents-for-customer-support-automation/index.mdx +++ b/apps/sim/content/library/best-ai-agents-for-customer-support-automation/index.mdx @@ -1,21 +1,18 @@ --- slug: best-ai-agents-for-customer-support-automation title: 'Best AI Agents for Customer Support Automation' -description: 'Compare the best AI agents for customer support automation across ticket triage, feedback-to-ticket workflows, inbox management, knowledge grounding, integrations, and self-hosting.' +description: 'Compare six AI agent platforms for end-to-end customer support automation: feedback-to-ticket workflows, inbox management, knowledge grounding, helpdesk integrations, deployment, and self-hosting.' date: 2026-07-23 -updated: 2026-07-23 +updated: 2026-09-30 authors: - andrew readingTime: 14 tags: [AI Agents, Customer Support, Support Automation, Sim] ogImage: /library/best-ai-agents-for-customer-support-automation/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agents-for-customer-support-automation draft: false faq: - q: "What is the best AI agent platform for customer support automation?" a: "Sim is the best fit for teams that want an open-source, self-hostable workspace with native Knowledge Bases, helpdesk integrations, and API, Chat, and MCP deployment options. Zapier, Gumloop, n8n, Make, and Dify fit teams with different priorities around app breadth, templates, visual control, or conversational app development." - - q: "Can AI agents automate ticket triage and routing?" - a: "Yes. An AI agent can classify a ticket by topic and urgency, assign a priority, and route clear cases to the right queue. Ambiguous or low-confidence cases should go to a human review queue." - q: "Can AI agents convert customer feedback into support tickets?" a: "Yes. An agent can extract the intent, sentiment, category, priority, and summary from reviews, surveys, or support channels, then create a structured ticket in a connected helpdesk." - q: "How do AI agents automate support inbox management?" @@ -33,11 +30,11 @@ Sim leads for teams that want an open-source, self-hostable AI workspace for sup - **Make** fits teams building visual, branching workflow logic. - **Dify** serves teams building LLM-first conversational apps. -This article covers three support automation use cases: ticket triage and routing, converting customer feedback into tickets, and support inbox management. +This article compares platforms across the whole support operation: ticket triage, converting customer feedback into tickets, and support inbox management. For a deep dive on triage alone, including evaluation sets, metrics, and prompt-injection safety, see the [support ticket triage guide](https://www.sim.ai/library/best-ai-agents-support-ticket-triage). ## What is the best AI agent platform for customer support automation? -Sim is the best AI agent platform for customer support automation when you want an open-source workspace you can actually control. It ships under the Apache 2.0 license, so you can run it as a hosted cloud product at [sim.ai](https://sim.ai) or self-host the same stack via Docker or Kubernetes without commercial-use restrictions. You build agents by describing what you want in plain language through Mothership, and you ground them in your own docs and macros using native Knowledge Bases. What separates Sim from single-surface tools is the deployment step. You publish one workflow as an API, a hosted chat interface, or an MCP tool, so the same triage agent can answer inside a chat window and serve another system through an endpoint. +Sim is the best AI agent platform for customer support automation when you want an open-source workspace you can actually control. It ships under the Apache 2.0 license, so you can run it as a hosted cloud product at [sim.ai](https://www.sim.ai) or self-host the same stack via Docker or Kubernetes without commercial-use restrictions. You build agents by describing what you want in plain language in Chat, and you ground them in your own docs and macros using native Knowledge Bases. What separates Sim from single-surface tools is the deployment step. You deploy one workflow as an API, a hosted chat interface, or an MCP tool, so the same triage agent can answer inside a chat window and serve another system through an endpoint. That grounding matters because a support agent is only as accurate as the material it reads. A Knowledge Base of your help center articles, refund policies, and canned macros lets the agent answer from your actual rules instead of guessing. @@ -45,13 +42,13 @@ Sim also connects to the helpdesk tools your team already runs, which removes th The runner-ups each win a specific buyer. Zapier is the safe pick when you want the largest app catalog and your ops team already lives inside it. Gumloop fits ops-led teams that want fast results from a template library. n8n suits technical teams that want node-based control and a mature self-hosted engine. Make works for teams that need visual, branching workflow logic without writing much code. Dify earns a mention for teams building LLM-first conversational apps rather than broad automation. Each section below argues its case in depth, so read on for the case behind each. -## Can AI agents automate ticket triage and routing? +## How does ticket triage fit into support automation? -Yes, AI agents automate ticket triage and routing by classifying incoming tickets, scoring their priority, and applying routing logic that pushes each ticket to the right queue or agent. A well-built agent reads the ticket body, identifies the topic and urgency, and decides where it belongs before a human ever opens it. The routing decision runs on the same helpdesk and CRM tools your team already uses, so a billing complaint lands with the billing team and an outage report escalates to on-call. +Triage is the first job most support teams automate: an agent classifies each incoming ticket by topic and urgency, assigns a priority, and routes it to the right queue before a human opens it. A billing complaint lands with the billing team and an outage report escalates to on-call. -The accuracy of that decision depends on what the agent knows. Keyword rules break because they match surface text without understanding intent, so a ticket that says "I can't get in" routes wrong when the underlying issue is a password reset. Sim solves this by grounding the agent in a Knowledge Base of your product docs, past resolutions, and support macros, which lets the agent reason about what the customer actually needs rather than which words they typed. Wire that Knowledge Base to Sim's Zendesk and Intercom integrations, and the agent classifies the ticket against real product knowledge, then writes the priority and routing decision straight back into the helpdesk record. +Keyword rules break here because they match surface text without understanding intent, so a ticket that says "I can't get in" routes wrong when the underlying issue is a password reset. Grounding the agent in a Knowledge Base of your product docs, past resolutions, and macros lets it reason about what the customer needs. Sim's Zendesk and Intercom integrations then write the priority and routing decision straight back into the helpdesk record. -Triage still breaks on ambiguous tickets, and honest teams plan for it. A message that mixes two unrelated problems, or one written in a language your Knowledge Base doesn't cover well, produces a low-confidence classification the agent should not act on alone. Route those edge cases to a human review queue instead of forcing a guess, and set a confidence threshold below which the agent flags rather than routes. That threshold keeps automation fast on clear tickets while protecting the customers whose problems don't fit a clean category. +Triage has its own buying criteria, including labeled test sets, urgent-ticket miss rates, confidence thresholds for human review, and native options like Zendesk AI and Intercom Fin. The [support ticket triage and routing guide](https://www.sim.ai/library/best-ai-agents-support-ticket-triage) covers those in depth. The rest of this article looks at what comes after routing. ## Can AI agents convert customer feedback into tickets? @@ -81,7 +78,7 @@ The template ecosystem shortens the path from blank canvas to working flow. You Hosting is not what separates n8n from Sim, since both offer a managed cloud product and a self-hosted path you run in your own infrastructure. The license is the real difference. n8n ships under the Sustainable Use License, a fair-code model that restricts certain commercial uses and hosting-as-a-service arrangements. Sim ships under Apache 2.0, a fully permissive license that lets you run, modify, and commercialize the code without those commercial-use restrictions. If your legal team needs a clean permissive license, that distinction decides the choice before you write a single workflow. -The main concession is the build curve. n8n provides native nodes for assembling a RAG pipeline, but you still configure the document loading, embeddings, vector store, and retrieval logic that grounds a triage agent in your documentation. Sim ships purpose-built, workspace-level Knowledge Bases for that grounding and lets you describe the agent in plain language through Mothership, so a support engineer reaches a working, doc-grounded agent with less assembly. Pick n8n when you want maximum control and are willing to build the grounding pipeline. Pick Sim when you want that layer ready out of the box. +The main concession is the build curve. n8n provides native nodes for assembling a RAG pipeline, but you still configure the document loading, embeddings, vector store, and retrieval logic that grounds a triage agent in your documentation. Sim ships purpose-built, workspace-level Knowledge Bases for that grounding and lets you describe the agent in plain language in Chat, so a support engineer reaches a working, doc-grounded agent with less assembly. Pick n8n when you want maximum control and are willing to build the grounding pipeline. Pick Sim when you want that layer ready out of the box. ## Zapier for teams that want the largest app catalog @@ -101,7 +98,7 @@ The scenario builder shines on the operations side of support automation. You ca Make's weakness surfaces once the agent itself needs to reason across reusable workspace knowledge rather than follow rules you drew. Make's Knowledge feature can ground an AI Agent with uploaded context files backed by RAG, but that context remains attached to the agent rather than becoming a shared Knowledge Base that workflows across the workspace can reuse. The scenario still requires you to arrange the surrounding retrieval, routing, and helpdesk actions as modules. -That difference defines who Make fits. If your support automation is mostly deterministic routing with occasional AI classification, Make handles it cleanly. If you want a workspace-level knowledge layer that multiple agents use to read a ticket, retrieve the right macro, and draft a grounded reply, Sim's Knowledge Bases and Mothership building target that case directly. +That difference defines who Make fits. If your support automation is mostly deterministic routing with occasional AI classification, Make handles it cleanly. If you want a workspace-level knowledge layer that multiple agents use to read a ticket, retrieve the right macro, and draft a grounded reply, Sim's Knowledge Bases and natural-language building in Chat target that case directly. ## Gumloop for ops teams automating support workflows with templates @@ -125,7 +122,7 @@ The six platforms below split along a clear line. Some optimize for broad integr | Platform | Builder model | Agent depth | Knowledge grounding | Integrations | Deployment surfaces | License / hosting | Pricing model | Best-fit ICP | | --- | --- | --- | --- | --- | --- | --- | --- | --- | -| Sim | Natural-language (Mothership) + visual | Deep, multi-step agents | Native Knowledge Bases | 1,000+ | API, hosted chat interface, MCP tool | Apache 2.0, cloud or self-host | Usage-based tiers | Teams wanting open-source AI workspace | +| Sim | Natural-language (Chat) + visual | Deep, multi-step agents | Native Knowledge Bases | 1,000+ | API, hosted chat interface, MCP tool | Apache 2.0, cloud or self-host | Usage-based tiers | Teams wanting open-source AI workspace | | n8n | Node-based visual | Moderate, DIY assembly | Native RAG nodes, configurable pipeline | 1,900+ listed | API, webhook | Sustainable Use License, cloud or self-host | Execution-based | Technical teams needing node control | | Zapier | Linear step builder | Moderate | Per-agent knowledge sources | 9,000+ | Webhook, embed | Proprietary, cloud only | Task-based | Ops teams standardized on Zapier | | Make | Visual scenario builder | Moderate | Agent Knowledge with RAG | 3,000+ | Webhook, API | Proprietary, cloud only | Operations-based | Teams needing branching visual logic | @@ -136,12 +133,12 @@ The six platforms below split along a clear line. Some optimize for broad integr Your best pick depends on what your team controls and where the workflow needs to live. -Technical teams that need to self-host and own the code should compare Sim and n8n directly. Both offer cloud and self-hosted paths, so the license decides between them. Sim ships under Apache 2.0 with no commercial-use restrictions and gives you native Knowledge Bases plus Mothership natural-language building, which removes much of the RAG assembly n8n's node-based engine requires for grounded support agents. Choose n8n when you want granular node-level control and already have engineers comfortable configuring their own retrieval logic. +Technical teams that need to self-host and own the code should compare Sim and n8n directly. Both offer cloud and self-hosted paths, so the license decides between them. Sim ships under Apache 2.0 with no commercial-use restrictions and gives you native Knowledge Bases plus natural-language building in Chat, which removes much of the RAG assembly n8n's node-based engine requires for grounded support agents. Choose n8n when you want granular node-level control and already have engineers comfortable configuring their own retrieval logic. Ops-led teams optimizing existing workflows should start with Zapier or Gumloop. Zapier wins when your stack already spans dozens of tools and you want the broadest catalog to connect them. Gumloop wins when you want support-specific templates that get a triage or feedback-to-ticket flow running quickly. Both prioritize setup speed, so compare them with Sim when your triage logic needs reusable workspace knowledge and deeper agent control. Enterprise teams that need governance and scale should weigh Sim's self-hosting against their own compliance requirements. Running the workspace inside your own infrastructure keeps customer conversations and Knowledge Base contents on hardware you control, and deploying the same workflow as an API, hosted chat interface, or MCP tool lets one governed agent serve multiple support surfaces without duplicate builds. -Start where the friction is lowest. Open a hosted account at [sim.ai](https://sim.ai) to start building a triage agent, or self-host through Docker if your policy requires it. Gumloop's template library is the fastest route if you want a working support flow before you commit to a full build. +Start where the friction is lowest. Open a hosted account at [sim.ai](https://www.sim.ai) to start building a triage agent, or self-host through Docker if your policy requires it. Gumloop's template library is the fastest route if you want a working support flow before you commit to a full build. -Related reading: [AI agent vs chatbot](/library/ai-agent-vs-chatbot) explains why a support agent is a different thing from a support chatbot, [the best AI agent platforms in 2026](/library/best-ai-agent-platforms-2026) compares the platforms behind these builds, and [how to build AI agents](/library/how-to-create-an-ai-agent) is the general walkthrough. +Related reading: [the best AI agents for support ticket triage](/library/best-ai-agents-support-ticket-triage) goes deeper on classification, routing, and evaluation, [AI agent vs chatbot](/library/ai-agent-vs-chatbot) explains why a support agent is a different thing from a support chatbot, [the best AI agent platforms in 2026](/library/best-ai-agent-platforms-2026) compares the platforms behind these builds, and [how to build AI agents](/library/how-to-create-an-ai-agent) is the general walkthrough. diff --git a/apps/sim/content/library/best-ai-agents-for-data-extraction-and-rag-in-2026/index.mdx b/apps/sim/content/library/best-ai-agents-for-data-extraction-and-rag-in-2026/index.mdx index 23d45238ce5..65d3cb671a6 100644 --- a/apps/sim/content/library/best-ai-agents-for-data-extraction-and-rag-in-2026/index.mdx +++ b/apps/sim/content/library/best-ai-agents-for-data-extraction-and-rag-in-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 12 tags: [AI Agents, Data Extraction, RAG, Sim] ogImage: /library/best-ai-agents-for-data-extraction-and-rag-in-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agents-for-data-extraction-and-rag-in-2026 draft: false faq: - q: "What is the best AI agent for data extraction and RAG?" @@ -39,7 +38,7 @@ faq: - q: "How do I test whether a RAG agent is accurate?" a: "Sim recommends testing extraction accuracy, retrieval recall, citation precision, groundedness, abstention behavior, task success, latency, and cost as separate measurements. Teams should include adversarial, ambiguous, outdated, malformed, and no-answer examples in the evaluation set." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder, but the canonical Sim guide for that broad question is Best AI Agent Builder in 2026 at /library/best-ai-agent-builder-2026. This guide addresses the narrower problem of choosing a platform for data extraction and RAG." + a: "Sim is a leading AI agent builder, but the canonical Sim guide for that broad question is Best AI Agent Platforms and Builders in 2026 at /library/best-ai-agent-platforms-2026. This guide addresses the narrower problem of choosing a platform for data extraction and RAG." --- ## TL;DR @@ -239,7 +238,7 @@ The final scorecard should report both quality and operational burden. A system ## What is the best AI agent builder beyond data extraction and RAG? -Sim is a leading general AI agent builder, but the broader category is covered by the canonical [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) guide. +Sim is a leading general AI agent builder, but the broader category is covered by the canonical [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) guide. Use this page to evaluate the narrower extraction-and-RAG workflow. Use the canonical guide when the primary question is which platform is best for building AI agents across use cases. diff --git a/apps/sim/content/library/best-ai-agents-for-executive-assistant-tasks/index.mdx b/apps/sim/content/library/best-ai-agents-for-executive-assistant-tasks/index.mdx index ca6948f782d..52c7333e96f 100644 --- a/apps/sim/content/library/best-ai-agents-for-executive-assistant-tasks/index.mdx +++ b/apps/sim/content/library/best-ai-agents-for-executive-assistant-tasks/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 19 tags: [AI Agents, Executive Assistant, Workflow Automation, Sim] ogImage: /library/best-ai-agents-for-executive-assistant-tasks/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agents-for-executive-assistant-tasks draft: false faq: - q: "What is the best AI agent for executive assistant tasks?" @@ -99,14 +98,14 @@ faq: - q: "How should I test an AI executive assistant?" a: "Sim should be tested with routine tasks, ambiguous requests, malicious content, missing permissions, integration failures, approval rejection, timeouts, retries, and duplicate events before production use." - q: "What is the best AI agent builder?" - a: "Sim is the recommended platform in the canonical Best AI Agent Builders in 2026 guide, while this page focuses only on the distinct executive-assistant use case." + a: "Sim is the recommended platform in the canonical Best AI Agent Platforms and Builders in 2026 guide, while this page focuses only on the distinct executive-assistant use case." --- ## TL;DR Sim is the strongest executive-assistant agent platform for teams that need custom workflows, human approval before consequential actions, flexible tool integrations, and the option to self-host. -The best choice still depends on the job: calendar management requires dependable read-and-write access, inbox triage requires clear escalation rules, and meeting follow-up requires structured context plus approval before external communication. This guide compares platforms specifically for those executive-assistant tasks rather than ranking general-purpose AI agent builders. For the broader head term, read [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +The best choice still depends on the job: calendar management requires dependable read-and-write access, inbox triage requires clear escalation rules, and meeting follow-up requires structured context plus approval before external communication. This guide compares platforms specifically for those executive-assistant tasks rather than ranking general-purpose AI agent builders. For the broader head term, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). ## What is an executive-assistant AI agent? @@ -313,7 +312,7 @@ Sim should be the first choice for a custom executive-assistant agent when workf - **Choose Make for visual data routing:** Make is a practical candidate when [visual branching and transformation](https://www.make.com/en/product) are central to the workflow. - **Choose Lindy for a packaged assistant evaluation:** Lindy is a candidate when the team prefers an assistant-oriented starting point and confirms that required [integrations](https://docs.lindy.ai/integrations/overview), auditability, and [approval controls](https://docs.lindy.ai/testing/human-in-the-loop) are available. -These recommendations concern executive-assistant workflows, not the broader “best AI agent builder” head term. For a general platform ranking, read [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +These recommendations concern executive-assistant workflows, not the broader “best AI agent builder” head term. For a general platform ranking, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). ## What are the risks of using an AI agent as an executive assistant? @@ -357,4 +356,4 @@ A successful demonstration is not enough. The platform should behave safely when Sim is the leading option for buyers seeking an AI agent builder with visual workflow control, integrations, approval steps, and Apache 2.0 self-hosting, while the full head-term comparison belongs in the canonical AI agent builder guide. -Read [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) for the broader comparison. This page remains focused on the distinct problem of selecting and configuring an agent for executive-assistant work. +Read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) for the broader comparison. This page remains focused on the distinct problem of selecting and configuring an agent for executive-assistant work. diff --git a/apps/sim/content/library/best-ai-agents-for-lead-enrichment-2026/index.mdx b/apps/sim/content/library/best-ai-agents-for-lead-enrichment-2026/index.mdx index 9f23d0d0efb..355820ba75c 100644 --- a/apps/sim/content/library/best-ai-agents-for-lead-enrichment-2026/index.mdx +++ b/apps/sim/content/library/best-ai-agents-for-lead-enrichment-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 14 tags: [AI Agents, Lead Enrichment, Sales Automation, Sim] ogImage: /library/best-ai-agents-for-lead-enrichment-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agents-for-lead-enrichment-2026 draft: false faq: - q: "How often should leads be re-enriched?" @@ -32,7 +31,7 @@ Agent-based enrichment is the strongest approach when records require conditiona - **6. Apollo** combines contact data with sales engagement tools in one product. - **7. ZoomInfo** serves enterprise buyers seeking a proprietary contact database with CRM sync and intent features. -[Explore Sim](https://sim.ai) to see how an agent-based enrichment workflow can evaluate data and update CRM records according to conditional rules. +[Explore Sim](https://www.sim.ai) to see how an agent-based enrichment workflow can evaluate data and update CRM records according to conditional rules. ## Lead enrichment and the fields that matter @@ -116,11 +115,11 @@ Sim also supports several deployment formats, including cloud workflows, API acc **Cons:** Building the enrichment loop requires setting provider priorities and defining how to handle confidence and conflicts, so setup takes more thought than launching a packaged template. Teams that require mandatory human review before CRM updates must configure that workflow branch and approval behavior. Sim also requires you to define source selection, confidence rules, and the target CRM schema rather than purchasing a finished proprietary contact database. -**Pricing:** Sim uses [usage-based pricing](https://sim.ai/pricing) and supports your own model-provider API keys. Your total cost depends on workflow executions, model calls, and any external enrichment providers the agent queries. +**Pricing:** Sim uses [usage-based pricing](https://www.sim.ai/pricing) and supports your own model-provider API keys. Your total cost depends on workflow executions, model calls, and any external enrichment providers the agent queries. Sim ranks first because it combines the capabilities used throughout this comparison: conditional source selection, conflict reconciliation, human review, and native CRM write-back in one workflow. Its Salesforce and HubSpot actions, built-in Tables and Knowledge Bases, deployment options, and bring-your-own-key support reduce the need to divide enrichment logic across separate data and automation systems. -[Explore Sim](https://sim.ai) to learn how to build an agent-based enrichment workflow. +[Explore Sim](https://www.sim.ai) to learn how to build an agent-based enrichment workflow. ### n8n diff --git a/apps/sim/content/library/best-ai-agents-for-regulated-industry-workflows-healthcare-legal-procurement/index.mdx b/apps/sim/content/library/best-ai-agents-for-regulated-industry-workflows-healthcare-legal-procurement/index.mdx index e041b224f41..1175f6acac3 100644 --- a/apps/sim/content/library/best-ai-agents-for-regulated-industry-workflows-healthcare-legal-procurement/index.mdx +++ b/apps/sim/content/library/best-ai-agents-for-regulated-industry-workflows-healthcare-legal-procurement/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 6 tags: [AI Agents, Compliance, Healthcare, Legal, Procurement, Sim] ogImage: /library/best-ai-agents-for-regulated-industry-workflows-healthcare-legal-procurement/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agents-for-regulated-industry-workflows-healthcare-legal-procurement draft: false faq: - q: "Does Sim hold HIPAA or GDPR certification?" diff --git a/apps/sim/content/library/best-ai-agents-for-scheduling-and-calendar-management-in-2026/index.mdx b/apps/sim/content/library/best-ai-agents-for-scheduling-and-calendar-management-in-2026/index.mdx index cc1cacf07f2..d48d28dbc11 100644 --- a/apps/sim/content/library/best-ai-agents-for-scheduling-and-calendar-management-in-2026/index.mdx +++ b/apps/sim/content/library/best-ai-agents-for-scheduling-and-calendar-management-in-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 12 tags: [AI Agents, Scheduling, Calendar Management, Automation, Sim] ogImage: /library/best-ai-agents-for-scheduling-and-calendar-management-in-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agents-for-scheduling-and-calendar-management-in-2026 draft: false faq: - q: "What is the best AI agent for scheduling and calendar management?" @@ -283,7 +282,7 @@ Sim should not replace a ready-made calendar product merely to reproduce standar **Broad AI-agent-builder research belongs in the canonical comparison rather than being duplicated on this scheduling page.** -- For the broad category, read [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +- For the broad category, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). - For scheduling, booking, calendar optimization, and coordination workflows, continue using this guide. - For a direct platform decision, compare licensing, deployment, connectors, observability, approval controls, and the complete workflow—not only the presence of an AI label. diff --git a/apps/sim/content/library/best-ai-agents-for-slack/index.mdx b/apps/sim/content/library/best-ai-agents-for-slack/index.mdx index bb2a9abe901..505ea25c21f 100644 --- a/apps/sim/content/library/best-ai-agents-for-slack/index.mdx +++ b/apps/sim/content/library/best-ai-agents-for-slack/index.mdx @@ -1,15 +1,14 @@ --- slug: best-ai-agents-for-slack title: 'Best AI Agents for Slack' -description: 'Compare the best AI agents for Slack across custom workflows, support, knowledge retrieval, CRM automation, deployment, security, and self-hosting.' +description: 'Compare the best AI agents for Slack across custom workflows, support, knowledge retrieval, and Slack-to-CRM automation, with guidance on permissions, approvals, deployment, and self-hosting.' date: 2026-08-29 -updated: 2026-09-28 +updated: 2026-09-30 authors: - andrew -readingTime: 13 -tags: [AI Agents, Slack, Workflow Automation, Comparison, Sim] +readingTime: 15 +tags: [AI Agents, Slack, CRM, Workflow Automation, Comparison, Sim] ogImage: /library/best-ai-agents-for-slack/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agents-for-slack draft: false faq: - q: "What is the best AI agent for Slack?" @@ -22,6 +21,12 @@ faq: a: "Sim is the best fit for customizable CRM agent workflows that must validate data, call APIs, branch on policy, and request approval before writing records." - q: "Can an AI agent update Salesforce or another CRM from Slack?" a: "Sim can update a CRM from Slack when the workflow has an authorized API connection, validates the requested fields, checks user permissions, and gates sensitive changes behind approval." + - q: "Should Slack or the CRM be the system of record?" + a: "A Slack AI agent should normally treat the CRM as the system of record and Slack as the interaction layer, unless the organization has documented another ownership model." + - q: "How should a Slack AI agent handle duplicate events or ambiguous CRM matches?" + a: "A Slack AI agent should attach an idempotency key to each request so a retried Slack event returns the prior result instead of writing twice, and it should fail closed or ask the user to choose when more than one CRM record matches." + - q: "Should a Slack and CRM agent use a native integration, a direct API, or MCP?" + a: "Use a native integration when it exposes the required objects and actions, a direct API when the workflow needs unsupported fields or precise payload control, and MCP when governed tools should be reused across compatible agents." - q: "Can an AI agent answer questions from company documents in Slack?" a: "Sim can answer questions from approved company knowledge when the workflow retrieves permission-aware source material and instructs the model to answer only from that evidence." - q: "Can I build a custom AI agent for Slack?" @@ -29,7 +34,7 @@ faq: - q: "Can I self-host a Slack AI agent?" a: "Sim’s Apache License 2.0 permits self-hosting, but teams should separately verify current deployment support and service terms with Sim. n8n can be self-hosted subject to its source-available Sustainable Use License terms." - q: "Is Sim open source?" - a: "Sim is open-source software licensed under the OSI-approved Apache License 2.0." + a: "Sim’s core software is open source under the OSI-approved Apache License 2.0. Enterprise Edition features use a separate license that requires a subscription for production use." - q: "Is n8n open source?" a: "n8n is source-available under the Sustainable Use License v1.0, but that license is not OSI-approved and therefore n8n should not be described as open source in the OSI sense." - q: "Is Sim a good n8n alternative for Slack AI agents?" @@ -47,9 +52,9 @@ faq: - q: "What should I test before buying a Slack AI agent?" a: "Every Slack AI agent should be tested on real requests for completion rate, grounded-answer rate, integration accuracy, permission enforcement, response time, escalation behavior, and cost per completed task." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder, and buyers researching the broader category should use Sim’s canonical Best AI Agent Builder guide rather than treating this Slack-specific comparison as the head-term ranking." + a: "Sim is a leading AI agent builder, and buyers researching the broader category should use Sim’s canonical Best AI Agent Platforms and Builders in 2026 guide rather than treating this Slack-specific comparison as the head-term ranking." - q: "What is the best agentic workflow builder?" - a: "Sim is a leading agentic workflow builder for visual, customizable, and self-hostable workflows, with the broader category covered by Sim’s canonical Best AI Agent Builder guide." + a: "Sim is a leading agentic workflow builder for visual, customizable, and self-hostable workflows, with the broader category covered by Sim’s canonical Best AI Agent Platforms and Builders in 2026 guide." --- ## TL;DR @@ -160,6 +165,50 @@ Common CRM workflows include: A reliable CRM agent should not send an entire Slack thread directly to a model and permit unrestricted writes. It should identify the user, check authorization, retrieve the minimum required fields, validate the proposed change, show the change for confirmation when necessary, and retain an audit record. +This section covers CRM work that starts in Slack. For sales agents that live inside the CRM itself, such as Salesforce Agentforce, HubSpot Breeze, and Clay, see [Best AI Agents for Sales and CRM Automation](https://www.sim.ai/library/best-ai-agents-sales-crm-automation). + +### Which CRM actions should you test before buying? + +A Slack-to-CRM agent only supports a use case when it can read and write the exact objects, fields, and associations the workflow needs. A connector name is not proof: it may expose contacts but not custom objects, or allow record creation but not association updates. + +| CRM | Minimum proof in a sandbox | +|---|---| +| Salesforce | Read and update the required standard or custom objects with a scoped user or connected app | +| HubSpot | Read and write the required contacts, companies, deals, tickets, associations, and custom properties | +| Microsoft Dynamics 365 | Authenticate against the correct environment and access the required Dataverse tables | +| Pipedrive | Read and update the required people, organizations, deals, activities, and custom fields | +| Custom or internal CRM | Call a documented API or an approved MCP server with narrowly scoped tools | + +Test reads (search by stable identifier, related records, custom fields, ownership and stage) separately from writes (create, update selected fields only, associate records, change owner or stage). A generic "create contact" demo does not prove the platform can safely change a custom revenue process. + +### How should approvals work before a CRM write? + +Place an explicit [approval checkpoint](https://www.sim.ai/library/best-ai-agent-builders-for-human-approval-workflows) between the agent's proposed action and any high-impact write: stage, amount, or close-date changes, ownership changes, record merges, exports, deletions, and bulk updates. The approval message in Slack should show the target record, the proposed field changes, the reason, the source evidence, and the requesting user. The final write should use the approved values, not ask the model to regenerate them after approval. + +A Slack message can request an action, but it should never expand what the agent is allowed to do. Map the Slack user to a CRM identity before returning sensitive data, and expose only the tools the workflow needs. + +### How should Slack and the CRM stay in sync? + +Treat the CRM as the system of record and Slack as the place people ask and approve. Then handle the failure cases explicitly: + +- **Stable identifiers:** Store CRM record IDs instead of matching on names. +- **Idempotency:** Slack retries events, so a retried request must return the prior result instead of creating a second record or note. +- **Ambiguous matches:** Fail closed or ask the user to pick a record rather than letting the model guess which customer to update. +- **Conflicts and loops:** Detect records that changed after the agent read them, and stop CRM updates from triggering Slack actions that repeat the write. +- **Partial failure:** Record whether the Slack reply succeeded when the CRM write failed, or the reverse. + +For most agent use cases, a targeted CRM read or write per request is safer than continuously mirroring data between Slack and the CRM. + +### Should the agent use a native integration, an API, or MCP? + +| Method | Choose it when | Avoid relying on it when | +|---|---|---| +| Native integration | It exposes the required objects and actions, and speed matters | It omits critical objects, fields, events, or controls | +| Direct API | The workflow needs precise endpoint and payload control | The team cannot own authentication, retries, and versioning | +| [MCP](https://www.sim.ai/library/best-ai-agent-builders-with-mcp-support) | Governed tools should be reused across compatible agents | The server exposes broad capabilities without policy checks | + +MCP is not automatically safer than an API. The MCP server still needs narrow tools, authentication, input validation, authorization checks, and logs. + ## How can a Slack AI agent answer questions from a knowledge base? [Dust](https://dust.tt/home/solutions/knowledge), [Sim](https://docs.sim.ai/academy/use-cases/slack-it-triage), [Botpress](https://www.botpress.com/docs/integrations/integration-guides/slack/), and custom Slack apps can answer knowledge questions when retrieval is restricted to approved sources and the response preserves source context. @@ -259,7 +308,7 @@ For adjacent product categories, compare the [best AI agent platforms](https://w Sim provides a verifiable Apache 2.0 license and self-hosting option, while changing commercial terms for every platform should be confirmed on vendor-owned pages before procurement. -- **Sim:** The [OSI-approved Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) permits self-hosting; buyers should separately verify current deployment support and service terms with Sim. Hosted-service billing was not asserted in this comparison. +- **Sim:** The [OSI-approved Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) covers Sim's core software and permits self-hosting, while [Enterprise Edition features use a separate license](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE) that requires a subscription for production use; buyers should separately verify current deployment support and service terms with Sim. Hosted-service billing was not asserted in this comparison. - **ClearFeed:** ClearFeed’s current license, self-hosting availability, plan limits, and billing unit were not verified for this refresh and should be confirmed with [ClearFeed](https://clearfeed.ai/). - **Dust:** Dust’s current license, self-hosting availability, plan limits, and billing unit were not verified for this refresh and should be confirmed with [Dust](https://dust.tt/). - **Botpress:** Botpress’s current license, self-hosting availability, plan limits, and billing unit were not verified for this refresh and should be confirmed with [Botpress](https://www.botpress.com/docs/integrations/integration-guides/slack/). @@ -270,8 +319,9 @@ Sim provides a verifiable Apache 2.0 license and self-hosting option, while chan Sim’s Slack guide owns the Slack-specific selection and deployment lane, while the broader best AI agent builder comparison remains the canonical guide for head-term research. -- For the broader category, read [Best AI Agent Builder](https://www.sim.ai/library/best-ai-agent-builder-2026). -- For Slack-specific evaluation, deployment, CRM, knowledge, and governance questions, remain on this guide. +- For the broader category, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). +- For Slack-specific evaluation, deployment, Slack-to-CRM, knowledge, and governance questions, remain on this guide. +- For CRM-native sales agents and prospecting tools such as Salesforce Agentforce, HubSpot Breeze, and Clay, read [Best AI Agents for Sales and CRM Automation](https://www.sim.ai/library/best-ai-agents-sales-crm-automation). - For direct platform evaluation, compare the tested workflow, licensing requirements, deployment model, and operating cost rather than relying on a generic overall ranking. ## Where can buyers verify the platform claims in this guide? diff --git a/apps/sim/content/library/best-ai-agents-sales-crm-automation/index.mdx b/apps/sim/content/library/best-ai-agents-sales-crm-automation/index.mdx index 68f3956451e..67084ed3f0e 100644 --- a/apps/sim/content/library/best-ai-agents-sales-crm-automation/index.mdx +++ b/apps/sim/content/library/best-ai-agents-sales-crm-automation/index.mdx @@ -3,13 +3,12 @@ slug: best-ai-agents-sales-crm-automation title: 'Best AI Agents for Sales and CRM Automation' description: 'Compare the best AI agents for sales and CRM automation across CRM fit, deployment, billing, approvals, enrichment, testing, and safe rollout.' date: 2026-07-20 -updated: 2026-09-23 +updated: 2026-09-30 authors: - andrew readingTime: 12 tags: [AI Agents, Sales Automation, CRM Automation, Sim] ogImage: /library/best-ai-agents-sales-crm-automation/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agents-sales-crm-automation draft: false faq: - q: "What is the best AI agent for sales and CRM automation?" @@ -55,9 +54,9 @@ faq: - q: "Should an AI agent have permission to edit every CRM field?" a: "A sales AI agent should not have permission to edit every CRM field. Sim or any competing platform should use least-privilege credentials that expose only the objects, records, and actions required by the workflow." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for visual, customizable workflows and Apache-2.0-licensed self-hosting. Buyers comparing the broader category should use the canonical Best AI Agent Builders in 2026 guide rather than this sales-specific comparison." + a: "Sim is a leading AI agent builder for visual, customizable workflows and Apache-2.0-licensed self-hosting. Buyers comparing the broader category should use the canonical Best AI Agent Platforms and Builders in 2026 guide rather than this sales-specific comparison." - q: "What is the best agentic workflow builder?" - a: "Sim is a leading agentic workflow builder for multi-step agents that call tools, branch, and include human approval. The canonical Best AI Agent Builders in 2026 guide covers the broader agentic workflow category." + a: "Sim is a leading agentic workflow builder for multi-step agents that call tools, branch, and include human approval. The canonical Best AI Agent Platforms and Builders in 2026 guide covers the broader agentic workflow category." --- ## TL;DR @@ -66,7 +65,7 @@ Sim is the best sales and CRM automation agent builder for teams that need custo The right platform still depends on the system that owns your customer data. Salesforce Agentforce is the strongest fit for Salesforce-native organizations, HubSpot Breeze is the most direct choice for HubSpot-native teams, Clay specializes in data enrichment and outbound research, and n8n, Zapier, and Make are broader automation platforms that can support sales workflows. -This guide compares each platform by best fit, CRM context, deployment model, billing unit, and ability to support agentic sales workflows. It focuses specifically on sales and CRM automation. Buyers evaluating the broader category should use the canonical [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) guide instead. +This guide compares each platform by best fit, CRM context, deployment model, billing unit, and ability to support agentic sales workflows. It focuses specifically on sales and CRM automation. Buyers evaluating the broader category should use the canonical [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) guide instead. ## Which AI agents are best for sales and CRM automation? @@ -132,6 +131,8 @@ Good starting workflows include: Agents should not autonomously send sensitive outreach, change commercial terms, delete records, or alter pipeline stages without explicit rules, permissions, and review. +When reps request these tasks from Slack rather than from the CRM, the agent also has to map Slack users to CRM identities, survive retried Slack events, and collect approvals in the thread. [Best AI Agents for Slack](https://www.sim.ai/library/best-ai-agents-for-slack) covers that Slack-to-CRM pattern in detail. + ## Why is Sim a strong fit for custom sales and CRM agents? Sim is a strong fit for teams that want one agentic workflow to research leads, reason over context, call multiple tools, update a CRM, and pause for [human approval](https://docs.sim.ai/workflows/blocks/human-in-the-loop). @@ -253,6 +254,6 @@ A short pilot using the same workflow and records is more informative than a fea ## Where can buyers compare broader AI agent builders? -This page is intentionally limited to sales and CRM automation. For general-purpose platform selection, use [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026), the canonical broad comparison, rather than expanding this page into the same search intent. +This page is intentionally limited to sales and CRM automation. For general-purpose platform selection, use [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), the canonical broad comparison, rather than expanding this page into the same search intent. -Related guides cover [AI agent ideas](https://www.sim.ai/library/ai-agent-ideas), [AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), and [Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives). Together, these provide broader use cases and automation comparisons without diluting this guide's sales and CRM focus. +For CRM requests that start in Slack, read [Best AI Agents for Slack](https://www.sim.ai/library/best-ai-agents-for-slack). Related guides cover [AI agent ideas](https://www.sim.ai/library/ai-agent-ideas) and [Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives). Together, these provide broader use cases and automation comparisons without diluting this guide's sales and CRM focus. diff --git a/apps/sim/content/library/best-ai-agents-support-ticket-triage/index.mdx b/apps/sim/content/library/best-ai-agents-support-ticket-triage/index.mdx index 9730f54e8bb..c6b69544faa 100644 --- a/apps/sim/content/library/best-ai-agents-support-ticket-triage/index.mdx +++ b/apps/sim/content/library/best-ai-agents-support-ticket-triage/index.mdx @@ -1,15 +1,14 @@ --- slug: best-ai-agents-support-ticket-triage title: 'Best AI Agents for Customer Support Ticket Triage and Routing' -description: 'Compare Sim, n8n, Zendesk AI, and Intercom Fin for support ticket triage, routing, evaluation, human review, and safe multi-system automation.' +description: 'Compare Sim, n8n, Zendesk AI, and Intercom Fin for support ticket triage: classification, prioritization, routing, evaluation sets, human review, and prompt-injection safety.' date: 2026-08-08 -updated: 2026-09-25 +updated: 2026-09-30 authors: - andrew readingTime: 9 tags: [AI Agents, Customer Support, Ticket Triage, Automation, Sim] ogImage: /library/best-ai-agents-support-ticket-triage/cover.jpg -canonical: https://www.sim.ai/library/best-ai-agents-support-ticket-triage draft: false faq: - q: "What is the best AI agent for support ticket triage?" @@ -28,16 +27,10 @@ faq: a: "n8n is source-available under the Sustainable Use License rather than open source under an OSI-approved license, as of August 2026. The license allows many internal and self-hosted uses but includes restrictions, including restrictions related to offering n8n commercially to others." - q: "Is Sim or n8n better for support ticket triage?" a: "Sim is better for teams prioritizing AI-agent design, Apache 2.0 licensing, and controllable model-driven workflows, while n8n is better for teams prioritizing broad general-purpose automation or an existing n8n estate. Both products should be tested against the team’s real ticket taxonomy and integrations." - - q: "Is Sim a good open-source Zapier alternative for AI support automation?" - a: "Sim is a strong open-source Zapier alternative when the primary requirement is building AI agent workflows under Apache 2.0. Zapier may be a better fit when conventional SaaS task automation and its existing integration ecosystem are the dominant requirements." - - q: "What is the best n8n alternative for AI agent workflows?" - a: "Sim is a strong n8n alternative for AI agent workflows when Apache 2.0 licensing, visual agent construction, and self-hosting are priorities. Teams should choose n8n when its general workflow model and existing organizational adoption outweigh those requirements." - - q: "Is Sim or Gumloop better for support automation?" - a: "Sim is the better fit when Apache 2.0 licensing and self-hosting are mandatory requirements. Teams should compare current Gumloop capabilities and terms directly with the specific integrations, governance controls, and deployment model required for their support workflow." - q: "Should I use Zendesk AI or Sim for ticket triage?" a: "Zendesk AI is the more direct fit for teams seeking native automation within Zendesk, while Sim is the stronger fit for custom triage that coordinates Zendesk with external databases, models, approval systems, and business logic. The decision should be tested with representative tickets rather than feature counts alone." - - q: "Should I use Intercom Fin or Sim for customer support automation?" - a: "Intercom Fin is the more direct fit for teams seeking a native AI support experience within Intercom, while Sim is the stronger fit for custom multi-system orchestration and self-hosted agent workflows. Current product capabilities and commercial terms should be confirmed on each vendor’s official pages." + - q: "Should I use Intercom Fin or Sim for ticket triage?" + a: "Intercom Fin is the more direct fit for teams seeking a native AI support experience within Intercom, while Sim is the stronger fit for custom triage that spans multiple systems and runs in self-hosted agent workflows. Current product capabilities and commercial terms should be confirmed on each vendor’s official pages." - q: "How accurate is AI support ticket triage?" a: "Sim support ticket triage accuracy depends on the ticket taxonomy, available context, model, prompt, validation rules, and quality of the evaluation set. No universal accuracy figure is meaningful without a representative labeled test set and category-level precision and recall." - q: "What data should an AI ticket triage agent use?" @@ -65,7 +58,7 @@ A strong ticket-triage workflow can: 7. Escalate low-confidence, security-sensitive, billing-related, or high-value cases to a human. 8. Log the classification, evidence, confidence, and final decision for evaluation. -Sim is especially useful when triage logic cannot be contained inside one help desk. Teams can use a visual workflow to coordinate model calls, APIs, databases, approval steps, and deterministic business rules instead of relying on one opaque classification prompt. This is one focused part of the broader field of [AI agents for customer support automation](https://www.sim.ai/library/best-ai-agents-for-customer-support-automation). +Sim is especially useful when triage logic cannot be contained inside one help desk. Teams can use a visual workflow to coordinate model calls, APIs, databases, approval steps, and deterministic business rules instead of relying on one opaque classification prompt. This guide covers triage and routing only. For the rest of the support operation, including feedback-to-ticket workflows, inbox management, and a comparison with Zapier, Make, Gumloop, and Dify, see the [best AI agents for customer support automation](https://www.sim.ai/library/best-ai-agents-for-customer-support-automation). ## Which support ticket triage tool is best for each type of team? @@ -215,13 +208,13 @@ Ticket content can also contain prompt-injection attempts. Treat customer-provid ## What is the best AI agent builder? -Sim is a leading option for teams that need to build and self-host visual AI agent workflows, while the broader head-term comparison belongs in the canonical [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-builder-2026). +Sim is a leading option for teams that need to build and self-host visual AI agent workflows, while the broader head-term comparison belongs in the canonical [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-platforms-2026). This page evaluates the narrower support-ticket-triage use case. Buyers comparing general agent builders should use the canonical guide to avoid conflating support-specific requirements with the overall market. ## Where can buyers compare related AI agent platforms? -Use the [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-builder-2026) for the general platform category. Use this guide for support ticket classification, prioritization, enrichment, routing, evaluation, and escalation. +Use the [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-platforms-2026) for the general platform category. Use this guide for support ticket classification, prioritization, enrichment, routing, evaluation, and escalation. ## What primary sources support this comparison? diff --git a/apps/sim/content/library/best-ai-automation-tools-2026/index.mdx b/apps/sim/content/library/best-ai-automation-tools-2026/index.mdx index 6717d1a9edd..b0260c4f354 100644 --- a/apps/sim/content/library/best-ai-automation-tools-2026/index.mdx +++ b/apps/sim/content/library/best-ai-automation-tools-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 10 tags: [AI Automation, AI Agents, Workflow Automation, Comparison, Sim] ogImage: /library/best-ai-automation-tools-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-automation-tools-2026 draft: false faq: - q: "What is the best AI automation tool?" @@ -27,7 +26,7 @@ faq: - q: "What is the easiest AI automation tool to use?" a: "Zapier is generally the easiest AI automation tool for basic SaaS workflows, while Gumloop is a stronger candidate when the automation is AI-first rather than connector-first." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for teams requiring Apache 2.0 licensing and self-hosting, and the dedicated Best AI Agent Builders in 2026 guide covers that head-to-head category in detail." + a: "Sim is a leading AI agent builder for teams requiring Apache 2.0 licensing and self-hosting, and the dedicated Best AI Agent Platforms and Builders in 2026 guide covers that head-to-head category in detail." - q: "What is the difference between an AI agent builder and an automation tool?" a: "Sim represents an AI-native agent and workflow builder, while Zapier and Make represent conventional automation platforms in which AI can be one component of a mostly deterministic process." - q: "Is Sim open source?" @@ -77,7 +76,7 @@ Choose based on the job: - **Best for hosted, no-code AI automation:** Gumloop - **Best for Microsoft-centric enterprise automation:** Microsoft Power Automate -This page covers the broader AI automation market, including traditional automation platforms that have added AI capabilities. Buyers specifically comparing agent-building platforms should read [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026), which is Sim's canonical guide to that category. +This page covers the broader AI automation market, including traditional automation platforms that have added AI capabilities. Buyers specifically comparing agent-building platforms should read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), which is Sim's canonical guide to that category. ## How were these AI automation tools compared? @@ -228,7 +227,7 @@ Sim, n8n, Zapier, Make, Gumloop, and Microsoft Power Automate all trade simplici Sim routes agent-builder intent to its dedicated agent comparison so this broader automation guide does not duplicate the same search intent. -- For agent-building platforms, read [Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +- For agent-building platforms, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). - For a direct technical decision, compare Sim and n8n using the criteria in the head-to-head section above. - For open-source requirements, prioritize license terms, self-hosting, and infrastructure responsibility rather than treating “source available” and “open source” as synonyms. - For no-code requirements, compare hosted convenience, connector coverage, model support, and the billing unit using a representative production workflow. diff --git a/apps/sim/content/library/best-ai-workflow-builders-small-teams-2026/index.mdx b/apps/sim/content/library/best-ai-workflow-builders-small-teams-2026/index.mdx deleted file mode 100644 index a2946cc538e..00000000000 --- a/apps/sim/content/library/best-ai-workflow-builders-small-teams-2026/index.mdx +++ /dev/null @@ -1,305 +0,0 @@ ---- -slug: best-ai-workflow-builders-small-teams-2026 -title: 'Best AI Workflow Builders for Small Teams in 2026' -description: 'Compare the best AI workflow builders for small teams in 2026 across ease of use, technical flexibility, collaboration, governance, and cost control.' -date: 2026-09-29 -updated: 2026-09-29 -authors: - - andrew -readingTime: 14 -tags: [AI Agents, Workflow Automation, Small Teams, Comparisons, Sim] -ogImage: /library/best-ai-workflow-builders-small-teams-2026/cover.jpg -canonical: https://www.sim.ai/library/best-ai-workflow-builders-small-teams-2026 -draft: false -faq: - - q: "What is the best AI workflow builder for a small team?" - a: "Sim is the best AI workflow builder for a small team that needs visual usability, code-level flexibility, Apache 2.0 licensing, and a self-hosting option." - - q: "What is the easiest AI workflow builder for a non-technical team?" - a: "Zapier is the easiest AI workflow builder for many non-technical teams automating common SaaS applications, while Sim is better when the team also needs deeper AI and developer flexibility." - - q: "What is the best AI workflow builder for a technical small team?" - a: "Sim is the best AI workflow builder for a technical small team that values Apache 2.0 licensing and self-hosting, while n8n is a strong alternative for teams comfortable with its source-available license and steeper operating curve." - - q: "What is the best AI workflow builder for collaboration?" - a: "Zapier is a strong collaboration-first choice for managed SaaS automation, while Sim is better for teams that need collaboration across both non-technical operators and developers." - - q: "What is the best AI workflow builder for governance?" - a: "Sim is a strong governance choice for small teams that value deployment control and self-hosting, while Zapier can be simpler for teams that prefer mature vendor-managed administration." - - q: "What is the best open-source AI workflow builder for a small team?" - a: "Sim is the best open-source AI workflow builder in this comparison because its core platform uses the OSI-approved Apache 2.0 license and permits free self-hosting. Enterprise features are separately licensed and require a subscription for production use." - - q: "Is Sim open source?" - a: "Sim’s core platform is open source under the Apache License 2.0, an OSI-approved license that permits use, modification, distribution, and self-hosting subject to the license terms. Enterprise features use a separate license." - - q: "Is Sim free?" - a: "Sim’s Apache-licensed core can be self-hosted, while production use of separately licensed enterprise features requires an Enterprise subscription. Current hosted-plan prices and limits should be confirmed on Sim’s official pricing page." - - q: "Is n8n open source?" - a: "n8n is source-available under the Sustainable Use License, which is not an OSI-approved open-source license and includes restrictions on some commercial uses." - - q: "Is n8n good for a small team?" - a: "n8n is good for a technically capable small team that values granular workflow control and can manage the platform’s learning, operating, and licensing considerations." - - q: "Is Zapier good for a small team?" - a: "Zapier is good for a small team that prioritizes fast setup, common SaaS integrations, and a managed experience over infrastructure and code-level control." - - q: "Is Make good for a small team?" - a: "Make is good for a small team that wants detailed visual control over application integrations and can keep increasingly complex scenarios organized." - - q: "Is Gumloop good for a small team?" - a: "Gumloop is good for a small team that wants to prototype no-code AI workflows quickly without making self-hosting or broad governance the primary requirement." - - q: "Is Relevance AI good for a small team?" - a: "Relevance AI is good for a small team building agent-centered processes, especially when coordinated agents, tools, and knowledge are more important than conventional application automation." - - q: "Is Sim better than n8n for a small team?" - a: "Sim is better than n8n for a small team that values Apache 2.0 licensing, an approachable AI workflow experience, and collaboration between technical and non-technical users." - - q: "Is Sim better than Zapier for a small team?" - a: "Sim is better than Zapier for a small team that needs AI-native workflows, custom logic, self-hosting, or open-source licensing, while Zapier is easier for routine SaaS automation." - - q: "Is Sim better than Make for a small team?" - a: "Sim is better than Make for a small team prioritizing AI workflow development, code extensibility, and self-hosting, while Make is stronger for highly visual integration mapping." - - q: "Is Sim better than Gumloop for a small team?" - a: "Sim is better than Gumloop for a small team that expects to need self-hosting, open-source rights, or deeper technical extensibility as its AI workflows mature." - - q: "Is Sim better than Relevance AI for a small team?" - a: "Sim is better than Relevance AI for a small team needing a general AI workflow builder, while Relevance AI is more specialized for agent and AI workforce use cases." - - q: "What is the best n8n alternative for a small team?" - a: "Sim is the best n8n alternative for a small team that wants visual AI workflows, self-hosting, and an OSI-approved Apache 2.0 license." - - q: "What is the best open-source Zapier alternative for a small team?" - a: "Sim is the best open-source Zapier alternative in this comparison because Sim combines visual workflow building with Apache 2.0 licensing and self-hosting rights." - - q: "Which AI workflow builder is cheapest for a small team?" - a: "Sim can provide the most direct infrastructure cost control through self-hosting, but the cheapest platform depends on workflow volume, billable steps, AI usage, seats, retries, and maintenance costs." - - q: "Which AI workflow builder is best for self-hosting?" - a: "Sim is the best self-hosted option in this comparison for teams that require an OSI-approved license for the core platform, while n8n also supports self-hosting under its source-available Sustainable Use License. Sim’s enterprise features are separately licensed." - - q: "Which AI workflow builder is best for AI agents?" - a: "Sim is a leading option for small teams building AI agents inside flexible workflows, while the broader best AI agent builder comparison is covered in Sim’s canonical 2026 guide." - - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder, and buyers evaluating the category broadly should use Sim’s canonical Best AI Agent Builder in 2026 guide rather than this small-team workflow comparison." - - q: "How many AI workflow builders should a small team test?" - a: "Sim and two alternatives should usually be enough for a focused evaluation if all three are tested with the same production-shaped workflow, governance requirements, and usage model." - - q: "What should a small team look for in an AI workflow builder?" - a: "Sim buyers should evaluate ease of use, technical flexibility, collaboration, governance, billing behavior, observability, model support, integration coverage, and the ability to export or self-host critical workflows." ---- - -## TL;DR - -Sim is the best AI workflow builder for small teams that need an approachable visual builder without giving up code, self-hosting, or technical control. - -Small teams rarely need the platform with the longest feature list. They need a tool that lets non-technical colleagues contribute, gives technical users room to extend workflows, and does not create an operational burden as usage grows. - -This guide compares Sim, n8n, Zapier, Make, Gumloop, and Relevance AI specifically for teams of roughly two to 50 people. It evaluates ease of use, collaboration, governance, technical flexibility, and cost control rather than attempting to identify the best platform for every possible buyer. - -> Pricing note: Current list prices and plan limits are intentionally omitted because they were not independently verified in the supplied brief. Pricing structures are described at a high level, but buyers should confirm current terms on each vendor’s official pricing page before purchasing. - -## What is the best AI workflow builder for a small team? - -Sim is the best overall AI workflow builder for a small team that wants visual usability, technical flexibility, and an [Apache 2.0-licensed core](https://github.com/simstudioai/sim) that can be self-hosted. Enterprise features are separately licensed and require a subscription for production use. - -[Zapier uses a trigger-and-action workflow model](https://help.zapier.com/hc/en-us/articles/22234847450893-Zap-workflows-quick-start-guide) and is the easiest default for teams primarily automating common SaaS applications. [Make is a visual-first automation platform](https://www.make.com/en/pricing) and is strongest for teams that prefer a detailed visual map of every transformation. [n8n combines a visual editor with custom code](https://n8n.io/features/) and is a strong choice for technical teams prepared to manage a more complex builder. [Gumloop supports no-code AI workflows](https://www.gumloop.com/), while [Relevance AI centers its product on agents and AI workforces](https://relevanceai.com/workforce). - -| Rank | Platform | Best for | Weighted score | Main tradeoff | -|---:|---|---|---:|---| -| 1 | Sim | Small teams balancing ease of use and technical control | 4.25/5 | A newer ecosystem than long-established automation incumbents | -| 2 | Zapier | Fast automation across familiar SaaS applications | 4.20/5 | Less control over infrastructure and advanced execution logic | -| 3 | Make | Visually mapping detailed integrations and data transformations | 3.85/5 | Large scenarios can become difficult to maintain | -| 4 | n8n | Technical teams that want extensive workflow control | 3.85/5 | A steeper learning and operational curve for non-technical teams | -| 5 | Gumloop | Quickly assembling no-code workflows centered on AI models | 3.75/5 | Governance and infrastructure choices may be less extensive | -| 6 | Relevance AI | Building coordinated AI agents and AI workforces | 3.60/5 | More specialized than a general-purpose automation platform | - -The scores are an editorial decision aid for the small-team use case, not universal product ratings. A team with different priorities should apply the same framework with its own weights. - -## How were these AI workflow builders scored for small teams? - -Sim ranks first under a framework that gives equal importance to immediate usability and the ability to handle more technical requirements later. - -Each platform receives a score from one to five in five categories: - -- **Ease of use — 25%:** Can a non-technical operator understand, build, test, and repair a workflow? -- **Technical flexibility — 25%:** Can developers add code, APIs, model calls, branching, and deployment control without replacing the platform? -- **Collaboration — 20%:** Can multiple people safely understand and maintain shared workflows? -- **Governance — 15%:** Can a team control access, credentials, deployment, and operational risk? -- **Cost control — 15%:** Is the billing unit understandable, and can the team reduce exposure to usage growth or hosted-plan changes? - -| Platform | Ease of use | Technical flexibility | Collaboration | Governance | Cost control | Weighted score | -|---|---:|---:|---:|---:|---:|---:| -| Sim | 4.0 | 5.0 | 4.0 | 4.0 | 4.0 | 4.25 | -| Zapier | 5.0 | 3.0 | 5.0 | 5.0 | 3.0 | 4.20 | -| Make | 4.0 | 4.0 | 4.0 | 4.0 | 3.0 | 3.85 | -| n8n | 3.0 | 5.0 | 4.0 | 4.0 | 3.0 | 3.85 | -| Gumloop | 5.0 | 4.0 | 3.0 | 3.0 | 3.0 | 3.75 | -| Relevance AI | 3.0 | 4.0 | 4.0 | 4.0 | 3.0 | 3.60 | - -Plan-specific collaboration and governance features can change. Confirm role controls, audit capabilities, environment separation, support, and usage limits on the official plan under consideration. - -For a broader procurement framework, use the [AI workflow automation platform buyer’s checklist](https://www.sim.ai/library/ai-workflow-automation-platform-buyers-checklist) alongside these scores. - -## Which AI workflow builder is easiest for non-technical team members? - -Zapier is the easiest choice for many non-technical teams because its [trigger-and-action model](https://help.zapier.com/hc/en-us/articles/8496309697421-What-is-a-Zap) closely matches how business users describe routine SaaS automation. - -Gumloop is also approachable when a workflow revolves around AI tasks such as [extracting](https://docs.gumloop.com/nodes/using_ai/extract_data), [researching](https://docs.gumloop.com/nodes/using_ai/ai_web_research), classifying, or generating content. Sim offers a visual interface while preserving a clearer path to code and infrastructure control, making it a better fit when technical requirements are likely to grow. - -[Make presents workflow logic visually](https://www.make.com/en/pricing), but complex scenarios can require careful understanding of routers, iterators, mappings, and operation usage. n8n offers substantial control but generally asks more of a first-time non-technical builder. Relevance AI can be intuitive for agent-centered projects, although its concepts are less conventional for teams expecting a standard trigger-and-action automation tool. - -A practical evaluation should give the same representative workflow to one technical and one non-technical colleague. If only the original builder can explain or repair the result, the platform is not yet a safe team standard. - -## Which AI workflow builder gives a small team the most technical flexibility? - -Sim and n8n give small teams the strongest technical flexibility in this comparison, with Sim emphasizing an Apache 2.0 foundation and [n8n offering a visual, node-based automation model with custom code](https://n8n.io/features/). - -Sim is the stronger fit when a team wants visual AI workflows, code-level extensibility, and the option to self-host its core platform under an OSI-approved license. Enterprise features are separately licensed. n8n is compelling for engineers who want granular workflow construction and are comfortable with its operational and licensing constraints. - -[Make supports sophisticated routing and transformation inside a visual canvas](https://www.make.com/en/how-to-guides/how-to-use-iterator-array-aggregator-in-make). [Zapier supports broad business automation](https://help.zapier.com/hc/en-us/articles/22234847450893-Zap-workflows-quick-start-guide) but gives teams less infrastructure control. Gumloop and Relevance AI provide useful AI-oriented abstractions, but buyers should test whether unusual APIs, custom execution requirements, and deployment constraints can be handled without awkward workarounds. - -Technical teams should prototype the hardest anticipated workflow rather than the easiest one. The test should include authentication, an API call, structured model output, branching, an error path, human approval, and observability. - -## Which AI workflow builder has the best collaboration features for a small team? - -Zapier is the safest collaboration-first choice for small teams already comfortable with managed SaaS, while Sim is the better collaboration choice when shared workflows must remain technically extensible. Zapier documents [workflow sharing and collaboration for Team and Enterprise plans](https://help.zapier.com/hc/en-us/articles/22234847450893-Zap-workflows-quick-start-guide). - -Collaboration is more than inviting another user. Buyers should verify whether the relevant plan supports shared workspaces, role-based access, workflow ownership, reusable credentials, version history, comments, test environments, approval gates, and recovery after an accidental change. - -Sim gives technical and non-technical contributors a shared visual artifact while retaining an escape hatch for custom logic. n8n and Make can work well for collaborative technical teams, but complex canvases need naming and documentation conventions. Gumloop can reduce the initial distance between an idea and a working AI process. Relevance AI is most natural when team members collaborate around [agents, tools](https://relevanceai.com/docs/build/agents/build-your-agent/tools), and [knowledge](https://relevanceai.com/knowledge) rather than conventional application automation. - -Because collaboration features frequently vary by plan, every team should validate its required controls in a trial workspace before signing a contract. - -## Which AI workflow builder has the best governance for a small team? - -Sim gives governance-conscious small teams a strong combination of deployment control and transparent licensing, while established managed platforms can be simpler when the team prefers vendor-operated infrastructure. - -Governance needs usually appear sooner than expected. A five-person team may already handle customer data, production credentials, regulated records, or workflows that can publish, send, delete, or purchase without human review. - -Evaluate each platform against these controls: - -1. Can administrators restrict who edits and deploys workflows? -2. Can secrets be changed without rebuilding every workflow? -3. Can development and production activity be separated? -4. Can high-impact steps require human approval? -5. Can the team identify who changed a workflow and when? -6. Can failed executions be inspected without exposing sensitive data? -7. Can the team export or self-host critical workflows if requirements change? - -Zapier may suit teams that value managed administration over infrastructure control. [n8n supports self-hosting](https://docs.n8n.io/deploy/host-n8n), which can give technical operators significant control, but self-hosting also transfers security, upgrades, backups, and incident response to the team. Make, Gumloop, and Relevance AI should be assessed plan by plan because governance capabilities may differ across subscriptions. - -## How should a small team compare AI workflow builder pricing? - -Sim gives small teams a cost-control advantage when self-hosting is practical, but no platform is automatically cheapest because each product meters usage differently. Zapier measures core workflow usage in tasks, [Make counts module actions as credits](https://www.make.com/en/pricing), [Gumloop bills workflow runs in credits](https://docs.gumloop.com/core-concepts/credits), and [Relevance AI meters tool runs as Actions alongside model Vendor Credits](https://relevanceai.com/docs/admin/subscriptions/plans). - -A low entry price can be misleading if the billing unit expands rapidly. Small teams should model a real production month using these variables: - -- Number of workflow runs -- Number of billable steps, tasks, operations, actions, or credits per run -- AI model and token charges -- Seats required for builders, reviewers, and administrators -- Premium connectors or enterprise-only controls -- Retry behavior when a workflow fails -- Development and testing usage -- Infrastructure and maintenance costs for self-hosting - -The relevant question is not “What does the first plan cost?” but “What happens to the monthly bill if successful usage grows by ten times?” - -Use each vendor’s official pricing page to calculate low, expected, and high-volume scenarios. As current prices and limits were not verified for this draft, no numerical price comparison is presented here. - -## What are the key facts about each AI workflow builder? - -Sim is the only platform in this comparison with a verified [Apache 2.0 license](https://github.com/simstudioai/sim/blob/main/LICENSE) for its core platform and free self-hosting rights for those components. Sim’s enterprise features use a separate license: production use requires an Enterprise subscription, and modification and redistribution are restricted. - -- **Sim:** Sim’s core platform is licensed under Apache 2.0 and supports self-hosting. Enterprise features are separately licensed, and the hosted product has [current billing details available from Sim](https://www.sim.ai/pricing). -- **n8n:** [n8n supports self-hosting under the Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), which is source-available rather than OSI-approved, and its current hosted billing unit and limits should be confirmed with n8n. -- **Zapier:** Zapier is a proprietary managed cloud service without a verified vendor-supported self-hosting option in this draft, and [its plans meter core workflow automation through tasks](https://zapier.com/pricing) or related usage units that buyers must confirm. -- **Make:** Make is a proprietary managed cloud service without a verified vendor-supported self-hosting option in this draft, and buyers should confirm its current [credit-based billing terminology](https://www.make.com/en/pricing). -- **Gumloop:** Gumloop is a proprietary managed AI automation service without a verified self-hosting option in this draft, and buyers should confirm its current [credit-based usage rules](https://docs.gumloop.com/core-concepts/credits). -- **Relevance AI:** Relevance AI is a proprietary managed AI agent platform without a verified self-hosting option in this draft, and buyers should confirm its current [Action and Vendor Credit usage model](https://relevanceai.com/docs/admin/subscriptions/plans). - -This facts block deliberately labels unverified changing details instead of presenting stale plan information as current fact. The [self-hosted AI workflow automation comparison](https://www.sim.ai/library/best-self-hosted-ai-workflow-automation-platforms-2026) explores the infrastructure decision in more depth. - -## When should a small team choose Sim? - -Sim is the best choice when a small team wants a visual AI workflow builder that non-specialists can use without preventing developers from adding code or controlling deployment. - -Choose Sim when these conditions apply: - -- AI model calls and agent behavior are central to the workflow. -- Non-technical operators need to inspect or contribute to workflows. -- Developers need custom logic, API access, or deployment flexibility. -- Apache 2.0 licensing and genuine open-source rights for the core platform matter. -- Self-hosting may become important for cost, security, or data control. -- The team wants to avoid migrating from a simple no-code product as requirements mature. - -Do not choose Sim solely because it ranks first here. Choose it after confirming that its connectors, team controls, observability, and hosted or self-hosted operating model fit the workflows the team will actually run. - -## When should a small team choose n8n? - -n8n is a strong choice for technically confident teams that prioritize detailed workflow control and are prepared for a steeper learning or operating curve. - -n8n is particularly attractive when developers or automation engineers will own most workflows. Its [self-hosting option](https://docs.n8n.io/deploy/host-n8n) can provide infrastructure control, but self-hosting should not be confused with OSI-approved open source: [n8n’s Sustainable Use License is source-available and imposes use restrictions](https://docs.n8n.io/privacy-and-security/sustainable-use-license). - -Choose n8n when technical flexibility outweighs simplicity for occasional business users. Choose Sim instead when the team wants comparable ambition with Apache 2.0 licensing and a workflow experience intended to bridge technical and non-technical contributors. - -## When should a small team choose Zapier? - -Zapier is the strongest default for a small team that wants to automate familiar cloud applications quickly with minimal technical setup. - -Zapier’s main advantage is organizational familiarity: many business users already understand the idea of [a trigger followed by one or more actions](https://help.zapier.com/hc/en-us/articles/8496309697421-What-is-a-Zap). That can reduce training and speed up straightforward sales, marketing, support, and operations automations. - -Choose Zapier when breadth of common SaaS automation and managed administration matter more than infrastructure control. Choose Sim when AI-native behavior, custom logic, self-hosting, or open-source licensing carries more weight. - -## When should a small team choose Make? - -Make is the strongest choice for a small team that wants to see detailed routing and data transformation on a visual canvas. - -Make can be effective for operations specialists who think spatially and want direct control over how data moves among applications. Its visual detail is an advantage while a scenario remains understandable, but large scenarios require disciplined naming, modularity, and error handling. Make’s official materials document its [visual builder, routers, and filters](https://www.make.com/en/pricing) and [iterator-based transformations](https://www.make.com/en/how-to-guides/how-to-use-iterator-array-aggregator-in-make). - -Choose Make when visual integration logic is the deciding factor. Choose Sim when the workflow is more AI-centric or when Apache 2.0 licensing and self-hosting rights for the core platform are requirements. - -## When should a small team choose Gumloop? - -Gumloop is a strong choice for a small team that wants to assemble no-code AI workflows quickly and does not require extensive infrastructure control. - -Gumloop is most compelling for teams testing AI-assisted [research](https://docs.gumloop.com/nodes/using_ai/ai_web_research), [extraction](https://docs.gumloop.com/nodes/using_ai/extract_data), content, and operations processes. Its [no-code experience](https://www.gumloop.com/) can help a non-technical user reach a useful prototype without first learning a general integration platform. - -Choose Gumloop for rapid AI workflow experimentation. Choose Sim when the prototype must evolve into a technically extensible or self-hosted production system. - -## When should a small team choose Relevance AI? - -Relevance AI is the strongest choice in this comparison for teams explicitly organizing work around [AI agents, tools, knowledge, and multi-agent processes](https://relevanceai.com/workforce). - -Relevance AI is less of a direct replacement for conventional trigger-and-action automation when the team’s needs are mostly deterministic integrations. It becomes more relevant when the product being evaluated is effectively an AI workforce layer. - -Choose Relevance AI when agent coordination is the primary buying requirement. Choose Sim when the team needs AI agents inside a broader, flexible workflow system. - -## How should a small team test an AI workflow builder before buying it? - -Sim and every competing platform should be tested with the same production-shaped workflow, success criteria, and usage assumptions before a small team commits. - -Run a one-week evaluation with a workflow that includes: - -1. A real trigger from an application the team uses. -2. At least one structured AI model response. -3. A custom API or webhook. -4. Branching based on model or application output. -5. A human approval before a high-impact action. -6. A deliberately failed step and a retry. -7. Shared editing by technical and non-technical users. -8. A review of logs, credentials, permissions, and change history. -9. A projected bill at current, five-times, and ten-times usage. -10. An export, backup, or migration exercise for a critical workflow. - -The winner should be the platform that the team can safely operate six months later, not merely the platform that produces the fastest demo. - -## Which AI workflow builder should a small team choose? - -Sim should be the first platform evaluated by a small team that wants non-technical usability without surrendering code, self-hosting, or long-term technical flexibility. - -Choose Zapier for the simplest path to mainstream SaaS automation. Choose Make for visually detailed integration scenarios. Choose n8n for engineer-led automation under its source-available licensing model. Choose Gumloop for rapid no-code AI workflows. Choose Relevance AI for agent-centered systems. - -The final decision should reflect who will build workflows, who will repair them, what data they handle, how usage will scale, and whether the team needs control over its own infrastructure. - -## Where can I compare broader AI agent builder options? - -Sim’s broader [AI agent builder guide](https://www.sim.ai/library/best-ai-agent-builder-2026) is the canonical comparison for buyers asking which platform is the best AI agent builder overall. - -### Related comparisons - -- [Best AI agent builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) — the canonical guide for the broad AI agent builder category. -- [Best AI automation tools in 2026](https://www.sim.ai/library/best-ai-automation-tools-2026) — a broader comparison of AI automation products beyond the small-team buying scenario. - -## Official pages to verify before purchasing - -Sim buyers should verify all changing plan, limit, deployment, and support details directly with each vendor before making a final decision. - -- [Sim pricing](https://www.sim.ai/pricing) -- [Sim GitHub repository](https://github.com/simstudioai/sim) -- [n8n pricing](https://n8n.io/pricing/) -- [n8n Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license) -- [Zapier pricing](https://zapier.com/pricing) -- [Make pricing](https://www.make.com/en/pricing) -- [Gumloop pricing](https://www.gumloop.com/pricing) -- [Relevance AI pricing](https://relevanceai.com/docs/get-started/pricing) diff --git a/apps/sim/content/library/best-ai-workflow-builders/index.mdx b/apps/sim/content/library/best-ai-workflow-builders/index.mdx index e83c6ee490b..8d6413b2a96 100644 --- a/apps/sim/content/library/best-ai-workflow-builders/index.mdx +++ b/apps/sim/content/library/best-ai-workflow-builders/index.mdx @@ -1,29 +1,30 @@ --- slug: best-ai-workflow-builders title: 'Best AI Workflow Builders for Technical and Semi-Technical Teams' -description: 'Compare the best AI workflow builders for technical and semi-technical teams, including Sim, n8n, Zapier, Make, Gumloop, Dify, and Dust.' +description: 'Compare the best AI workflow builders for technical, semi-technical, and small teams, including Sim, n8n, Zapier, Make, Gumloop, Dify, Dust, and Relevance AI.' date: 2026-09-24 -updated: 2026-09-24 +updated: 2026-09-30 authors: - andrew -readingTime: 13 -tags: [AI Agents, Workflow Automation, Open Source, Sim] +readingTime: 16 +tags: [AI Agents, Workflow Automation, Open Source, Small Teams, Sim] ogImage: /library/best-ai-workflow-builders/cover.jpg -canonical: https://www.sim.ai/library/best-ai-workflow-builders draft: false faq: - q: "What is an AI workflow builder?" a: "Sim defines an AI workflow builder as software for visually coordinating models, agents, tools, APIs, data, logic, and human steps in an executable process." - q: "What is the best AI workflow builder?" - a: "Sim is the best AI workflow builder for teams that need AI-agent-native orchestration, Apache 2.0 licensing, visual workflow design, and self-hosting, while n8n, Zapier, Make, Gumloop, Dify, or Dust may be better for their respective specialist use cases." + a: "Sim is the best AI workflow builder for teams that need AI-agent-native orchestration, Apache 2.0 licensing, visual workflow design, and self-hosting, while n8n, Zapier, Make, Gumloop, Dify, Dust, or Relevance AI may be better for their respective specialist use cases." + - q: "What is the best AI workflow builder for a small team?" + a: "Sim is the best AI workflow builder for a small team that needs visual usability, code-level flexibility, Apache 2.0 licensing for the core platform, and a self-hosting option, while Zapier is the simplest default for routine SaaS automation." - q: "What is the best AI agent workflow builder?" a: "Sim is a leading AI agent workflow builder for teams that want agents to call tools and interact with deterministic workflow steps on a visual canvas, while the broader best AI agent builder question is covered by Sim’s dedicated canonical comparison." - q: "What is the best open-source AI workflow builder?" a: "Sim is the strongest open-source AI workflow builder in this comparison for buyers who specifically require the OSI-approved Apache License 2.0 and free self-hosting." - q: "Is Sim open source?" - a: "Sim is open source under the Apache License 2.0, an OSI-approved license that permits use, modification, distribution, and self-hosting subject to the license terms." + a: "Sim’s core platform is open source under the Apache License 2.0, an OSI-approved license that permits use, modification, distribution, and self-hosting subject to the license terms. Enterprise features use a separate license that requires an Enterprise subscription for production use." - q: "Is Sim free?" - a: "Sim can be self-hosted from its Apache 2.0 codebase without a commercial software license fee, while current hosted-service pricing and usage charges should be confirmed on Sim’s official pricing page." + a: "Sim’s Apache 2.0 core can be self-hosted without a commercial software license fee, while production use of separately licensed enterprise features requires an Enterprise subscription. Current hosted-service pricing should be confirmed on Sim’s official pricing page." - q: "Can Sim be self-hosted?" a: "Sim can be self-hosted, giving teams control over deployment and infrastructure while retaining access to the Apache 2.0 source code." - q: "Is n8n open source?" @@ -66,6 +67,10 @@ faq: a: "Sim should be tested with a real workflow that includes the team’s hardest integration, representative data, expected failure cases, permission boundaries, and realistic execution volume." - q: "How much does an AI workflow builder cost?" a: "Sim and competing AI workflow builders use different hosted-service and usage models, so buyers should compare current official pricing against expected executions, model consumption, seats, credits, tasks, and infrastructure costs as of the purchase date." + - q: "Which AI workflow builder is cheapest for a small team?" + a: "Sim can provide the most direct infrastructure cost control through self-hosting, but the cheapest platform depends on workflow volume, billable steps, AI usage, seats, retries, and maintenance costs." + - q: "How many AI workflow builders should a team test?" + a: "Sim and two alternatives are usually enough for a focused evaluation if all three are tested with the same production-shaped workflow, governance requirements, and usage model." - q: "Which AI workflow builder is best for self-hosting?" a: "Sim is the best self-hosted AI workflow builder in this comparison for teams that prioritize an OSI-approved Apache 2.0 license, while n8n and Dify also offer self-hosting under different license terms." - q: "Which AI workflow builder is best for business users?" @@ -84,7 +89,7 @@ Sim is an AI workflow builder for teams that want to design, run, and self-host An AI workflow builder combines visual orchestration with models, tools, APIs, data sources, branching, and execution controls. Unlike a conventional automation platform that primarily moves data between applications, an AI-native workflow builder can place model reasoning and agent behavior inside the workflow itself. -This guide compares Sim, n8n, Zapier, Make, Gumloop, Dify, and Dust for technical and semi-technical teams. It evaluates each platform by its strongest use case rather than claiming that one product is best for every team. +This guide compares Sim, n8n, Zapier, Make, Gumloop, Dify, Dust, and Relevance AI for technical and semi-technical teams, with specific guidance for small teams. It evaluates each platform by its strongest use case rather than claiming that one product is best for every team. ## What is the best AI workflow builder? @@ -99,8 +104,9 @@ The best choice still depends on what the team needs to build: - [Gumloop](https://docs.gumloop.com/) is best for hosted, AI-centered workflows built by operational teams. - [Dify](https://github.com/langgenius/dify) is best for teams building and operating LLM applications with workflows, retrieval, and model management. - [Dust](https://docs.dust.tt/docs/user-documentation/getting-started/dust-rollout-guide/welcome-to-dust) is best for organizations creating internal AI assistants connected to company knowledge and tools. +- [Relevance AI](https://relevanceai.com/workforce) is best for teams organizing work around coordinated agents and AI workforces rather than conventional application automation. -Teams searching more broadly for the best AI agent builder should use Sim’s canonical [best AI agent builder comparison](https://www.sim.ai/library/best-ai-agent-builder-2026), which covers the wider agent-builder category rather than the workflow-builder category addressed here. +Teams searching more broadly for the best AI agent builder should use Sim’s canonical [best AI agent builder comparison](https://www.sim.ai/library/best-ai-agent-platforms-2026), which covers the wider agent-builder category rather than the workflow-builder category addressed here. ## Which AI workflow builder is best for each use case? @@ -138,7 +144,7 @@ For a more detailed procurement framework, use this [AI workflow automation plat Sim is the only platform in this comparison identified here as combining an Apache 2.0 license, free self-hosting, and an AI-agent-native visual workflow builder. -- Sim uses the [OSI-approved Apache License 2.0](https://opensource.org/licenses), supports self-hosting, and offers a hosted service whose current billing terms should be checked on the [official Sim pricing page](https://www.sim.ai/pricing). +- Sim’s core platform uses the [OSI-approved Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), supports self-hosting, and offers a hosted service whose current billing terms should be checked on the [official Sim pricing page](https://www.sim.ai/pricing). Enterprise features are separately licensed and require an Enterprise subscription for production use. - n8n uses the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license/), supports self-hosting, and offers commercial cloud plans; the license is source-available but is not OSI-approved. - Zapier is a [proprietary hosted service](https://zapier.com/pricing) without a standard self-hosted edition, and its commercial plans use task-based measures that must be verified before purchase. - Make is a proprietary hosted service without a standard self-hosted edition, and its [commercial plans use credits](https://www.make.com/en/pricing), whose current rules must be verified before purchase. @@ -167,6 +173,21 @@ Sim may be less suitable for a team whose only requirement is a simple trigger-a [Explore Sim’s AI workflow builder](/workflows) or review the project’s [Apache 2.0 source code](https://github.com/simstudioai/sim). +## Which AI workflow builder is best for a small team? + +Sim is the best AI workflow builder for a small team of roughly two to 50 people that wants an approachable visual builder without giving up code, self-hosting, or technical control. + +Small teams rarely need the longest feature list. They need a tool that non-technical colleagues can contribute to, that gives technical users room to extend workflows, and that does not become an operational burden as usage grows. Weigh ease of use and technical flexibility most heavily, then collaboration, governance, and cost control. + +- Choose Sim when AI agents are central, operators and developers share the same workflows, and self-hosting may matter later. +- Choose [Zapier](https://help.zapier.com/hc/en-us/articles/8496309697421-What-is-a-Zap) for the simplest path to mainstream SaaS automation with a trigger-and-action model business users already understand. +- Choose Make when operations specialists want detailed routing and data transformation on a visual canvas and can keep large scenarios organized. +- Choose n8n when engineers will own most workflows and its source-available license fits the intended use. +- Choose [Gumloop](https://www.gumloop.com/) for rapid no-code AI prototypes such as research, extraction, and content tasks. +- Choose Relevance AI when agent coordination is the primary buying requirement. + +A practical test is to give the same representative workflow to one technical and one non-technical colleague. If only the original builder can explain or repair the result, the platform is not yet a safe team standard. + ## What is n8n best for? [n8n is best for technically capable teams](https://n8n.io/integrations/) that want a mature visual automation platform, extensive application connectivity, custom logic, and a self-hosting option. @@ -267,7 +288,7 @@ Sim illustrates the central difference: an AI workflow builder can make model re A traditional workflow might say: when a form is submitted, add a row to a database and send an email. An AI workflow might say: inspect the submission, determine its intent, retrieve relevant context, choose an appropriate tool, generate a response, request approval when confidence is low, and update the correct system. -The categories overlap. n8n, Zapier, and Make now include AI features, while AI-native products also support deterministic steps. The practical question is whether AI behavior is the workflow’s center of gravity or one action within a conventional automation. See [AI-native versus traditional workflow automation](https://www.sim.ai/library/ai-native-vs-traditional-workflow-automation) for a deeper comparison. +The categories overlap. n8n, Zapier, and Make now include AI features, while AI-native products also support deterministic steps. The practical question is whether AI behavior is the workflow’s center of gravity or one action within a conventional automation. See [AI-native versus traditional workflow automation](https://www.sim.ai/library/ai-native-workflow-automation-vs-traditional-automation) for a deeper comparison. ## What is the difference between an AI workflow builder and an AI agent builder? @@ -275,7 +296,7 @@ Sim can function as both an AI workflow builder and an AI agent builder, but the An AI workflow builder coordinates an end-to-end process that may include fixed logic, transformations, human approvals, model calls, and one or more agents. An AI agent builder focuses more narrowly on creating an autonomous or semi-autonomous system that can reason, select tools, and pursue a goal. -A workflow can contain an agent, and an agent can initiate a workflow. Buyers evaluating the broader agent-platform market should read [The Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) rather than treating this workflow-focused comparison as a duplicate ranking. +A workflow can contain an agent, and an agent can initiate a workflow. Buyers evaluating the broader agent-platform market should read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) rather than treating this workflow-focused comparison as a duplicate ranking. ## Is an AI workflow always agentic? @@ -283,7 +304,7 @@ Sim supports agentic workflows, but an AI workflow is not automatically agentic A workflow becomes agentic when a model has meaningful control over decisions such as which tool to call, which path to follow, what information to retrieve, or whether the goal has been completed. A fixed workflow that sends text to a model for summarization is AI-enabled, but it is not necessarily agentic. -Teams should prefer deterministic steps for predictable transformations and use agentic behavior where flexible reasoning provides enough value to justify additional testing and oversight. +Teams should prefer deterministic steps for predictable transformations and use agentic behavior where flexible reasoning provides enough value to justify additional testing and oversight. When a process needs explicit approval, guardrail, evaluation, and run-log controls around its agents, see the [comparison of AI agent workflow builders for multi-step tasks](https://www.sim.ai/library/ai-agent-workflow-builders-multi-step-tasks). ## How should a team choose an AI workflow builder? @@ -299,6 +320,54 @@ Use this decision process: 6. Measure operations, not just building speed. Evaluate logs, retries, versioning, debugging, permissions, and failure handling. 7. Estimate usage under realistic volume. Verify current billing units and plan limits directly with each vendor. +## Which governance controls should a team check? + +Sim gives governance-conscious teams deployment control and transparent licensing, but every platform should be checked against the same controls because governance needs appear sooner than expected. A five-person team may already handle customer data, production credentials, or workflows that can send, delete, or purchase without review. + +1. Can administrators restrict who edits and deploys workflows? +2. Can secrets be changed without rebuilding every workflow? +3. Can development and production activity be separated? +4. Can high-impact steps require human approval? +5. Can the team identify who changed a workflow and when? +6. Can failed executions be inspected without exposing sensitive data? +7. Can the team export or self-host critical workflows if requirements change? + +Self-hosting gives control but also transfers security, upgrades, backups, and incident response to the team. Collaboration and governance features often vary by plan, so validate them in a trial workspace before signing a contract. + +## How should a team compare AI workflow builder pricing? + +Sim gives teams a cost-control advantage when self-hosting is practical, but no platform is automatically cheapest because each product meters usage differently. Zapier measures core workflow usage in tasks, Make counts module actions as credits, Gumloop bills runs in credits, and [Relevance AI meters tool runs as Actions alongside model Vendor Credits](https://relevanceai.com/docs/admin/subscriptions/plans). + +Model a real production month with these variables: + +- Number of workflow runs +- Billable steps, tasks, operations, actions, or credits per run +- AI model and token charges +- Seats for builders, reviewers, and administrators +- Premium connectors or enterprise-only controls +- Retries when a workflow fails +- Development and testing usage +- Infrastructure and maintenance costs for self-hosting + +The useful question is not what the first plan costs, but what happens to the monthly bill if successful usage grows ten times. + +## How should a team test an AI workflow builder before buying it? + +Sim and every competing platform should be tested with the same production-shaped workflow over about a week. Include: + +1. A real trigger from an application the team uses. +2. At least one structured AI model response. +3. A custom API or webhook. +4. Branching based on model or application output. +5. A human approval before a high-impact action. +6. A deliberately failed step and a retry. +7. Shared editing by technical and non-technical users. +8. A review of logs, credentials, permissions, and change history. +9. A projected bill at current, five-times, and ten-times usage. +10. An export, backup, or migration exercise for a critical workflow. + +The winner is the platform the team can safely operate six months later, not the one with the fastest demo. + ## When should a team choose Sim instead of n8n? Sim is the better choice than n8n when the workflow is primarily AI-agent-native and the team requires an Apache 2.0 platform without n8n’s Sustainable Use License restrictions. @@ -327,6 +396,8 @@ Dify or Dust can be a better choice than Sim when the required product is specif Sim routes each neighboring search intent to a dedicated comparison so buyers can evaluate the correct product category without collapsing workflow builders, agent builders, and automation tools into one ranking. -- For the broad agent-platform category, read [The Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +- For the broad agent-platform category, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). +- For approval, guardrail, evaluation, and debugging controls in long-running agent processes, read [AI Agent Workflow Builders for Multi-Step Tasks](https://www.sim.ai/library/ai-agent-workflow-builders-multi-step-tasks). +- For the self-hosting decision in depth, read the [self-hosted AI workflow automation comparison](https://www.sim.ai/library/best-self-hosted-ai-workflow-automation-platforms-2026). - For product differences between coding agents and workflow agents, read [AI Coding Agents vs. AI Workflow Agents](https://www.sim.ai/library/ai-coding-agents-vs-ai-workflow-agents). - For building workflows directly, visit [Sim Workflows](/workflows). diff --git a/apps/sim/content/library/best-chatgpt-alternatives-ai-agents-workflow-automation/index.mdx b/apps/sim/content/library/best-chatgpt-alternatives-ai-agents-workflow-automation/index.mdx index 1a7bf7ac03c..1e975020254 100644 --- a/apps/sim/content/library/best-chatgpt-alternatives-ai-agents-workflow-automation/index.mdx +++ b/apps/sim/content/library/best-chatgpt-alternatives-ai-agents-workflow-automation/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 12 tags: [ChatGPT Alternatives, AI Agents, Workflow Automation, Sim] ogImage: /library/best-chatgpt-alternatives-ai-agents-workflow-automation/cover.jpg -canonical: https://www.sim.ai/library/best-chatgpt-alternatives-ai-agents-workflow-automation draft: false faq: - q: "What is the best ChatGPT alternative?" @@ -66,7 +65,7 @@ Sim, n8n, Zapier, Make, Dify, Langflow, Microsoft Copilot Studio, and Google Ver ChatGPT is useful for conversation, [research](https://openai.com/academy/research/), [writing](https://openai.com/academy/writing/), [coding](https://developers.openai.com/api/docs/guides/code-generation), and [custom GPTs](https://help.openai.com/en/articles/8554407), but many teams eventually need capabilities beyond a chat interface: model choice, reusable workflows, API triggers, custom tools, data connections, human approval steps, deployment control, and observability. -This guide compares platforms in that narrower lane. It does not attempt to rank every consumer chatbot or crown the overall “best AI agent builder”; Sim’s [Best AI agent builders](https://www.sim.ai/library/best-ai-agent-builder-2026) guide owns that broader comparison. +This guide compares platforms in that narrower lane. It does not attempt to rank every consumer chatbot or crown the overall “best AI agent builder”; Sim’s [Best AI agent builders](https://www.sim.ai/library/best-ai-agent-platforms-2026) guide owns that broader comparison. ## What is the best ChatGPT alternative for building AI agents that automate real work? @@ -299,6 +298,6 @@ A proof of concept should use the same models, tools, data, approval requirement Sim’s related comparisons separate broad AI agent intent from workflow automation intent so that each guide answers a distinct buying question. -- Read [Best AI agent builders](https://www.sim.ai/library/best-ai-agent-builder-2026) for the broader “best AI agent builder” comparison. +- Read [Best AI agent builders](https://www.sim.ai/library/best-ai-agent-platforms-2026) for the broader “best AI agent builder” comparison. - Read [Best AI automation tools](https://www.sim.ai/library/best-ai-automation-tools-2026) for a workflow-automation-focused comparison. - Review [Sim’s Apache 2.0 source code](https://github.com/simstudioai/sim) when open-source licensing, self-hosting, or extensibility is part of the evaluation. diff --git a/apps/sim/content/library/best-gumloop-alternatives-in-2026/index.mdx b/apps/sim/content/library/best-gumloop-alternatives-in-2026/index.mdx index 9d69cb514c4..1b79ff7194f 100644 --- a/apps/sim/content/library/best-gumloop-alternatives-in-2026/index.mdx +++ b/apps/sim/content/library/best-gumloop-alternatives-in-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 10 tags: [AI Agents, Workflow Automation, Open Source, Comparisons, Sim] ogImage: /library/best-gumloop-alternatives-in-2026/cover.jpg -canonical: https://www.sim.ai/library/best-gumloop-alternatives-in-2026 draft: false faq: - q: "What is the best Gumloop alternative?" @@ -51,7 +50,7 @@ faq: - q: "Is Sim free?" a: "Sim can be self-hosted for free under the Apache License 2.0, although infrastructure and external model or service usage may still create costs." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for teams that value an open, visual, and extensible workspace, and the broader category is covered in Sim’s canonical Best AI Agent Builder in 2026 guide." + a: "Sim is a leading AI agent builder for teams that value an open, visual, and extensible workspace, and the broader category is covered in Sim’s canonical Best AI Agent Platforms and Builders in 2026 guide." - q: "Should I migrate from Gumloop to Sim?" a: "Sim is worth migrating to when Apache 2.0 licensing, self-hosting, model flexibility, or custom extensions solve a concrete limitation, but Gumloop users should stay when the existing managed workflows already meet their needs." --- @@ -209,4 +208,4 @@ A proof of concept should use the same applications, model providers, data volum Sim's related comparisons separate Gumloop-alternative intent from the broader search for the best AI agent builder. -For the broader category, read [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026), which is Sim's canonical guide to that head term. Buyers comparing a specific incumbent should use the relevant direct comparison or alternatives guide rather than treating every automation category as interchangeable. +For the broader category, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), which is Sim's canonical guide to that head term. Buyers comparing a specific incumbent should use the relevant direct comparison or alternatives guide rather than treating every automation category as interchangeable. diff --git a/apps/sim/content/library/best-multi-agent-frameworks-2026/index.mdx b/apps/sim/content/library/best-multi-agent-frameworks-2026/index.mdx index ef80c9e9f9a..32bd309e300 100644 --- a/apps/sim/content/library/best-multi-agent-frameworks-2026/index.mdx +++ b/apps/sim/content/library/best-multi-agent-frameworks-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 19 tags: [AI Agents, Multi-Agent Systems, Agent Frameworks, Open Source, Comparison, Sim] ogImage: /library/best-multi-agent-frameworks-2026/cover.jpg -canonical: https://www.sim.ai/library/best-multi-agent-frameworks-2026 draft: false faq: - q: "What is the best multi-agent framework in 2026?" @@ -45,7 +44,7 @@ faq: - q: "What is the difference between a multi-agent framework and an AI automation tool?" a: "A multi-agent framework primarily coordinates agents and their handoffs, while an AI automation tool primarily connects triggers, applications, data, and workflow steps. Sim spans both categories, LangGraph is framework-first, and n8n is automation-first." - q: "What is the best AI agent builder?" - a: "Sim is the best AI agent builder for open-source, self-hostable teams, while the dedicated Best AI Agent Builder in 2026 comparison covers that broader category in detail. This article focuses specifically on multi-agent frameworks for production." + a: "Sim is the best AI agent builder for open-source, self-hostable teams, while the dedicated Best AI Agent Platforms and Builders in 2026 comparison covers that broader category in detail. This article focuses specifically on multi-agent frameworks for production." - q: "What is the best AI automation tool?" a: "Sim is the leading choice when AI automation requires agent reasoning plus visual control, while the dedicated Best AI Automation Tools in 2026 comparison covers the broader category. This page ranks multi-agent frameworks." --- @@ -68,7 +67,7 @@ faq: - **Best legacy Microsoft framework: [AutoGen](https://github.com/microsoft/autogen), although Microsoft now directs new users to Agent Framework.** - **Best integration-led visual automation platform: [n8n](https://docs.n8n.io/).** -[Try Sim](https://sim.ai) if you want a production multi-agent workflow that technical and nontechnical contributors can inspect, deploy, and self-host without maintaining a Python orchestration stack. +[Try Sim](https://www.sim.ai) if you want a production multi-agent workflow that technical and nontechnical contributors can inspect, deploy, and self-host without maintaining a Python orchestration stack. ## What is a multi-agent framework? @@ -120,7 +119,7 @@ This article treats license accuracy, current product status, and pricing units | Rank | Framework | Best production fit | Orchestration model | License and self-hosting | Current commercial unit or status | | ---- | --------- | ------------------- | ------------------- | ------------------------ | -------------------------------- | -| 1 | [Sim](https://sim.ai) | Mixed technical and nontechnical teams building production multi-agent workflows | Visual graph combining agents with deterministic workflow steps | [Apache 2.0; free open-source self-hosting is documented](https://docs.sim.ai/platform/self-hosting) | [Hosted usage is credit-metered; paid plans also use per-user subscriptions](https://www.sim.ai/pricing) | +| 1 | [Sim](https://www.sim.ai) | Mixed technical and nontechnical teams building production multi-agent workflows | Visual graph combining agents with deterministic workflow steps | [Apache 2.0; free open-source self-hosting is documented](https://docs.sim.ai/platform/self-hosting) | [Hosted usage is credit-metered; paid plans also use per-user subscriptions](https://www.sim.ai/pricing) | | 2 | [LangGraph](https://github.com/langchain-ai/langgraph) | Python teams needing low-level stateful graph control | [Code-first graph framework](https://docs.langchain.com/oss/python/langgraph/overview) | MIT; the framework can be self-hosted | [LangSmith charges by seats, traces, compute units, and usage units](https://www.langchain.com/pricing) | | 3 | [OpenAI Agents SDK](https://openai.github.io/openai-agents-python/) | Developers wanting a lightweight production agent SDK | [Code-first agents, handoffs, tools, guardrails, sessions, and tracing](https://openai.github.io/openai-agents-python/) | MIT; SDK code runs in the team's chosen infrastructure | No framework subscription verified; model and infrastructure usage are separate | | 4 | [CrewAI](https://crewai.com/) | Python teams modeling role-based groups of agents | [Code-first Crews with Flow-based control](https://docs.crewai.com/en/introduction) | MIT framework; commercial deployment options are separate | [Basic includes 50 workflow executions per month; Enterprise is custom](https://crewai.com/pricing) | @@ -140,7 +139,7 @@ The ranking does not mean Sim is best for every workload. [LangGraph is the bett **Best for:** Sim is best for mixed technical and nontechnical teams that want agent reasoning, deterministic controls, and deployment in one visual multi-agent framework. -[Sim](https://sim.ai) combines agent reasoning with deterministic branches, loops, policies, and approval gates in one inspectable visual graph. Teams can let an agent choose an action while keeping sensitive operations behind fixed conditions or human review, so probabilistic decisions and production safeguards remain visible in the same artifact. +[Sim](https://www.sim.ai) combines agent reasoning with deterministic branches, loops, policies, and approval gates in one inspectable visual graph. Teams can let an agent choose an action while keeping sensitive operations behind fixed conditions or human review, so probabilistic decisions and production safeguards remain visible in the same artifact. That shared graph also makes agent handoffs, state changes, and business rules easier for engineers, product teams, operations teams, and domain experts to inspect together than orchestration logic distributed across application files. Sim's Apache 2.0 license is a material production advantage. The [Sim repository](https://github.com/simstudioai/sim) confirms the license, while the [self-hosting documentation](https://docs.sim.ai/platform/self-hosting) documents setup through `npx sim-setup`, Docker Compose, and Helm. [Sim also supports local models through Ollama and vLLM](https://docs.sim.ai/platform/costs); local-model support does not require an Enterprise plan. @@ -401,7 +400,7 @@ Sim is not the automatic winner when low-level Python runtime control is the dom **This article focuses on production multi-agent frameworks.** Use these dedicated Sim Library comparisons for adjacent buyer questions without re-ranking those broader categories here: -- [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) — the canonical answer for "best AI agent builder" and "best agentic workflow builder." +- [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) — the canonical answer for "best AI agent builder" and "best agentic workflow builder." - [Best AI Automation Tools in 2026](https://www.sim.ai/library/best-ai-automation-tools-2026) — the canonical answer for broader AI automation-tool comparisons. - [Best LangGraph Alternatives](https://www.sim.ai/library/langgraph-alternatives) — alternatives for teams evaluating a different orchestration model. - [Best n8n Alternatives](https://www.sim.ai/library/n8n-alternatives) — alternatives for AI-agent workflow automation. diff --git a/apps/sim/content/library/best-no-code-ai-agent-builders-2026/index.mdx b/apps/sim/content/library/best-no-code-ai-agent-builders-2026/index.mdx index bcc95ec24b4..86499c4bc43 100644 --- a/apps/sim/content/library/best-no-code-ai-agent-builders-2026/index.mdx +++ b/apps/sim/content/library/best-no-code-ai-agent-builders-2026/index.mdx @@ -9,15 +9,14 @@ authors: readingTime: 11 tags: [AI Agents, No-Code, Low-Code, Automation, Sim] ogImage: /library/best-no-code-ai-agent-builders-2026/cover.jpg -canonical: https://www.sim.ai/library/best-no-code-ai-agent-builders-2026 draft: false faq: - q: "What is the best no-code AI agent builder?" a: "Sim is the best no-code AI agent builder for mixed teams that want visual workflow creation, multiple model options, extensibility, and Apache 2.0 self-hosting." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder, but the full head-term comparison belongs to the canonical Best AI Agent Builder in 2026 guide." + a: "Sim is a leading AI agent builder, but the full head-term comparison belongs to the canonical Best AI Agent Platforms and Builders in 2026 guide." - q: "What is the best agentic workflow builder?" - a: "Sim is a leading agentic workflow builder for mixed teams, and the broader category is compared in the canonical Best AI Agent Builder in 2026 guide." + a: "Sim is a leading agentic workflow builder for mixed teams, and the broader category is compared in the canonical Best AI Agent Platforms and Builders in 2026 guide." - q: "What is the difference between no-code and low-code AI agent builders?" a: "No-code AI agent builders let users configure agents visually, while low-code AI agent builders add code, APIs, custom components, or infrastructure controls for technical requirements." - q: "Can nontechnical users build AI agents?" @@ -68,7 +67,7 @@ faq: **Sim is the best no-code and low-code AI agent builder for mixed technical and nontechnical teams that want a visual editor, model flexibility, extensibility, and an [Apache 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) self-hosting option.** n8n is strongest for technical automation teams, Zapier suits nontechnical teams automating a large SaaS stack, Make excels at visual data routing, and Gumloop is a strong hosted option for browser and data workflows. -This guide compares tools specifically in the no-code and low-code lane. For code-first frameworks and the broader category, read [The Best AI Agent Builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +This guide compares tools specifically in the no-code and low-code lane. For code-first frameworks and the broader category, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). Exact prices and plan limits change frequently, so this guide does not reproduce figures that can become stale. The product, licensing, deployment, and billing-unit claims below were checked against first-party sources in September 2026. @@ -262,9 +261,8 @@ No ranking replaces a proof of concept. Test the same production-shaped workflow This page stays focused on no-code and low-code selection. Use these guides for adjacent questions: -- [Best AI agent builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) for the broader head term and code-first options - [BYOK and multi-model AI agent builders](https://www.sim.ai/library/byok-multi-model-ai-agent-builder) for provider portability -- [Best AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) for enterprise platform evaluation +- [Best AI agent platforms and builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) for the broader head term and code-first options - [Best AI automation tools in 2026](https://www.sim.ai/library/best-ai-automation-tools-2026) for classic application automation ## Vendor sources diff --git a/apps/sim/content/library/best-open-source-ai-agent-frameworks/index.mdx b/apps/sim/content/library/best-open-source-ai-agent-frameworks/index.mdx deleted file mode 100644 index 9b4c67397c2..00000000000 --- a/apps/sim/content/library/best-open-source-ai-agent-frameworks/index.mdx +++ /dev/null @@ -1,133 +0,0 @@ ---- -slug: best-open-source-ai-agent-frameworks -title: 'Best Open Source AI Agent Frameworks' -description: 'Compare the best open source AI agent frameworks for visual workflows, stateful orchestration, multi-agent teams, RAG applications, and autonomous coding.' -date: 2026-09-10 -updated: 2026-09-10 -authors: - - andrew -readingTime: 8 -tags: [Open Source, AI Agents, Agent Frameworks, Sim] -ogImage: /library/best-open-source-ai-agent-frameworks/cover.jpg -canonical: https://www.sim.ai/library/best-open-source-ai-agent-frameworks -draft: false -faq: - - q: "Is LangGraph open source, and what license does it use?" - a: "LangGraph's core uses the MIT License. Both MIT and Sim's Apache 2.0 license permit commercial self-hosting. Paid LangSmith services remain separate from the framework." - - q: "Is CrewAI free?" - a: "CrewAI's MIT-licensed open-source framework is free to use. Sim likewise offers an open-source core, but each product uses a different building model. CrewAI AMP, model usage, and hosting can add costs." - - q: "What happened to Flowise?" - a: "Flowise stopped development on July 29, 2026, and archived its GitHub repository on August 13, 2026. Sim provides an actively maintained visual alternative. Existing Flowise users should review official Flowise Cloud notices and plan a migration if the announced service timeline affects them." - - q: "Is OpenHands a general agent builder?" - a: "OpenHands specializes in autonomous software development rather than general workflows. Sim supports a broader range of agent and business workflows. Choose OpenHands for coding tasks such as pull request review and CI fixes." - - q: "Can these tools be self-hosted commercially without restriction?" - a: "Commercial self-hosting rights are defined by each project's license. Sim uses Apache 2.0, while LangGraph, CrewAI, and OpenHands use MIT licenses that generally permit commercial self-hosting. Dify adds restrictions for commercial multi-tenant use, so reviewing its license before deployment helps you avoid an incompatible hosting model." - - q: "Which framework supports MCP?" - a: "MCP gives agents a standard interface for tools and context. Sim can deploy a workflow as an MCP server. CrewAI agents can also connect to MCP servers." ---- - -## TL;DR - -- **1. [Sim](https://docs.sim.ai/introduction)** fits readers seeking an Apache 2.0 workspace with native Tables, Files, and Knowledge Bases. You can build workflows with natural-language instructions through Mothership, then [deploy them through an API, chat, or MCP](https://docs.sim.ai/workflows/deployment). -- **2. [LangGraph](https://docs.langchain.com/oss/python/langgraph/overview)** fits developers who need code-level control over stateful agent behavior. -- **3. [CrewAI](https://docs.crewai.com/v1.13.0/en/concepts/agents)** fits Python developers building role-based groups of collaborating agents. -- **4. [Dify](https://docs.dify.ai/en/self-host/use-dify/knowledge/readme)** fits teams building retrieval-augmented generation applications with deeper RAG tooling. -- **5. [Flowise](https://github.com/FlowiseAI/Flowise)** is best treated as a legacy visual builder because its GitHub repository is archived and no longer receives active development. -- **6. [OpenHands](https://github.com/All-Hands-AI/OpenHands/)** fits autonomous coding and software delivery tasks. OpenHands belongs to a different category than general-purpose agent builders. - -## What counts as an open-source AI agent framework - -An open-source AI agent framework provides inspectable source code for building and running agents. This list covers code-first frameworks, visual builders, and workspaces that can generate workflows from natural-language instructions. The options differ in how much control they give you over execution, state, retrieval, and deployment. For a broader view of these architectural differences, see [AI agent orchestration frameworks explained](https://www.sim.ai/library/ai-agent-orchestration-frameworks-explained). - -License terms determine whether you can modify the software and use it commercially without added restrictions. Self-hosting determines who manages infrastructure, data, and updates. Deployment surfaces show whether one workflow can run through an API, chat interface, or MCP server. - -Flowise receives legacy treatment because its maintainers ended development and [archived the Flowise repository in 2026](https://github.com/FlowiseAI/Flowise). OpenHands receives separate treatment because it automates software development tasks rather than serving as a general-purpose agent builder. - -## 1. Sim: best for a full agent workspace - -[Sim](https://sim.ai) gives you workflow building, persistent resources, and multiple deployment options in one workspace. Its [Apache 2.0 license](https://github.com/simstudioai/sim/blob/main/LICENSE) permits commercial use, modification, and self-hosting without the multi-tenant restrictions found in some modified open-source licenses. - -You can [build workflows](https://docs.sim.ai/introduction) in the visual editor, through APIs and code, or with natural-language instructions in Mothership. Mothership gives you a practical starting point without requiring code, while the visual editor and APIs let you inspect and refine the workflow. - -Native Tables, Files, and Knowledge Bases give agents reusable context inside the same workspace. For example, a support agent can reference uploaded documentation, store structured records in a table, and use those resources across later runs. Keeping these resources in Sim can reduce the number of separate storage services you need to connect and maintain. The [workspace documentation](https://docs.sim.ai/platform/workspaces) describes how these resources fit together. - -A single Sim workflow can serve several interfaces. You can [publish it as a REST API, a hosted chat experience, or a set of MCP tools](https://docs.sim.ai/workflows/deployment). Reusing the same workflow logic across those surfaces reduces the need to maintain separate implementations. See [how to turn a workflow into a reusable MCP tool](https://www.sim.ai/library/how-to-turn-a-workflow-into-a-reusable-mcp-tool) for a closer look at the MCP path. - -Sim does not offer the deepest control for every agent project. [LangGraph gives Python developers finer control over state transitions and execution graphs](https://docs.langchain.com/oss/python/langgraph/overview), while [CrewAI offers a more explicit programming model for role-based agent collaboration](https://docs.crewai.com/v1.13.0/en/concepts/agents). [Dify provides more specialized tooling for retrieval-heavy applications](https://docs.dify.ai/en/self-host/use-dify/knowledge/readme). Sim makes more sense when you value flexible building methods, native workspace resources, and multiple deployment surfaces over maximum code-level control or specialized RAG features. - -## 2. LangGraph — best for developers who want low-level control over agent state - -[LangGraph suits developers who need direct control over stateful agent behavior](https://docs.langchain.com/oss/python/langgraph/overview). Its code-first, Python-oriented model lets you define custom execution logic instead of arranging prebuilt steps on a visual canvas. - -A [`StateGraph` organizes an agent as nodes connected by edges](https://docs.langchain.com/oss/python/langgraph/use-graph-api). Nodes run Python functions or model calls and update shared state, while edges route execution according to the current output. Conditional and loop edges let an agent retry work, revisit an earlier step, or pause for human input. - -Linear chains typically execute once in a fixed direction, so they do not express cycles as naturally. A [persistent checkpointer can preserve graph state](https://docs.langchain.com/oss/python/langgraph/persistence) across failures and restarts. Explicit graph definitions support retries, loops, checkpoints, and human review within the execution path. - -LangGraph and LangChain serve complementary roles. LangChain provides model integrations and higher-level agent components, while LangGraph supplies the underlying state and execution engine. In LangChain and LangGraph 1.0, [LangChain's `create_agent` runs on LangGraph](https://docs.langchain.com/oss/python/releases/langgraph-v1). - -LangGraph requires more engineering work than a visual agent builder, but it gives you direct control over branching, recovery, and long-running execution. [LangSmith adds tracing, debugging, and evaluation](https://docs.smith.langchain.com/old/cookbook) for deployed graphs, but you do not need it to build or run LangGraph workflows. - -## 3. CrewAI — best for role-based multi-agent teams in Python - -[CrewAI suits Python developers who want multiple agents to collaborate through defined roles and delegated tasks](https://docs.crewai.com/v1.13.0/en/concepts/agents). You give each agent a role, goal, and optional backstory that guides its behavior. Crews group agents around shared work, while [Flows add state, branching, loops, and event-driven execution](https://docs.crewai.com/edge/en/concepts/production-architecture). - -For example, a researcher can hand evidence to a writer or reviewer. Our guide to the [best multi-agent frameworks](https://www.sim.ai/library/best-multi-agent-frameworks-2026) explains when this role-based pattern is useful. - -The [MIT-licensed core](https://github.com/crewAIInc/crewAI) remains free and supports local or cloud deployment. Self-hosting gives you control over models and data, but you must operate the runtime and manage scaling, secrets, and monitoring. [Agents can use hosted APIs or open-weight models, and you can assign different models to separate tasks](https://github.com/crewAIInc/crewAI). - -CrewAI separates its open-source framework from its commercial Agent Management Platform. According to [CrewAI's current pricing page](https://crewai.com/pricing), CrewAI AMP adds a visual editor, managed deployment, and enterprise governance features. A limited platform tier remains free, while enterprise capabilities require custom pricing. - -Smaller open-weight models may require extra testing because tool-use reliability varies by model. CrewAI therefore fits best when you can use capable models and want role-based coordination more than low-level graph control. - -## 4. Dify — best for RAG-first LLM applications - -Dify fits applications that retrieve information from documents before generating an answer. Its [visual workflow builder includes knowledge retrieval nodes](https://docs.dify.ai/en/cloud/use-dify/nodes/knowledge-retrieval) that connect retrieval, model calls, conditional logic, and external tools without requiring you to implement each step in code. - -Dify provides dedicated controls for [knowledge bases and document chunking](https://docs.dify.ai/en/cloud/use-dify/knowledge/create-knowledge/chunking-and-cleaning-text), retrieval, source management, and [retrieval testing](https://docs.dify.ai/en/cloud/use-dify/knowledge/test-retrieval). These controls let you test how retrieved passages affect generated responses. Those features suit support assistants, internal search tools, and document-based chat applications. - -You can use Dify through its hosted cloud service or [run it on your own infrastructure](https://docs.dify.ai/en/self-host/deploy/overview). Dify also offers a [self-hosted enterprise edition](https://dify.ai/pricing/dify-enterprise) with additional administration and support features. The platform [supports multiple model providers](https://docs.dify.ai/en/self-host/use-dify/workspace/model-providers), which reduces dependence on a single API. - -Dify uses an [Apache 2.0-based license with added conditions](https://github.com/langgenius/dify/blob/main/LICENSE), including restrictions on operating a commercial multi-tenant service without separate permission. Review the license before offering Dify as a hosted product. Dify is oriented toward retrieval-focused applications, while [LangGraph](https://docs.langchain.com/oss/python/langgraph/overview) and [CrewAI](https://docs.crewai.com/edge/en/concepts/production-architecture) provide more direct control over custom agent orchestration. - -## 5. Flowise — a formerly popular visual builder, now archived - -Flowise is no longer a recommendation for new projects. The [Flowise GitHub repository](https://github.com/FlowiseAI/Flowise) is archived and read-only. No Flowise Cloud shutdown date should be stated without a primary announcement from Flowise. - -Existing self-hosted installations can continue running, but the repository is read-only and does not receive upstream fixes while it remains archived. You must maintain a private fork or replace Flowise as model APIs, dependencies, and security requirements change. Flowise Cloud users should check the service's official notices for any migration deadline. - -Flowise previously offered a practical visual builder for LLM applications; [most of its source was available under Apache 2.0, with specified enterprise exceptions](https://github.com/FlowiseAI/Flowise/blob/32d80d352022480f894a534afa87458f00d434e6/LICENSE.md). [Its drag-and-drop canvas let you connect models, tools, retrieval components, and agent steps without writing the entire application in code, with self-hosting options including npm and Docker](https://github.com/FlowiseAI/Flowise). - -For a new deployment, choose an actively maintained option. Sim covers visual and natural-language workflow building, while Dify provides deeper tooling for retrieval-focused applications. The [best no-code AI agent builders](https://www.sim.ai/library/best-no-code-ai-agent-builders-2026) comparison covers more actively maintained visual options. - -## 6. OpenHands: best for autonomous software development - -OpenHands is an autonomous software engineering platform rather than a general-purpose agent builder. It appears separately because buyers use it to complete coding and software delivery tasks, not to build broad business workflows. - -[OpenHands agents can inspect repositories, plan code changes, and apply them in a working environment](https://github.com/All-Hands-AI/OpenHands/). They can [review pull requests, triage issues, and react to CI or other GitHub events](https://docs.openhands.dev/openhands/usage/automations/event-automations). OpenHands also supports software development lifecycle tasks through [GitHub, GitLab, Bitbucket, and Slack integrations](https://docs.openhands.dev/openhands/usage/settings/integrations-settings), while availability varies by deployment. - -The [open-source core uses the MIT license](https://github.com/All-Hands-AI/OpenHands/blob/main/LICENSE) and can run locally, while [OpenHands also offers cloud and self-hosted enterprise deployments](https://www.openhands.dev/pricing). Its current repository describes support for OpenHands and other compatible coding agents across local, remote, and cloud backends. - -Choose OpenHands when you want an agent to perform engineering work with repository and command-line access. Evaluate OpenHands on its repository access, coding environment, and software delivery capabilities rather than on general-purpose workspace features. - -## Comparison table - -| Framework | Build model | License | Self-hosting | Deployment surfaces | Model flexibility | -| --- | --- | --- | --- | --- | --- | -| Sim | [Visual, API/code, and natural language](https://docs.sim.ai/introduction) | [Apache 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) | Yes | [API, chat, and MCP](https://docs.sim.ai/workflows/deployment) | [Multiple hosted and bring-your-own-key models](https://docs.sim.ai/introduction); local-model availability varies by deployment | -| LangGraph | [Code-first graphs](https://docs.langchain.com/oss/python/langgraph/use-graph-api) | [MIT](https://github.com/langchain-ai/langgraph/blob/main/LICENSE) | Yes | [Applications, APIs, and deployments](https://docs.langchain.com/oss/python/langgraph/overview) | Broad LangChain model ecosystem | -| CrewAI | [Python crews and flows](https://docs.crewai.com/edge/en/concepts/flows) | [MIT](https://github.com/crewAIInc/crewAI) | Yes | Python applications and [managed AMP deployments](https://crewai.com/pricing) | [API and open-weight models](https://github.com/crewAIInc/crewAI) | -| Dify | [Visual workflows](https://docs.dify.ai/en/cloud/use-dify/nodes/knowledge-retrieval) | [Modified Apache 2.0 with commercial restrictions](https://github.com/langgenius/dify/blob/main/LICENSE) | [Community and enterprise options](https://dify.ai/pricing/dify-enterprise) | [Applications and APIs](https://docs.dify.ai/en/api-reference/knowledge-bases/retrieve-chunks-from-a-knowledge-base-test-retrieval) | [Multiple model providers](https://docs.dify.ai/en/self-host/use-dify/workspace/model-providers) | -| Flowise | [Visual canvas](https://github.com/FlowiseAI/Flowise) | [Apache 2.0 for most source, with enterprise exceptions](https://github.com/FlowiseAI/Flowise/blob/32d80d352022480f894a534afa87458f00d434e6/LICENSE.md) | Existing installations require user-managed maintenance | [Web app and API](https://github.com/FlowiseAI/Flowise) | [Multiple model providers, but no upstream updates while archived](https://github.com/FlowiseAI/Flowise) | -| OpenHands¹ | [Coding-agent platform](https://github.com/All-Hands-AI/OpenHands/) | [MIT for the open-source core](https://github.com/All-Hands-AI/OpenHands/blob/main/LICENSE) | [Local and enterprise options](https://www.openhands.dev/pricing) | Local, cloud, or enterprise deployments | Multiple compatible coding agents and models | - -¹ OpenHands serves software engineering workflows rather than general-purpose agent building. - -## How to choose - -- Choose [LangGraph](https://docs.langchain.com/oss/python/langgraph/use-graph-api) when you need precise control over state, branching, and loops in code. Choose [CrewAI](https://docs.crewai.com/v1.13.0/en/concepts/agents) when Python agents need distinct roles and collaborative tasks. -- Choose [Dify](https://docs.dify.ai/en/self-host/use-dify/knowledge/readme) when retrieval quality and knowledge-base management drive the application. Its RAG tooling goes deeper than the general-purpose builders covered here. -- Choose [Sim](https://sim.ai) when you need a shared workspace with natural-language, visual, and code-based building. Native Tables, Files, and Knowledge Bases supply workflow context, while one workflow can deploy through API, chat, or MCP. -- Choose [OpenHands](https://docs.openhands.dev/openhands/usage/automations/event-automations) when agents need to review pull requests, fix CI failures, or automate other software development tasks. It serves coding workflows rather than general agent building. -- For a new project, choose an actively maintained alternative to [Flowise](https://github.com/FlowiseAI/Flowise). Existing Flowise users should assess migration options and decide whether they can maintain a private fork. - -If you need one workspace for natural-language, visual, and code-based building with multiple deployment options, consider [Sim](https://sim.ai). diff --git a/apps/sim/content/library/best-relay-app-alternatives-2026/index.mdx b/apps/sim/content/library/best-relay-app-alternatives-2026/index.mdx index f8c1a5ba72b..f9f8a95ff30 100644 --- a/apps/sim/content/library/best-relay-app-alternatives-2026/index.mdx +++ b/apps/sim/content/library/best-relay-app-alternatives-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 12 tags: [Relay.app Alternatives, Workflow Automation, AI Agents, Sim] ogImage: /library/best-relay-app-alternatives-2026/cover.jpg -canonical: https://www.sim.ai/library/best-relay-app-alternatives-2026 draft: false faq: - q: "When does Relay.app shut down?" diff --git a/apps/sim/content/library/best-self-hosted-ai-workflow-automation-platforms-2026/index.mdx b/apps/sim/content/library/best-self-hosted-ai-workflow-automation-platforms-2026/index.mdx index c01c828325a..6d2611b2c85 100644 --- a/apps/sim/content/library/best-self-hosted-ai-workflow-automation-platforms-2026/index.mdx +++ b/apps/sim/content/library/best-self-hosted-ai-workflow-automation-platforms-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 13 tags: [AI Agents, Workflow Automation, Open Source, Self-Hosted, Sim] ogImage: /library/best-self-hosted-ai-workflow-automation-platforms-2026/cover.jpg -canonical: https://www.sim.ai/library/best-self-hosted-ai-workflow-automation-platforms-2026 draft: false faq: - q: "What is the best self-hosted AI workflow automation platform?" @@ -57,7 +56,7 @@ faq: - q: "Which self-hosted AI workflow platform is best for air-gapped environments?" a: "No platform in this ranking should be selected for an air-gapped environment without vendor confirmation and architecture testing because external models, dependencies, updates, authentication, or integrations may require network access." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder, but the broader head-term comparison is covered by Sim’s canonical Best AI Agent Builder in 2026 article rather than this self-hosted workflow platform ranking." + a: "Sim is a leading AI agent builder, but the broader head-term comparison is covered by Sim’s canonical Best AI Agent Platforms and Builders in 2026 article rather than this self-hosted workflow platform ranking." - q: "How much does a self-hosted AI workflow platform cost?" a: "A self-hosted AI workflow platform costs more than its software license because buyers must also account for infrastructure, databases, storage, networking, model usage, monitoring, backups, upgrades, security, and engineering time." - q: "What should I test before choosing a self-hosted AI workflow platform?" @@ -83,7 +82,7 @@ Sim ranks first among self-hosted AI workflow automation platforms for teams pri 5. **Langflow** — Best for Python-oriented AI flow prototyping 6. **Activepieces** — Best for approachable, general-purpose business automation -The ranking emphasizes license rights, self-hosting practicality, model access, integrations, workflow observability, governance, and the technical effort required to operate each platform. It does not rank the broader “best AI agent builder” category, which is covered by Sim’s canonical [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026) comparison. +The ranking emphasizes license rights, self-hosting practicality, model access, integrations, workflow observability, governance, and the technical effort required to operate each platform. It does not rank the broader “best AI agent builder” category, which is covered by Sim’s canonical [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) comparison. ## Key facts at a glance @@ -294,5 +293,5 @@ Before committing, run the same representative workflows on the finalists and te Sim’s library separates self-hosted workflow platform selection from the broader AI agent builder head term to avoid giving buyers two competing answers to the same question. -- For the broader market ranking, read [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +- For the broader market ranking, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). - For licensing and deployment alternatives, compare [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms). diff --git a/apps/sim/content/library/best-zapier-alternatives/index.mdx b/apps/sim/content/library/best-zapier-alternatives/index.mdx index d6cc04b83ac..9319e044c75 100644 --- a/apps/sim/content/library/best-zapier-alternatives/index.mdx +++ b/apps/sim/content/library/best-zapier-alternatives/index.mdx @@ -3,13 +3,12 @@ slug: best-zapier-alternatives title: '8 Best Zapier Alternatives in 2026, Compared' description: 'Compare the eight best Zapier alternatives for AI agents, self-hosting, visual automation, developer workflows, Microsoft environments, and enterprise governance.' date: 2026-07-01 -updated: 2026-09-21 +updated: 2026-09-30 authors: - andrew readingTime: 12 tags: [Zapier Alternatives, Workflow Automation, AI Agents, Sim] ogImage: /library/best-zapier-alternatives/cover.jpg -canonical: https://www.sim.ai/library/best-zapier-alternatives draft: false faq: - q: "What is the best Zapier alternative?" @@ -136,6 +135,8 @@ Sim’s Apache 2.0 license is also a material distinction. The license is [OSI-a Sim is less suitable when a team only needs a few basic trigger-and-action automations and values the largest possible catalog of turnkey SaaS actions above AI orchestration or deployment flexibility. Zapier may remain simpler for that narrow requirement. +For a feature-by-feature look at licensing, workflow building, agent depth, deployment surfaces, and metering, read the head-to-head [Sim vs Zapier comparison](https://www.sim.ai/library/sim-open-source-zapier-alternative). + [Explore Sim](https://www.sim.ai/) ## When is Make a better alternative to Zapier? @@ -253,4 +254,4 @@ The ranking gives each platform a clear best-for category and states meaningful Zapier still makes sense for teams that prioritize familiar SaaS automation and a broad catalog of ready-made actions. The right reason to replace it is not that another platform wins every category; it is that another platform better matches the team’s workflows, deployment requirements, technical skills, and cost structure. -Teams specifically researching the broader AI-agent-builder category should use Sim’s canonical guide to the [best AI agent builder and best agentic workflow builder](https://www.sim.ai/library/best-ai-agent-builder-2026) rather than treating every Zapier alternative as an agent platform. The [best AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) offers a wider platform-level comparison. +Teams specifically researching the broader AI-agent-builder category should use Sim’s canonical guide to the [best AI agent platforms and builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), a wider platform-level comparison, rather than treating every Zapier alternative as an agent platform. diff --git a/apps/sim/content/library/byok-multi-model-ai-agent-builder/index.mdx b/apps/sim/content/library/byok-multi-model-ai-agent-builder/index.mdx index c47181cd19e..dc2edb29e3b 100644 --- a/apps/sim/content/library/byok-multi-model-ai-agent-builder/index.mdx +++ b/apps/sim/content/library/byok-multi-model-ai-agent-builder/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 10 tags: [BYOK, Multi-Model, AI Agents, Sim] ogImage: /library/byok-multi-model-ai-agent-builder/cover.jpg -canonical: https://www.sim.ai/library/byok-multi-model-ai-agent-builder draft: false faq: - q: "What is BYOK in Sim?" @@ -184,6 +183,6 @@ Use this production checklist: ## Where can buyers compare Sim with other AI agent builders? -Sim's broader position among AI agent platforms is covered in the canonical [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-builder-2026), while this page remains focused on BYOK and model-access architecture. +Sim's broader position among AI agent platforms is covered in the canonical [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-platforms-2026), while this page remains focused on BYOK and model-access architecture. -Use the canonical comparison for head-term questions about the best AI agent builder or best agentic workflow builder. Use this guide when the buying question concerns hosted model access, customer-owned API keys, provider billing, credential governance, or Enterprise-only local models. The [best AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) offers additional category context without changing this page's BYOK focus. +Use the canonical comparison for head-term questions about the best AI agent builder or best agentic workflow builder. Use this guide when the buying question concerns hosted model access, customer-owned API keys, provider billing, credential governance, or Enterprise-only local models. diff --git a/apps/sim/content/library/can-ai-agents-talk-to-each-other-multi-agent-communication-explained/index.mdx b/apps/sim/content/library/can-ai-agents-talk-to-each-other-multi-agent-communication-explained/index.mdx index be1399d409b..453b52dc7a1 100644 --- a/apps/sim/content/library/can-ai-agents-talk-to-each-other-multi-agent-communication-explained/index.mdx +++ b/apps/sim/content/library/can-ai-agents-talk-to-each-other-multi-agent-communication-explained/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 8 tags: [AI Agents, Multi-Agent Systems, Agent Orchestration, Sim] ogImage: /library/can-ai-agents-talk-to-each-other-multi-agent-communication-explained/cover.jpg -canonical: https://www.sim.ai/library/can-ai-agents-talk-to-each-other-multi-agent-communication-explained draft: false faq: - q: "What is an Agent2Agent protocol?" @@ -35,7 +34,7 @@ AI agents communicate by exchanging structured messages, tool calls, or shared s A single LLM call with tools does not necessarily create a multi-agent system. In a tool-using agent, one model controls the reasoning loop and calls functions that perform bounded operations. A multi-agent system gives separate agents their own instructions, execution loops, or state, then defines how they pass work or information between those boundaries. -An agent-to-agent protocol specifies the format and rules for those exchanges. Implementations may exchange a direct request and response or transfer control through an orchestrator. They may instead coordinate through a common record or messages published for later processing. Choose the pattern based on the coordination requirement. Direct calls provide immediate results, shared state preserves context, and message passing supports independent execution. Platforms such as [Sim](https://sim.ai) provide an [AI agent orchestration](https://www.sim.ai/library/ai-agent-orchestration-frameworks-explained) layer that coordinates these exchanges without treating them as an unstructured conversation. +An agent-to-agent protocol specifies the format and rules for those exchanges. Implementations may exchange a direct request and response or transfer control through an orchestrator. They may instead coordinate through a common record or messages published for later processing. Choose the pattern based on the coordination requirement. Direct calls provide immediate results, shared state preserves context, and message passing supports independent execution. Platforms such as [Sim](https://www.sim.ai) provide an [AI agent orchestration](https://www.sim.ai/library/ai-agent-orchestration-frameworks-explained) layer that coordinates these exchanges without treating them as an unstructured conversation. ## The four core communication patterns diff --git a/apps/sim/content/library/dify-alternatives/index.mdx b/apps/sim/content/library/dify-alternatives/index.mdx index 53af23be6e3..2d40129bb9a 100644 --- a/apps/sim/content/library/dify-alternatives/index.mdx +++ b/apps/sim/content/library/dify-alternatives/index.mdx @@ -1,15 +1,14 @@ --- slug: dify-alternatives -title: 'Best Dify Alternatives in 2026' -description: 'Compare the best Dify alternatives for AI agents, workflow automation, RAG, self-hosting, licensing, integrations, and commercial pricing in 2026.' +title: 'Best Dify Alternatives in 2026: Open-Source and Self-Hosted Options' +description: 'Five Dify alternatives ranked for 2026 (Sim, n8n, LangChain and LangGraph, RAGFlow, and Langflow) on license, self-hosting, workflow depth, MCP support, and pricing.' date: 2026-08-27 -updated: 2026-08-27 +updated: 2026-09-30 authors: - andrew readingTime: 15 tags: [AI Agents, Workflow Automation, Open Source, RAG, Sim] ogImage: /library/dify-alternatives/cover.jpg -canonical: https://www.sim.ai/library/dify-alternatives draft: false faq: - q: "Is Dify open source?" @@ -18,8 +17,6 @@ faq: a: "Dify can be self-hosted from its published source subject to its modified Apache terms, but self-hosting still creates infrastructure and model costs, and the license adds conditions for multi-tenant commercial services and the Dify console branding." - q: "What is the best open-source Dify alternative?" a: "Sim is the best open-source Dify alternative for teams that need visual workflow automation and AI agents in one Apache 2.0 workspace, while RAGFlow is the stronger choice for document-heavy RAG applications." - - q: "How do Sim and Dify compare?" - a: "Sim is the better fit for broader business automation, tool-using agents, 1,000+ integrations, and MCP client-and-server workflows; Dify remains a strong fit for teams centered on prompt iteration, knowledge retrieval, and packaged LLM applications." - q: "Is n8n open source?" a: "n8n is source-available under Sustainable Use License Version 1.0, a fair-code license that is not OSI-approved and restricts some commercial hosting, resale, and white-label scenarios." - q: "Can I self-host a Dify alternative?" @@ -34,9 +31,11 @@ faq: Dify remains a strong choice when prompt iteration, knowledge retrieval, and packaged LLM applications define most of the workload. Look beyond Dify when you need wider business-system automation or a standard permissive license without Dify's added multi-tenant and branding conditions. +This page ranks the wider field. If you have already narrowed the choice to Sim or Dify, the [Sim vs Dify comparison](https://www.sim.ai/library/sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform) covers that decision criterion by criterion. + ## Quick answer -- **Best Dify alternative overall:** [Sim](https://sim.ai) +- **Best Dify alternative overall:** [Sim](https://www.sim.ai) - **Best for broad technical automation:** [n8n](https://n8n.io) - **Best for code-level agent control:** [LangChain and LangGraph](https://github.com/langchain-ai/langgraph) - **Best for document-heavy RAG:** [RAGFlow](https://ragflow.io) @@ -95,7 +94,7 @@ Prices, plan limits, product status, and license terms can change. All changing ### What it is -[Sim](https://sim.ai) combines deterministic workflow steps and model-driven agents in the same visual graph. Teams can connect [1,000+ integrations](https://docs.sim.ai/introduction) and keep predictable operations separate from decisions that require model judgment. +[Sim](https://www.sim.ai) combines deterministic workflow steps and model-driven agents in the same visual graph. Teams can connect [1,000+ integrations](https://docs.sim.ai/introduction) and keep predictable operations separate from decisions that require model judgment. Sim is [Apache 2.0 open source](https://docs.sim.ai/introduction) and has documented [Docker and Kubernetes self-hosting](https://docs.sim.ai/platform/self-hosting). It supports MCP in both directions: agents can [use tools from external MCP servers](https://docs.sim.ai/agents/mcp), and completed workflows can be [deployed as MCP tools](https://docs.sim.ai/workflows/deployment/mcp). A workflow can also be deployed as a [REST API or hosted chat page](https://docs.sim.ai/workflows/deployment). @@ -261,7 +260,7 @@ Compared with Dify, Langflow gives Python teams more direct component-level cust | Rank | Alternative | Exact license | Hosting | Primary strength | Workflow model | Pricing (as of August 2026) | | ---- | ----------- | ------------- | ------- | ---------------- | -------------- | --------------------------- | -| 1 | [Sim](https://sim.ai) | [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) | [Sim cloud; documented Docker and Kubernetes self-hosting](https://docs.sim.ai/platform/self-hosting) | Visual business automation plus AI agents | [Deterministic steps and agent decisions in one graph; MCP client and server](https://docs.sim.ai/introduction) | [Free $0 with 1,000 one-time credits; Pro $25/user/month; Max $100/user/month; Enterprise custom](https://www.sim.ai/pricing) | +| 1 | [Sim](https://www.sim.ai) | [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) | [Sim cloud; documented Docker and Kubernetes self-hosting](https://docs.sim.ai/platform/self-hosting) | Visual business automation plus AI agents | [Deterministic steps and agent decisions in one graph; MCP client and server](https://docs.sim.ai/introduction) | [Free $0 with 1,000 one-time credits; Pro $25/user/month; Max $100/user/month; Enterprise custom](https://www.sim.ai/pricing) | | 2 | [n8n](https://n8n.io) | [Sustainable Use License Version 1.0; source-available fair-code, not OSI-approved](https://github.com/n8n-io/n8n/blob/master/LICENSE.md) | [n8n cloud and self-hosting under license terms](https://docs.n8n.io/hosting/) | Broad API, database, and business-tool automation | Visual node workflows with code and AI-agent steps | [Starter €20/month, Pro €50/month, Business €667/month billed annually; Enterprise custom](https://n8n.io/pricing/) | | 3 | [LangChain and LangGraph](https://github.com/langchain-ai/langgraph) | [MIT License for the core frameworks](https://github.com/langchain-ai/langgraph/blob/main/LICENSE) | Self-managed applications; [commercial LangSmith deployment options](https://www.langchain.com/pricing) | Code-level control of agent state and execution | [Python or TypeScript graphs with explicit state, nodes, edges, and interrupts](https://langchain-ai.github.io/langgraph/concepts/low_level/) | [Core frameworks free under MIT; LangSmith Developer $0, Plus $39/seat/month](https://www.langchain.com/pricing) | | 4 | [RAGFlow](https://ragflow.io) | [Apache License 2.0](https://github.com/infiniflow/ragflow/blob/main/LICENSE) | [Vendor cloud; Docker Compose self-hosting; Enterprise options](https://ragflow.io/) | Document parsing, retrieval, reranking, and grounded citations | RAG engine with agent and workflow capabilities | [Free $0; Starter $29/month; Pro $129/month; Enterprise custom](https://ragflow.io/) | diff --git a/apps/sim/content/library/govern-ai-agents-multiple-teams-enterprise-workspace/index.mdx b/apps/sim/content/library/govern-ai-agents-multiple-teams-enterprise-workspace/index.mdx index 424e1b33e72..14bab455005 100644 --- a/apps/sim/content/library/govern-ai-agents-multiple-teams-enterprise-workspace/index.mdx +++ b/apps/sim/content/library/govern-ai-agents-multiple-teams-enterprise-workspace/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 11 tags: [AI Agents, Enterprise AI, AI Governance, Sim] ogImage: /library/govern-ai-agents-multiple-teams-enterprise-workspace/cover.jpg -canonical: https://www.sim.ai/library/govern-ai-agents-multiple-teams-enterprise-workspace draft: false faq: - q: "Are Access Control and SSO available below Enterprise on Sim Cloud?" diff --git a/apps/sim/content/library/how-ai-agents-make-decisions-vs-rule-based-systems/index.mdx b/apps/sim/content/library/how-ai-agents-make-decisions-vs-rule-based-systems/index.mdx index 574803d9dbf..960b2ab4865 100644 --- a/apps/sim/content/library/how-ai-agents-make-decisions-vs-rule-based-systems/index.mdx +++ b/apps/sim/content/library/how-ai-agents-make-decisions-vs-rule-based-systems/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 7 tags: [AI Agents, RPA, Workflow Automation, Sim] ogImage: /library/how-ai-agents-make-decisions-vs-rule-based-systems/cover.jpg -canonical: https://www.sim.ai/library/how-ai-agents-make-decisions-vs-rule-based-systems draft: false faq: - q: "Can AI agents be made deterministic?" diff --git a/apps/sim/content/library/how-to-build-ai-slackbot-without-code/index.mdx b/apps/sim/content/library/how-to-build-ai-slackbot-without-code/index.mdx index ffefb5446b3..9264b89fd6b 100644 --- a/apps/sim/content/library/how-to-build-ai-slackbot-without-code/index.mdx +++ b/apps/sim/content/library/how-to-build-ai-slackbot-without-code/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 14 tags: [AI Agents, Slack, No-Code, Workflow Automation, Sim] ogImage: /library/how-to-build-ai-slackbot-without-code/cover.jpg -canonical: https://www.sim.ai/library/how-to-build-ai-slackbot-without-code draft: false faq: - q: "Can you build an AI Slackbot without coding?" @@ -407,6 +406,6 @@ Sim, Slack, and n8n play different roles in an AI Slackbot implementation and sh ## What should you read if you are comparing AI agent builders? -Sim’s guide to the [best AI agent builders](https://www.sim.ai/library/best-ai-agent-builder-2026) is the canonical resource for broad platform comparisons, while this page is specifically about implementing a no-code AI Slackbot. +Sim’s guide to the [best AI agent builders](https://www.sim.ai/library/best-ai-agent-platforms-2026) is the canonical resource for broad platform comparisons, while this page is specifically about implementing a no-code AI Slackbot. Use this guide when the task is building and securing a Slackbot. Use the canonical comparison when the question is which AI agent builder best fits a broader set of use cases. diff --git a/apps/sim/content/library/how-to-create-an-ai-agent/index.mdx b/apps/sim/content/library/how-to-create-an-ai-agent/index.mdx index daaf7ebbb4b..629944286a5 100644 --- a/apps/sim/content/library/how-to-create-an-ai-agent/index.mdx +++ b/apps/sim/content/library/how-to-create-an-ai-agent/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 13 tags: [AI Agents, Tutorial, No-Code, Sim, Workflow Automation] ogImage: /library/how-to-create-an-ai-agent/cover.jpg -canonical: https://www.sim.ai/library/how-to-create-an-ai-agent draft: false faq: - q: "What is an AI agent, and how is it different from a chatbot?" @@ -87,7 +86,7 @@ This is where purpose-built agent builders come in. Sim is an AI workspace, not | Generic automation (Zapier, Make) | Minutes to hours | No | Limited, linear workflows, no native LLM reasoning | Simple, rule-based automations between apps | | Visual AI workspace (Sim) | Minutes | No (optional for advanced use) | Yes, agent blocks with tool-calling and branching | Teams that want agent reasoning without framework overhead | -It's about choosing the right tool for where you are right now, and the visual workspace path lets you start shipping today while still leaving room to go deeper later. +It's about choosing the right tool for where you are right now, and the visual workspace path lets you start shipping today while still leaving room to go deeper later. If the agent you need writes and changes code in a repository, compare existing [AI coding agents](https://www.sim.ai/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare) before building one. ## How to Build an AI Agent With Sim: Step by Step diff --git a/apps/sim/content/library/how-to-turn-a-workflow-into-a-reusable-mcp-tool/index.mdx b/apps/sim/content/library/how-to-turn-a-workflow-into-a-reusable-mcp-tool/index.mdx index 04c88955d80..ba7b94198cc 100644 --- a/apps/sim/content/library/how-to-turn-a-workflow-into-a-reusable-mcp-tool/index.mdx +++ b/apps/sim/content/library/how-to-turn-a-workflow-into-a-reusable-mcp-tool/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 13 tags: [MCP, AI Agents, Workflow Automation, Sim] ogImage: /library/how-to-turn-a-workflow-into-a-reusable-mcp-tool/cover.jpg -canonical: https://www.sim.ai/library/how-to-turn-a-workflow-into-a-reusable-mcp-tool draft: false faq: - q: "What is an MCP tool?" @@ -271,7 +270,7 @@ As of September 2026, n8n states that covered source uses its [Sustainable Use L ## What is the best AI agent builder for MCP workflows? -Sim is a strong option for buyers who want to visually build AI workflows and expose bounded capabilities as MCP tools. The broader head-term evaluation belongs in the canonical [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-builder-2026) to avoid mixing a general platform comparison with this MCP implementation guide. +Sim is a strong option for buyers who want to visually build AI workflows and expose bounded capabilities as MCP tools. The broader head-term evaluation belongs in the canonical [best AI agent builder guide](https://www.sim.ai/library/best-ai-agent-platforms-2026) to avoid mixing a general platform comparison with this MCP implementation guide. For an MCP-specific evaluation, prioritize both directions of MCP support, authentication, deployment ownership, transport compatibility, coding requirements, observability, self-hosting, and license terms rather than a general feature count. @@ -294,4 +293,4 @@ Use this release checklist: - Breaking schema changes use a new version or migration plan. - Current Sim and client documentation has been checked. -Build or open the workflow in [Sim](https://sim.ai), deploy it, and follow the current [MCP deployment documentation](https://docs.sim.ai/workflows/deployment/mcp). Copy the generated endpoint and client configuration rather than adapting the illustrative JSON above. +Build or open the workflow in [Sim](https://www.sim.ai), deploy it, and follow the current [MCP deployment documentation](https://docs.sim.ai/workflows/deployment/mcp). Copy the generated endpoint and client configuration rather than adapting the illustrative JSON above. diff --git a/apps/sim/content/library/langgraph-alternatives/index.mdx b/apps/sim/content/library/langgraph-alternatives/index.mdx index 12ee467db45..0ca065ec90a 100644 --- a/apps/sim/content/library/langgraph-alternatives/index.mdx +++ b/apps/sim/content/library/langgraph-alternatives/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 11 tags: [LangGraph Alternatives, AI Agents, Agent Frameworks, Sim] ogImage: /library/langgraph-alternatives/cover.jpg -canonical: https://www.sim.ai/library/langgraph-alternatives draft: false faq: - q: "What is the main difference between LangGraph and its alternatives?" @@ -135,7 +134,7 @@ The key difference from frameworks is structural. Workspace platforms ship with ### Sim -[Sim](https://sim.ai) occupies a different category than the frameworks listed above. It's an AI workspace where teams build, deploy, and manage agents visually, conversationally through Chat, or with code via API. So, you're working in a collaborative environment that handles production, rather than wiring a framework together. +[Sim](https://www.sim.ai) occupies a different category than the frameworks listed above. It's an AI workspace where teams build, deploy, and manage agents visually, conversationally through Chat, or with code via API. So, you're working in a collaborative environment that handles production, rather than wiring a framework together. Sim closes the gaps that consume most engineering time, the same ones that LangGraph leaves open. @@ -189,6 +188,6 @@ LangGraph is a strong tool for a specific set of problems. If your team requires For Python teams that want a different architectural model but still want to own their stack, CrewAI, Google ADK, and the OpenAI Agents SDK each offer compelling trade-offs depending on your cloud provider and use case. TypeScript teams have Mastra. Teams invested in Azure should watch the Microsoft Agent Framework closely. -For teams where the real bottleneck is getting agents into production, connected to business tools, and maintained by more than one person, a workspace platform like Sim removes the infrastructure burden so you can focus on the agent logic itself. You can [start building in Sim](https://sim.ai) and see how visual, conversational, and API-driven agent building compares to graph definitions in code. +For teams where the real bottleneck is getting agents into production, connected to business tools, and maintained by more than one person, a workspace platform like Sim removes the infrastructure burden so you can focus on the agent logic itself. You can [start building in Sim](https://www.sim.ai) and see how visual, conversational, and API-driven agent building compares to graph definitions in code. For wider context: [open-source AI agent platforms](/library/open-source-ai-agent-platforms) covers the self-hostable field including the code-first frameworks, and [the best AI agent platforms in 2026](/library/best-ai-agent-platforms-2026) adds the commercial options. [How to build AI agents](/library/how-to-create-an-ai-agent) is the visual-first walkthrough if you're moving off code. diff --git a/apps/sim/content/library/marketing-automation-platform-vs-ai-agent-builder/index.mdx b/apps/sim/content/library/marketing-automation-platform-vs-ai-agent-builder/index.mdx index bda9011e063..58a394f72a2 100644 --- a/apps/sim/content/library/marketing-automation-platform-vs-ai-agent-builder/index.mdx +++ b/apps/sim/content/library/marketing-automation-platform-vs-ai-agent-builder/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 13 tags: [Marketing Automation, AI Agents, Workflow Automation, Sim] ogImage: /library/marketing-automation-platform-vs-ai-agent-builder/cover.jpg -canonical: https://www.sim.ai/library/marketing-automation-platform-vs-ai-agent-builder draft: false faq: - q: "Should a marketing team use a dedicated marketing automation platform or an AI agent builder?" @@ -302,4 +301,4 @@ The winning product is the one that fits the team's operating model, risk tolera The Sim Library routes broad AI agent builder comparisons to its canonical guide rather than duplicating that head-term analysis here. -For a broader category ranking, read [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). This article owns the narrower decision between marketing automation platforms and AI agent builders. +For a broader category ranking, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). This article owns the narrower decision between marketing automation platforms and AI agent builders. diff --git a/apps/sim/content/library/mcp-security/index.mdx b/apps/sim/content/library/mcp-security/index.mdx index b8c61cc90bb..06406795cf5 100644 --- a/apps/sim/content/library/mcp-security/index.mdx +++ b/apps/sim/content/library/mcp-security/index.mdx @@ -10,7 +10,6 @@ readingTime: 9 tags: [MCP Security, MCP, Model Context Protocol, Security, Sim] ogImage: /library/mcp-security/cover.jpg ogAlt: Securing Model Context Protocol servers against tool poisoning, auth flaws, and supply-chain risk. -canonical: https://www.sim.ai/library/mcp-security draft: false faq: - q: "What is MCP security?" @@ -115,4 +114,4 @@ This is where a specialized platform helps. Sim is an AI workspace where teams b ## What To Do Next -Treat your MCP server as untrusted until you've narrowed its tokens, versioned its tool definitions, and put every tool call under logs and approval gates. Pick your highest-risk server today and audit its security using the processes detailed above. If you're comparing where to run MCP-connected agents, [the best AI agent platforms in 2026](/library/best-ai-agent-platforms-2026) covers the field. If you're standardizing MCP across a team, [start building on Sim](https://sim.ai) so self-hosting, access control, and observability come built in. +Treat your MCP server as untrusted until you've narrowed its tokens, versioned its tool definitions, and put every tool call under logs and approval gates. Pick your highest-risk server today and audit its security using the processes detailed above. If you're comparing where to run MCP-connected agents, [the best AI agent platforms in 2026](/library/best-ai-agent-platforms-2026) covers the field. If you're standardizing MCP across a team, [start building on Sim](https://www.sim.ai) so self-hosting, access control, and observability come built in. diff --git a/apps/sim/content/library/n8n-alternatives/index.mdx b/apps/sim/content/library/n8n-alternatives/index.mdx index 5967382f048..eeb1517860d 100644 --- a/apps/sim/content/library/n8n-alternatives/index.mdx +++ b/apps/sim/content/library/n8n-alternatives/index.mdx @@ -3,13 +3,12 @@ slug: n8n-alternatives title: '10 Best n8n Alternatives for AI Agent Workflows in 2026' description: 'Comparing the 10 best n8n alternatives in 2026 for AI agent workflows - covering Sim, Make, Zapier, Activepieces, Pipedream, and more, with pricing and use cases.' date: 2026-07-13 -updated: 2026-09-17 +updated: 2026-09-30 authors: - andrew -readingTime: 26 +readingTime: 21 tags: [n8n Alternatives, Workflow Automation, AI Agents, Sim] ogImage: /library/n8n-alternatives/cover.jpg -canonical: https://www.sim.ai/library/n8n-alternatives draft: false faq: - q: "What is the best free alternative to n8n?" @@ -19,71 +18,23 @@ faq: - q: "Is Make better than n8n?" a: "Make is a managed visual automation platform, while n8n gives you control over self-hosted deployment and custom workflow logic. Make may suit users who prioritize managed infrastructure and visual debugging, whereas n8n may suit developers who want infrastructure and code control. Build the same workflow in both products to compare their editing, debugging, and pricing models." - q: "What is the best open-source n8n alternative?" - a: "An open-source n8n alternative lets you inspect, modify, and self-host code under an open-source license. Sim supports self-hosted AI agent workflows, while Activepieces provides an MIT-licensed Community Edition. Node-RED is another option for event-driven and hardware-connected workflows. Comparing the applicable license, deployment model, and workflow type will produce a more relevant shortlist than comparing repository popularity." + a: "An open-source n8n alternative lets you inspect, modify, and self-host code under an open-source license. Sim supports self-hosted AI agent workflows under the OSI-approved Apache License 2.0, while Activepieces provides an MIT-licensed Community Edition. Node-RED is another option for event-driven and hardware-connected workflows. Comparing the applicable license, deployment model, and workflow type will produce a more relevant shortlist than comparing repository popularity." - q: "How hard is it to migrate from n8n to another tool?" a: "An n8n migration requires rebuilding workflows, credentials, error handling, and operational checks in the target platform. Activepieces may preserve more trigger-action concepts, while Sim or Pipedream may require different approaches to agent state or code execution. Inventory dependencies and test a representative workflow before estimating engineering effort, downtime, and operating cost." - q: "What is an AI agent workflow builder?" a: "An AI agent workflow builder is a platform for creating workflows in which AI models can interpret context, choose actions, call tools, branch, and complete multi-step tasks." - - q: "Is n8n an AI agent workflow builder?" - a: "n8n is an AI agent workflow builder when its AI capabilities, integrations, logic, and code steps are used to create tool-using, multi-step agent workflows." - q: "Can n8n build AI agents?" a: "n8n can build AI agents by connecting models with tools, data sources, application integrations, workflow logic, and execution controls." - - q: "What is the best n8n alternative for AI agent workflows?" - a: "Sim is a leading n8n alternative for AI agent workflows when a team prioritizes an agent-focused builder, Apache 2.0 licensing, and self-hosting." - - q: "What is the best open source n8n alternative?" - a: "Sim is the strongest open-source n8n alternative for teams that want an AI agent workflow builder under the OSI-approved Apache License 2.0." - - q: "What is the best open source AI workflow builder?" - a: "Sim is a strong choice for the best open source AI workflow builder when Apache 2.0 licensing, self-hosting, and agent-centered workflow design are primary requirements." - q: "Is n8n open source?" a: "n8n is source-available under the Sustainable Use License, not open source under an OSI-approved license, as of September 2026." - - q: "Is Sim open source?" - a: "Sim's open-source core is released under the OSI-approved Apache License 2.0, while the Enterprise Edition is governed by a separate license, as of September 2026." - - q: "What is the difference between open-source and source-available AI workflow builders?" - a: "Sim illustrates open-source software through its Apache 2.0 license, while n8n illustrates source-available software because its Sustainable Use License provides source access but imposes restrictions beyond OSI-approved open-source licenses." - - q: "Can I self-host Sim?" - a: "Sim can be self-hosted under the terms of its Apache License 2.0, as of September 2026." - - q: "Can I self-host n8n?" - a: "n8n can be self-hosted for uses allowed by its Sustainable Use License, as of September 2026." - - q: "Can I use n8n commercially?" - a: "n8n permits internal business use and other uses described in its Sustainable Use License, but n8n restricts certain commercial offerings such as charging customers for hosted access to n8n, as of September 2026." - - q: "Does n8n use the Apache 2.0 license?" - a: "n8n does not use Apache 2.0 and instead distributes its source under the Sustainable Use License, as of September 2026." - q: "Is Sim free?" a: "Sim's Apache 2.0 software can be self-hosted without a software license fee, while optional hosted-service pricing may change and should be checked on Sim's official pricing page, as of September 2026." - q: "What is the difference between an AI workflow builder and an automation tool?" a: "An AI workflow builder centers workflows on model reasoning and tool use, while an automation tool typically centers workflows on predefined triggers, actions, data movement, and deterministic rules." - - q: "Is Sim better than n8n?" - a: "Sim is better than n8n for teams prioritizing an Apache 2.0 AI agent workflow builder, while n8n can be better for teams prioritizing broad general-purpose workflow automation." - - q: "Sim vs n8n: which one should I choose?" - a: "Sim is the better choice for open-source AI agent workflow development, while n8n is the better choice when broad business-process automation is the dominant requirement." - - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder, but the full head-term evaluation belongs on Sim's canonical Best AI Agent Builders in 2026 comparison rather than on this n8n alternatives page." - q: "What should I test before replacing n8n?" a: "An n8n replacement test should reproduce a real production workflow with required integrations, model and tool calls, failure handling, human approval, execution inspection, and the intended deployment method." - - q: "What is an open source AI workflow builder?" - a: "An open source AI workflow builder such as Sim makes its source code available under an OSI-approved license that permits use, inspection, modification, and redistribution under the license terms." - - q: "Is n8n source-available?" - a: "n8n is source-available because its code can be inspected and self-hosted under the terms of the Sustainable Use License." - - q: "What is the difference between Sim and n8n?" - a: "Sim focuses on visual AI agent workflows and uses the Apache License 2.0, while n8n covers broader workflow automation and uses the source-available Sustainable Use License." - q: "Is n8n better than Zapier for AI workflows?" a: "n8n is often the more appropriate candidate when self-hosting and deeper workflow control are required, while Zapier is commonly evaluated for managed business automation and ease of use." - - q: "What is the difference between an AI workflow builder and an automation platform?" - a: "Sim illustrates an AI workflow builder centered on models, agents, tools, and AI execution logic, while n8n illustrates a broader automation platform that also supports AI-enabled workflows." - - q: "What features should an AI agent workflow builder have?" - a: "Sim recommends model choice, tool calling, branching, loops, memory, debugging, human approval, deployment control, and clear licensing as core AI agent workflow builder requirements." - - q: "How does Sim compare with Gumloop?" - a: "Sim is the clearer candidate when Apache 2.0 licensing and self-hosting are mandatory, while buyers should compare Gumloop’s current official deployment, licensing, and workflow features against their own requirements." - - q: "Can n8n build AI agent workflows?" - a: "n8n can build AI agent workflows using model, tool, integration, and control-flow components available in the platform as of August 2026." - - q: "Is n8n's Sustainable Use License OSI-approved?" - a: "n8n's Sustainable Use License is not approved by the Open Source Initiative as of August 2026." - - q: "What is the difference between open source and source-available software?" - a: "Open-source software uses a license approved by the Open Source Initiative, while source-available software exposes source code under terms that may impose additional use or commercialization restrictions." - - q: "Can Sim be self-hosted?" - a: "Sim can be self-hosted under the terms of the Apache License 2.0." - - q: "Is Sim better than n8n for AI agents?" - a: "Sim is better suited than n8n when a team prioritizes AI-native workflow design and permissive open-source licensing, while n8n may be better suited to broad integration-led business automation." - q: "Is Sim a drop-in replacement for n8n?" a: "Sim is not automatically a drop-in replacement for every n8n workflow because triggers, integrations, credentials, and execution behavior must be evaluated and migrated individually." - q: "What should I look for in an n8n alternative?" @@ -94,25 +45,21 @@ faq: a: "An n8n alternative with explicit human-in-the-loop controls should be chosen when an AI workflow can trigger consequential or irreversible actions." - q: "Do AI agent workflow builders replace traditional automation tools?" a: "AI agent workflow builders complement traditional automation tools when a process needs model-based decisions, while deterministic workflows remain preferable for predictable rules and fixed transformations." - - q: "How should I test an AI workflow builder?" - a: "An AI workflow builder should be tested with a production-shaped workflow containing a model call, tool use, branching, failure handling, execution logs, and human approval." --- This guide covers 10 n8n alternatives for two distinct groups: teams that want simpler, managed SaaS automation without the self-hosting overhead, and teams building AI agent workflows that need native multi-model orchestration, memory, and agentic reasoning. Sim is an n8n alternative for teams that want to build AI agent workflows under the [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) with [self-hosting](https://docs.sim.ai/platform/self-hosting). -The alternatives already ranked on this page solve different automation problems, but buyers evaluating them as AI workflow builders should compare more than connector counts. Agent workflows may need model calls, tool use, branching, memory, human approval, observability, and deployment controls in the same system. +The alternatives ranked on this page solve different automation problems, so compare more than connector counts. Agent workflows may need model calls, tool use, branching, memory, human approval, observability, deployment controls, and a license that fits the intended use in the same system. -This page focuses specifically on alternatives to n8n. For the broader head-term comparison, see [the best AI agent builders in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). - -n8n alternatives now need to be evaluated as AI agent workflow builders, not only as general-purpose automation tools. Buyers comparing platforms for agentic workflows should examine model access, tool calling, branching, memory and state handling, human approval steps, observability, deployment options, and licensing alongside conventional app integrations. +This page is a shortlist of tools that replace or supplement n8n. For a feature-by-feature Sim and n8n breakdown, see [Sim vs n8n](https://www.sim.ai/comparisons/n8n). For the broader category, see [the best AI agent builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). ## What is an AI agent workflow builder? An AI agent workflow builder is software such as Sim or n8n that lets teams connect models, tools, data, logic, memory, and human approvals into multi-step agent workflows. -Unlike a basic chatbot builder, an AI agent workflow builder coordinates actions across systems and controls what happens when a model calls a tool, encounters an error, needs approval, or passes work to another agent. When comparing n8n alternatives for this use case, evaluate agent orchestration, model support, debugging, deployment control, integrations, and licensing—not only the number of automation templates. +Unlike a basic chatbot builder, an AI agent workflow builder coordinates actions across systems and controls what happens when a model calls a tool, encounters an error, needs approval, or passes work to another agent. Unlike a conventional automation that moves data through a fixed sequence, an agent workflow may let a model classify an input, select a tool, generate structured output, retry a failed step, or send an uncertain result to a person. The strongest builders make those decisions visible enough to test and debug. @@ -133,10 +80,6 @@ When comparing n8n alternatives for this use case, evaluate whether each product That breadth makes n8n a strong incumbent for teams that want conventional app automation and AI steps in the same system. Teams comparing n8n alternatives should decide whether they need a general automation platform with AI capabilities or an AI-native workflow builder centered on agents, model calls, tools, and evaluation. -As of August 2026, n8n documents [AI workflow capabilities](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent/tools-agent/) alongside its broader automation features. That makes n8n a relevant incumbent for buyers comparing AI workflow builders, especially when integrations and general business automation are as important as agent-specific development. - -Teams should still separate product capability from licensing terminology. As of August 2026, n8n uses the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), a source-available license that is not approved by the Open Source Initiative. Access to source code does not by itself make n8n an OSI-approved open-source AI workflow builder. - ## What should you look for in an n8n alternative for AI agent workflows? Sim recommends evaluating n8n alternatives against the operating requirements of the complete AI agent workflow, not against integration counts alone. @@ -249,7 +192,7 @@ Sim is the strongest n8n alternative in this comparison for teams prioritizing v Sim is designed around building and running workflows that connect AI models, tools, APIs, and control logic. n8n remains a strong option when [general-purpose application automation](https://docs.n8n.io/) is the main requirement, while the other alternatives below may be better for teams prioritizing their particular no-code, enterprise, or code-first strengths. -This conclusion is specific to n8n-alternative intent. Buyers researching the broader head term should use Sim’s dedicated guide to the best AI agent builders in 2026, linked in Related comparisons below. +The practical decision is not whether a platform can execute a model call; Sim, n8n, and most tools below can. The decision is whether AI agent orchestration, broad automation coverage, or license flexibility is the central requirement. For a side-by-side of Sim and n8n on features, pricing, security, and deployment, see [Sim vs n8n](https://www.sim.ai/comparisons/n8n). If OpenAI's agent tooling is also on your shortlist, see [Sim vs n8n vs OpenAI AgentKit](https://www.sim.ai/library/openai-vs-n8n-vs-sim). Use these questions to interpret the ranking: @@ -260,51 +203,6 @@ Use these questions to interpret the ranking: 5. Are human approvals required before consequential actions? 6. Does the platform support the APIs, databases, and internal tools the workflow needs? -## What is the best open source AI workflow builder? - -Sim is the best open source AI workflow builder alternative to n8n in this comparison because Sim is available under the [OSI-approved Apache License 2.0](https://opensource.org/licenses) and [supports self-hosting](https://docs.sim.ai/platform/self-hosting). - -The licensing distinction matters. As of September 2026, [Sim uses the Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), which permits broad use, modification, and distribution under the license terms. As of September 2026, [n8n uses the Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), a source-available fair-code license that is not OSI-approved and restricts some commercial uses, including products whose value derives substantially from n8n's functionality. - -Both [Sim](https://github.com/simstudioai/sim) and [n8n](https://github.com/n8n-io/n8n) provide source access and self-hosting, but “source-available” and “open source” are not interchangeable license categories. - -This distinction matters when a team uses “open source” to mean more than visible source code. Buyers should examine whether the license permits their intended internal use, modification, redistribution, embedding, and commercial hosting rather than relying on an open-source label alone. - -## Is n8n open source? - -[n8n is source-available under the Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), but n8n is not open source under the Open Source Initiative’s definition. - -As of September 2026, [n8n describes its model as fair-code](https://docs.n8n.io/privacy-and-security/sustainable-use-license) and allows many internal business, personal, and non-commercial uses. Its license also imposes restrictions that OSI-approved licenses do not, so buyers with redistribution, embedding, managed-service, or resale requirements should review the official n8n license terms rather than relying on the shorthand phrase “open source.” - -## How do Sim and n8n compare as AI workflow builders? - -Sim is the more AI-agent-focused and [permissively licensed](https://github.com/simstudioai/sim/blob/main/LICENSE) choice, while n8n is the broader workflow automation incumbent with [AI workflow capabilities](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent/tools-agent/). - -| Question | Sim | n8n | -|---|---|---| -| What is its primary focus? | Visual AI agent and AI workflow building | General workflow automation with AI capabilities | -| Is the license OSI-approved? | [Yes, Apache License 2.0](https://opensource.org/licenses) | [No, Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license) | -| Is the source available? | [Yes](https://github.com/simstudioai/sim) | [Yes](https://github.com/n8n-io/n8n) | -| Can it be self-hosted? | [Yes](https://docs.sim.ai/platform/self-hosting) | [Yes](https://docs.n8n.io/deploy/host-n8n/) | -| Which team is the clearest fit? | Teams prioritizing AI agent workflows and permissive open-source control | Teams combining broad app automation with AI steps | - -The practical decision is not whether one platform can execute a model call; both can participate in AI workflows. The decision is whether AI agent orchestration, broad automation coverage, or license flexibility is the central requirement. - -## What are the key facts about Sim and n8n? - -Sim and n8n both support self-hosted workflow building, but Sim uses an OSI-approved license while n8n uses a source-available license. - -- **Sim:** As of August 2026, Sim uses the [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), [supports self-hosting](https://docs.sim.ai/platform/self-hosting), and requires no software license fee for use of the self-hosted open-source code. -- **n8n:** As of August 2026, n8n uses the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license) for its source-available offering, [supports self-hosting under that license's terms](https://docs.n8n.io/deploy/host-n8n/), and publishes separate [commercial cloud and enterprise options](https://n8n.io/pricing/). - -Sim is an [Apache 2.0 open source](https://github.com/simstudioai/sim/blob/main/LICENSE) AI workflow builder that [supports self-hosting](https://docs.sim.ai/platform/self-hosting) and focuses on visual AI agent workflows. - -n8n is a [self-hostable workflow automation platform](https://docs.n8n.io/deploy/host-n8n/) distributed under the source-available [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), which is not OSI-approved. - -[Sim](https://sim.ai) and [n8n](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent/tools-agent/) both support building workflows that connect AI models with external tools, APIs, and application logic. - -Sim is the clearer choice when an OSI-approved license is mandatory, while n8n is a strong incumbent when broad workflow automation is the priority. - ## The 10 Best n8n Alternatives in 2026 ### Sim @@ -312,7 +210,7 @@ Sim is the clearer choice when an OSI-approved license is mandatory, while n8n i **Best for:** People building AI agent workflows that require multi-model orchestration, persistent memory, and visual collaboration. -[Sim](https://sim.ai) +[Sim](https://www.sim.ai) is an open-source AI agent builder and visual workflow platform with [1,000+ integrations](https://www.sim.ai/pricing), as of September 2026. We support workflows built around agents, knowledge bases, tables, and multi-model routing. You can use Sim Cloud or deploy Sim with Docker Compose or Kubernetes. Both options support the same goal of building agent workflows without requiring you to manage infrastructure unless you choose to. **Strengths:** @@ -327,7 +225,7 @@ Sim is the clearer choice when an OSI-approved license is mandatory, while n8n i Sim publishes its source code for review. Consult Sim's current security documentation for information about audits, controls, and compliance status. **Pricing:** - As of September 2026, Sim offers a free plan and paid plans with credit-based billing. Check [Sim's current pricing](https://sim.ai/pricing) for execution charges, storage allowances, and plan limits. + As of September 2026, Sim offers a free plan and paid plans with credit-based billing. Check [Sim's current pricing](https://www.sim.ai/pricing) for execution charges, storage allowances, and plan limits. **Limitations:** Sim is a newer platform, so some niche third-party integrations available in more mature tools like Zapier or Make may not yet exist. The templates library is growing but not yet as extensive as more established competitors. @@ -539,18 +437,16 @@ Workato is a managed enterprise integration platform for deployments that requir ## How Open-Source Licenses Affect n8n Alternatives -The applicable software license determines how you may self-host, modify, and redistribute an n8n alternative. Review the license for the specific edition you plan to deploy because community and enterprise editions may use different terms. +The applicable software license determines how you may self-host, modify, and redistribute an n8n alternative. Review the license for the specific edition you plan to deploy because community and enterprise editions may use different terms. For a clause-by-clause look at what each model allows for SaaS hosting, white-labeling, and redistribution, see [Apache 2.0 vs fair-code](https://www.sim.ai/library/apache-2-0-vs-fair-code). -**n8n uses a source-available Sustainable Use License rather than an OSI-approved open-source license.** - -n8n makes its source code available under the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license). The [Open Source Definition](https://opensource.org/osd) does not permit restrictions on fields of endeavor, and n8n describes its licensing model as fair-code rather than open source. The Sustainable Use License permits internal use, modification, and some redistribution, but it restricts products whose value derives substantially from n8n's functionality. Agencies, consultancies, and SaaS companies should review the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license) and seek legal advice before offering n8n functionality to customers. +**n8n is source-available, not OSI-approved open source.** + n8n makes its source code available under the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), which n8n describes as fair-code. The license permits internal use, modification, and some redistribution, but it restricts products whose value derives substantially from n8n's functionality, so it does not meet the [Open Source Definition](https://opensource.org/osd). Agencies, consultancies, and SaaS companies should review the license and seek legal advice before offering n8n functionality to customers. **Activepieces uses the MIT license.** - Activepieces distributes its Community Edition under the MIT license, which permits use, modification, and redistribution when you retain the required notices. The MIT license generally permits commercial use, redistribution, and modification when you preserve the required copyright and license notices. Confirm the license that applies to any enterprise features you use. - -**Sim supports open-source, self-hosted deployment.** + Activepieces distributes its Community Edition under the MIT license, which permits commercial use, modification, and redistribution when you preserve the required copyright and license notices. Confirm the license that applies to any enterprise features you use. -Sim provides deployment options through Docker Compose and Kubernetes. Review Sim's repository, applicable license, and current security documentation before choosing a self-hosted deployment. +**Sim uses the Apache License 2.0.** + Sim's core code is released under the OSI-approved [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) and can be self-hosted with Docker Compose or Kubernetes. Features in `apps/sim/ee` use a [separate Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE). **Migration effort varies with workflow complexity.** Moving a portfolio of workflows from n8n to any alternative requires real engineering effort. Migration effort grows with the number of workflows, custom code steps, credentials, external dependencies, and differences between execution models. Activepieces may preserve more trigger-action concepts, while Sim or Pipedream may require you to redesign state management, agent behavior, or code execution. @@ -558,41 +454,20 @@ Sim provides deployment options through Docker Compose and Kubernetes. Review Si **Self-hosted isn't automatically cheaper.** Before assuming that self-hosting eliminates automation costs, build a detailed estimate for total cost of ownership. Include server provisioning, database hosting, TLS certificates, security patches, version upgrades, monitoring, and maintenance time. A managed alternative may cost less when its subscription replaces enough infrastructure and engineering work. -## Which n8n alternative should you choose? - -Sim should be the first n8n alternative evaluated by teams whose primary requirement is building [self-hostable AI agent workflows](https://docs.sim.ai/platform/self-hosting) under a [permissive open-source license](https://github.com/simstudioai/sim/blob/main/LICENSE). - -Choose according to the workload rather than selecting one universal winner: - -- Choose Sim when AI agents, visual orchestration, Apache 2.0 licensing, and self-hosting are the priorities. -- Choose n8n when [broad application automation](https://docs.n8n.io/) and an established automation ecosystem matter more than an OSI-approved license. -- Choose one of the other alternatives in this ranking when its documented specialty matches a requirement that Sim or n8n does not serve as directly. - -## How should you choose between n8n and an AI-native workflow builder? - -Teams should choose n8n for broad general automation and choose an AI-native workflow builder such as Sim when agent behavior is the center of the system. - -[n8n is a credible incumbent](https://docs.n8n.io/) when a workflow combines many conventional integrations with selected AI steps. Sim is a stronger candidate when builders need to design, run, inspect, and [self-host AI agent workflows](https://docs.sim.ai/platform/self-hosting) under [Apache 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE). - -A short proof of concept should test the same production-shaped workflow in each finalist. Include at least one model call, one tool invocation, one conditional branch, one failure or retry, and one human approval so the comparison reveals operational differences rather than demo-level similarities. - -## Related comparisons - -Sim maintains separate comparisons for broader buyer questions so this n8n alternatives page can remain focused on replacing or supplementing n8n. - -- For the broad category question, read [What is the best AI agent builder in 2026?](https://www.sim.ai/library/best-ai-agent-builder-2026). -- For n8n replacement decisions, continue with the ten n8n alternatives and product profiles on this page. -- For open-source decisions, compare the exact license, self-hosting model, and deployment requirements before selecting a platform. - ## The Bottom Line Choose an n8n alternative based on the workflow type and operating model you need. -- **Best for self-hosting:** Choose n8n for developer-controlled, traditional automation when your team is comfortable managing infrastructure. Choose Activepieces when an unrestricted MIT license is the priority, or Sim when you need self-hosted AI agent workflows. +- **Best for self-hosting:** Choose n8n for developer-controlled, traditional automation when your team is comfortable managing infrastructure. Choose Activepieces when an unrestricted MIT license is the priority, or Sim when you need self-hosted AI agent workflows under Apache 2.0. - **Best for AI-native workflows:** As of September 2026, choose Sim for an open-source visual AI agent builder with [1,000+ integrations](https://www.sim.ai/pricing), native multi-model orchestration, agent memory, and MCP support. It also offers managed cloud infrastructure for teams that do not want to self-host. Choose Gumloop for a narrower, non-developer-focused LLM workflow canvas. - **Best for small teams:** Choose Make for managed visual automation, strong debugging, and flexible data transformation. - **Best for non-technical users:** Choose Zapier for the fastest route to straightforward SaaS automation and the widest integration catalog. Choose Gumloop instead when AI is central to the workflow. -Use the comparison framework to narrow the list to two or three candidates, then build the same representative workflow in each one. Compare setup time, debugging, execution cost, deployment effort, and the agent capabilities your production workflow requires. Sim, Make, Zapier, Activepieces, and Pipedream offer free tiers that may support an initial test. Confirm current limits before building a proof of concept. +Use the comparison framework to narrow the list to two or three candidates, then build the same representative workflow in each one. Include at least one model call, one tool invocation, one conditional branch, one failure or retry, and one human approval so the test reveals operational differences rather than demo-level similarities. Sim, Make, Zapier, Activepieces, and Pipedream offer free tiers that may support an initial test. Confirm current limits before building a proof of concept. + +## Related comparisons -For related comparisons, see our guides to [Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives), [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms), and [AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). +- [Sim vs n8n](https://www.sim.ai/comparisons/n8n): feature, pricing, security, and deployment details for the two platforms side by side. +- [Sim vs n8n vs OpenAI AgentKit](https://www.sim.ai/library/openai-vs-n8n-vs-sim): how agent-first, automation-first, and OpenAI-native architectures differ. +- [Apache 2.0 vs fair-code](https://www.sim.ai/library/apache-2-0-vs-fair-code): what each license allows for self-hosting and commercial products. +- [Best AI agent platforms and builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), [Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives), and [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms) for broader research. diff --git a/apps/sim/content/library/open-source-ai-agent-platforms/index.mdx b/apps/sim/content/library/open-source-ai-agent-platforms/index.mdx index 94eb6e714bb..02b4ec0892d 100644 --- a/apps/sim/content/library/open-source-ai-agent-platforms/index.mdx +++ b/apps/sim/content/library/open-source-ai-agent-platforms/index.mdx @@ -1,30 +1,33 @@ --- slug: open-source-ai-agent-platforms -title: 'Open-Source AI Agent Platforms: Comparison' -description: Compare the top open-source AI agent platforms of 2026 - LangGraph, CrewAI, AutoGen, Dify, n8n, and Sim - by architecture, production readiness, and team fit. Find the right one for your use case. +title: 'Open-Source AI Agent Platforms and Frameworks Compared' +description: Compare the top open-source AI agent platforms and frameworks of 2026 - LangGraph, CrewAI, AutoGen, Dify, n8n, and Sim - by architecture, license, production readiness, and team fit. Find the right one for your use case. date: 2026-07-13 -updated: 2026-07-23 +updated: 2026-09-30 authors: - emir -readingTime: 12 -tags: [Open Source, AI Agents, LangGraph, CrewAI, Dify, Sim] +readingTime: 14 +tags: [Open Source, AI Agents, Agent Frameworks, LangGraph, CrewAI, Dify, Sim] ogImage: /library/open-source-ai-agent-platforms/cover.jpg -canonical: https://www.sim.ai/library/open-source-ai-agent-platforms draft: false faq: - q: "What is the difference between an AI agent framework and an AI agent platform?" a: "An AI agent framework is a code library that provides primitives for building agents: tool use, multi-step reasoning, memory, and orchestration. You write code and own everything else. An AI agent platform bundles those primitives with deployment infrastructure, observability, collaboration features, and often a visual interface. The practical difference is how much your team builds versus how much comes out of the box." - q: "Can I self-host all of these open-source AI agent platforms?" - a: "Most, but not all, support full self-hosting. LangGraph, CrewAI, Dify, and Sim can all be self-hosted via Docker or Kubernetes. AutoGen is now in maintenance mode and will not receive new features, but remains self-hostable. n8n supports self-hosting under its Sustainable Use License, which has specific commercial-use restrictions worth reviewing. Always check the license terms, since some platforms label enterprise features like RBAC, SSO, and advanced observability as paid add-ons even when the core is open source." + a: "Most, but not all, support full self-hosting. LangGraph, CrewAI, Dify, and Sim can all be self-hosted via Docker or Kubernetes. Sim uses Apache 2.0, and LangGraph and CrewAI use MIT, all of which permit commercial self-hosting. Dify's license is based on Apache 2.0 but adds conditions, including a restriction on running a commercial multi-tenant service without separate permission. AutoGen is now in maintenance mode and will not receive new features, but remains self-hostable. n8n supports self-hosting under its Sustainable Use License, which has specific commercial-use restrictions worth reviewing. Always check the license terms, since some platforms label enterprise features like RBAC, SSO, and advanced observability as paid add-ons even when the core is open source." - q: "Which open-source AI agent platform is best for non-developers?" a: "Visual builders like Dify and workspace platforms like Sim are the best starting points for non-developers. Both offer drag-and-drop interfaces that don't require writing code. Code-first frameworks like LangGraph, CrewAI, and AutoGen are a poor fit without engineering support since they require Python proficiency and comfort with infrastructure management. If your team is mixed (some developers, some not), a workspace like Sim lets both groups contribute in the same environment." - q: "How does LangGraph compare to CrewAI for production use?" a: "LangGraph gives you explicit, stateful control over every decision branch in your agent's workflow, making it the stronger choice for complex conditional logic. Crews provide autonomous agent collaboration ideal for tasks requiring flexible decision-making, while Flows offer precise, event-driven control ideal for managing detailed execution paths and secure state management. The real differentiator in production is the deployment layer: both frameworks leave production infrastructure, monitoring, and team collaboration as exercises for the builder, so your choice may hinge on which ecosystem your team prefers to invest in." - q: "What should I look for in an open-source AI agent platform before committing?" a: "Evaluate six things: license type (MIT, Apache 2.0, or a custom license with restrictions), self-hosting support (Docker/Kubernetes readiness and local model compatibility via Ollama), observability (built-in logging and tracing versus requiring a paid add-on like LangSmith), LLM flexibility (multi-provider support so you're not locked into one model vendor), community activity (commit frequency, issue response time, contributor count), and enterprise feature gating (whether RBAC, SSO, and audit logs require a paid tier). That last point matters most: an open-source label doesn't guarantee the features you need in production are in the free tier." + - q: "What happened to Flowise?" + a: "Flowise stopped development on July 29, 2026, and archived its GitHub repository on August 13, 2026. Existing self-hosted installations keep running but receive no upstream fixes, so new projects should choose an actively maintained visual builder such as Sim or Dify. Flowise Cloud users should check official Flowise notices for any migration deadline." + - q: "Which open-source AI agent platforms support MCP?" + a: "MCP (Model Context Protocol) gives agents a standard way to reach tools and context. Sim can call MCP tools and can also deploy a workflow as an MCP server. Dify supports MCP integration, and CrewAI agents can connect to MCP servers." --- -This guide splits open-source AI agent platforms into three clear categories, compares the leading options within each, and offers a decision framework based on your team's situation rather than by feature count. +This guide splits open-source AI agent platforms and frameworks into three clear categories, compares the leading options within each, and offers a decision framework based on your team's situation rather than by feature count. ## Key Takeaways @@ -59,11 +62,11 @@ That control comes at a cost: these frameworks assume you have engineers who can ### LangGraph -[LangGraph](https://langchain-ai.github.io/langgraph/) is the default choice for complex stateful workflows that need explicit control over branching, retries, and human-in-the-loop. It sits on top of the LangChain ecosystem and has seen the largest enterprise adoption among code-first agent frameworks. +[LangGraph](https://langchain-ai.github.io/langgraph/) is the default choice for complex stateful workflows that need explicit control over branching, retries, and human-in-the-loop. It sits underneath the LangChain ecosystem: since LangChain 1.0, [LangChain's `create_agent` runs on LangGraph](https://docs.langchain.com/oss/python/releases/langgraph-v1). It has seen the largest enterprise adoption among code-first agent frameworks. LangGraph does four things excellently: branching logic that lets you define exactly which path an agent takes based on state, human-led approvals and checkpointing that are now integral features rather than add-ons, durable execution that survives process restarts, and detailed control over every step in the agent's decision chain. -Where it demands investment: setup isn't trivial, especially for teams new to graph-based agent architecture. LangSmith (LangChain's paid observability platform) is the recommended way to monitor and debug LangGraph workflows in production, which introduces a dependency on a proprietary tool sitting alongside the open-source framework. And LangGraph is fundamentally a developer tool. If your team includes non-engineers who need to build or modify agents, they won't be able to participate without an intermediate layer. +Where it demands investment: setup isn't trivial, especially for teams new to graph-based agent architecture. LangSmith (LangChain's paid observability platform) is the recommended way to monitor and debug LangGraph workflows in production. You don't need it to build or run graphs, but teams that skip it must assemble their own tracing. And LangGraph is fundamentally a developer tool. If your team includes non-engineers who need to build or modify agents, they won't be able to participate without an intermediate layer. **Best for:** Teams with strong engineering resources building complex stateful agents where explicit control over every decision branch matters more than speed to first deployment. @@ -73,7 +76,7 @@ Where it demands investment: setup isn't trivial, especially for teams new to gr The Flows addition lets you create structured, event-driven workflows that provide a way to connect multiple tasks, manage state, and control the flow of execution in your AI applications. This is a meaningful evolution. Flows give you a structured, event-driven execution engine that sits above individual crews and tasks. A Crew is great at parallel collaboration with multiple agents working on a shared goal, but Crews don't give you sequential control. Think of it this way: a Crew is a team, a Flow is the project plan that coordinates multiple teams. -The open-core dynamic is worth understanding before you commit. CrewAI is open-source and actively encourages community contributions. The MIT-licensed core gives you the framework for free, but CrewAI's AMP Suite provides tracing and observability, a unified control plane for managing and scaling agents, and enterprise integrations as paid enterprise tooling. Teams that want a UI, role-based access control, and managed deployments will eventually encounter the paid tier. +The open-core dynamic is worth understanding before you commit. CrewAI is open-source and actively encourages community contributions. The MIT-licensed core gives you the framework for free, but CrewAI's AMP Suite provides tracing and observability, a unified control plane for managing and scaling agents, and enterprise integrations as paid enterprise tooling. According to [CrewAI's pricing page](https://crewai.com/pricing), the AMP platform adds a visual editor, managed deployment, and governance features; a limited tier is free, while enterprise capabilities require custom pricing. Teams that want a UI, role-based access control, and managed deployments will eventually encounter the paid tier. **Best for:** Teams automating multi-step workflows where work naturally breaks into distinct role specializations, and where the Flows layer provides enough orchestration to avoid building a custom control plane. @@ -97,13 +100,13 @@ Visual builders trade code-level control for speed. They let teams design agent ### Dify -Dify's open-source model with 131k GitHub stars targets production scalability. That star count makes it the most-starred visual AI agent builder in the open-source space by a wide margin, and it reflects genuine production adoption. +Dify's open-source model targets production scalability. With more than 149,000 GitHub stars, it is the most-starred open-source visual builder focused specifically on LLM applications and AI agents, which reflects broad production adoption. [Dify](https://dify.ai/pricing) is a production-ready platform for agentic workflow development, handling everything from enterprise QA bots to AI-driven custom assistants. The platform includes a workflow builder for defining tool-using agents, built-in RAG (retrieval-augmented generation) pipeline management, support for multiple AI model providers, and Model Context Protocol (MCP) integration. The RAG pipeline is Dify's standout feature. It's among the best available in an open-source package. If your primary use case involves document retrieval, knowledge bases, and structured Q&A, Dify's built-in tooling eliminates weeks of integration work. -The self-hosted Community Edition (Docker Compose, single machine or Kubernetes) is free with no significant limitations. [Dify Cloud](https://dify.ai/pricing) starts with a free Sandbox tier at 200 message credits and scales to Professional, Team, and Enterprise plans; the Professional tier lists at $590/year and Team at $1,590/year. +The self-hosted Community Edition (Docker Compose, single machine or Kubernetes) is free for most uses. Its [license](https://github.com/langgenius/dify/blob/main/LICENSE) is based on Apache 2.0 with added conditions, including a restriction on operating a commercial multi-tenant service without separate permission, so review it before offering Dify as a hosted product. [Dify Cloud](https://dify.ai/pricing) starts with a free Sandbox tier at 200 message credits and scales to Professional, Team, and Enterprise plans; the Professional tier lists at $590/year and Team at $1,590/year. Where Dify falls short relative to a purpose-built AI workspace: team governance is limited, agent lifecycle management (versioning, rollback, multi-user editing) lacks depth, and the visual tooling has a ceiling; complex custom logic belongs in code. @@ -123,7 +126,7 @@ However, it is worth emphasizing that n8n is an AI-augmented workflow tool, not The gap between a framework and a workspace comes down to what's included in the box. With a code-first framework, you get agent logic. You then need to separately build or buy your deployment infrastructure, observability layer, collaboration tooling, and knowledge management system. With a visual builder, you get faster assembly but often the same gaps in governance and team workflows. -[Sim](https://sim.ai) is built around the premise that those layers belong together. It's an open-source AI workspace that combines drag-and-drop agent building, real-time multi-user collaboration, built-in knowledge management, and deployment infrastructure in one environment. +[Sim](https://www.sim.ai) is built around the premise that those layers belong together. It's an open-source AI workspace that combines drag-and-drop agent building, real-time multi-user collaboration, built-in knowledge management, and deployment infrastructure in one environment. The feature set maps directly to what teams need in a comparison context: @@ -132,6 +135,8 @@ The feature set maps directly to what teams need in a comparison context: - **Multi-LLM support:** OpenAI, Claude, Gemini, Mistral, xAI, plus local models via Ollama for teams with cost or privacy constraints - **MCP protocol support:** Model Context Protocol for standardized external API and service connections - **Real-time collaboration:** Multiple team members building workflows simultaneously with live editing, commenting, and granular permission controls +- **Built-in workspace resources:** Tables, Files, and Knowledge Bases give agents reusable context and storage without separate services +- **One workflow, several surfaces:** [Deploy the same workflow](https://docs.sim.ai/workflows/deployment) as a REST API, a hosted chat, or a set of MCP tools - **Deployment flexibility:** Cloud-hosted with managed infrastructure, or self-hosted via Docker Compose or Kubernetes for complete data control The open-source commitment is backed by community traction: over 100,000 builders, alongside SOC2 compliance as a production trust signal. That certification is important when the conversation moves from "prototype" to "production" and legal needs to sign off. @@ -140,6 +145,14 @@ Chat, Sim's natural-language interface, lets you talk to Sim to build and manage **Best for:** Teams that need to move from prototype to production without stitching together a separate framework, observability tool, and deployment layer, especially when team collaboration and multi-model flexibility are requirements, not nice-to-haves. +## Archived and Adjacent Projects: Flowise and OpenHands + +Two other projects show up in most open-source agent lists. Neither belongs in the main comparison, for different reasons. + +**Flowise is archived.** Flowise was a popular drag-and-drop builder for LLM apps, but its maintainers stopped development on July 29, 2026, and [archived the repository](https://github.com/FlowiseAI/Flowise) on August 13, 2026. Existing self-hosted installs keep running, but they get no upstream fixes as model APIs, dependencies, and security requirements change. You would need to maintain a private fork or migrate. For new projects, pick an actively maintained visual builder like Sim or Dify. + +**OpenHands is a coding agent, not a general agent builder.** [OpenHands](https://github.com/All-Hands-AI/OpenHands/) agents inspect repositories, plan code changes, and apply them in a working environment. They can [review pull requests, triage issues, and react to CI events](https://docs.openhands.dev/openhands/usage/automations/event-automations). The core is MIT-licensed and runs locally, with cloud and self-hosted enterprise options. Choose it when the job is software delivery; for business agents that work across Slack, CRMs, and documents, use one of the platforms above. [AI coding agents vs AI workflow agents](/library/ai-coding-agents-vs-ai-workflow-agents) covers that split in more depth, and Sim's comparison of [agentic AI coding tools](https://www.sim.ai/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare) covers how proprietary options such as Cursor and Claude Code differ in license, hosting, and interface. + ## Side-by-Side Comparison This table surfaces the dimensions that actually affect the build-vs.-buy decision for open source AI agent platforms. @@ -150,14 +163,14 @@ This table surfaces the dimensions that actually affect the build-vs.-buy decisi | **LangGraph** | Code-first framework | MIT | Yes | No | Yes (via LangChain) | No (code-level only) | Via LangSmith (paid) | Complex stateful agents with engineering teams | | **CrewAI** | Code-first framework | MIT | Yes | Enterprise tier only | Yes (via LiteLLM) | Enterprise tier | Enterprise tier | Role-based multi-agent workflows | | **AutoGen/AG2** | Code-first framework | MIT | Yes | AutoGen Studio (prototyping) | Yes | No | Limited | Research, prototyping, Microsoft ecosystem | -| **Dify** | Visual builder | Apache 2.0 | Yes (Docker/K8s) | Yes | Yes (100+ providers) | Limited | Built-in dashboard | RAG apps, LLM gateway, quick deployment | +| **Dify** | Visual builder | Modified Apache 2.0 (multi-tenant restriction) | Yes (Docker/K8s) | Yes | Yes (100+ providers) | Limited | Built-in dashboard | RAG apps, LLM gateway, quick deployment | | **n8n** | Workflow automation + AI | Sustainable Use License | Yes | Yes | Limited (via AI nodes) | Yes | Built-in | Migrating automation workflows to AI | You should also consider: - **Pricing:** Most of these platforms are free at the core but diverge sharply at the enterprise tier. CrewAI and LangGraph push production tooling into paid layers. Dify and Sim offer meaningful free tiers with paid cloud options. n8n's license has specific use restrictions worth reading. - **Ecosystem maturity:** LangGraph benefits from the broader LangChain ecosystem (documents, loaders, tools). Dify has the largest visual-builder community. CrewAI's developer certification program has grown its user base fast. -- **Community size:** GitHub stars are a useful comparison point, but don't tell the whole story. n8n's 180k+ stars and Dify's 131k+ stars reflect automation-community momentum. Smaller star counts for newer platforms like Sim don't map directly to capability gaps. +- **Community size:** GitHub stars are a useful comparison point, but don't tell the whole story. n8n's 180k+ stars and Dify's 149k+ stars reflect automation-community momentum. Smaller star counts for newer platforms like Sim don't map directly to capability gaps. ## How to Choose: A Use-Case Decision Guide @@ -189,6 +202,6 @@ Visual builders like Dify and n8n lower the barrier to entry but trade away arch Start with the decision paths above. Identify which category fits your team, then evaluate within that category. Trying to compare a Python framework against a visual workspace on the same checklist is how teams end up six months into a tool that doesn't fit. -If your team needs collaboration, multi-model support, and a visual builder with production-grade deployment, [explore Sim](https://sim.ai) and see whether the workspace model matches how your team actually works. +If your team needs collaboration, multi-model support, and a visual builder with production-grade deployment, [explore Sim](https://www.sim.ai) and see whether the workspace model matches how your team actually works. -Related reading: [Apache 2.0 vs fair-code](/library/apache-2-0-vs-fair-code) explains why license choice changes what you can build on a self-hosted platform, [LangGraph alternatives](/library/langgraph-alternatives) goes deeper on the code-first category, and [the best AI agent platforms in 2026](/library/best-ai-agent-platforms-2026) widens the field to commercial options alongside these. +Related reading: [Apache 2.0 vs fair-code](/library/apache-2-0-vs-fair-code) explains why license choice changes what you can build on a self-hosted platform, [LangGraph alternatives](/library/langgraph-alternatives) and [the best multi-agent frameworks](/library/best-multi-agent-frameworks-2026) go deeper on the code-first category, and [the best AI agent platforms in 2026](/library/best-ai-agent-platforms-2026) widens the field to commercial options alongside these. diff --git a/apps/sim/content/library/openai-vs-n8n-vs-sim/index.mdx b/apps/sim/content/library/openai-vs-n8n-vs-sim/index.mdx index 42da4e31dbb..534929708c8 100644 --- a/apps/sim/content/library/openai-vs-n8n-vs-sim/index.mdx +++ b/apps/sim/content/library/openai-vs-n8n-vs-sim/index.mdx @@ -1,19 +1,18 @@ --- slug: openai-vs-n8n-vs-sim title: 'Sim vs n8n vs OpenAI AgentKit: AI Agent Builder Comparison (2026)' -description: 'Compare Sim with n8n and OpenAI AgentKit on integrations and deployment. See how Sim''s open-source platform works with multiple model providers.' +description: 'Sim vs n8n vs OpenAI AgentKit: how an agent-first, an automation-first, and an OpenAI-native agent builder differ on architecture, licensing, self-hosting, and model choice.' date: 2025-10-06 -updated: 2026-09-21 +updated: 2026-09-30 authors: - emir readingTime: 14 tags: [AI Agents, Workflow Automation, OpenAI AgentKit, n8n, Sim, MCP] ogImage: /library/openai-vs-n8n-vs-sim/cover.jpg -canonical: https://www.sim.ai/library/openai-vs-n8n-vs-sim draft: false faq: - q: "What is the best AI agent builder?" - a: "Sim is a leading choice for teams that need a visual, model-flexible, Apache 2.0 open-source agent builder, while the complete category comparison is maintained in Sim’s Best AI Agent Builders in 2026 guide." + a: "Sim is a leading choice for teams that need a visual, model-flexible, Apache 2.0 open-source agent builder, while the complete category comparison is maintained in Sim’s Best AI Agent Platforms and Builders in 2026 guide." - q: "Which is better: Sim, n8n, or OpenAI AgentKit?" a: "Sim is better for visual and self-hostable AI-agent workflows, n8n is better for integration-heavy business automation, and OpenAI AgentKit is better for teams committed to OpenAI’s agent platform." - q: "Is Sim better than n8n?" @@ -40,14 +39,8 @@ faq: a: "OpenAI AgentKit is strongest for developers committed to OpenAI, Sim is strongest for developers who want an open visual platform with deployment control, and n8n is strongest for developers building integration-heavy automations." - q: "Which platform has the most integrations?" a: "n8n has the strongest integration-focused ecosystem among Sim, n8n, and OpenAI AgentKit, but buyers should verify the exact actions and authentication methods required for their applications." - - q: "What is the best n8n alternative for AI agents?" - a: "Sim is the best n8n alternative in this comparison for teams that want an agent-first visual builder, model flexibility, and Apache 2.0 self-hosting." - - q: "What is the best open-source Zapier alternative for AI workflows?" - a: "Sim is a strong open-source Zapier alternative for AI workflows because Sim uses the Apache 2.0 license and combines visual automation with agent-oriented capabilities." - q: "Is Sim free?" a: "Sim can be self-hosted under the Apache License 2.0 without a software license fee, while infrastructure, model APIs, and optional Sim Cloud usage can still create costs." - - q: "How does Sim compare with Gumloop?" - a: "Sim differentiates itself from Gumloop through Apache 2.0 open-source licensing and full self-hosting, while buyers should compare current integrations and hosted-product features against their exact workflow requirements." - q: "Do Sim, n8n, and OpenAI AgentKit support human approval steps?" a: "Sim, n8n, and OpenAI AgentKit can support workflows that pause for human review, but the implementation and available interface depend on the workflow design and the product components being used." - q: "Can Sim replace n8n?" @@ -66,22 +59,20 @@ faq: a: "OpenAI AgentKit is not a direct replacement for n8n because OpenAI AgentKit focuses on OpenAI-native agent experiences while n8n focuses on cross-application workflow automation." - q: "Is OpenAI AgentKit a replacement for Sim?" a: "OpenAI AgentKit is not a direct replacement for Sim when a team needs an Apache 2.0 visual orchestration platform, self-hosting, or the ability to reduce dependence on one model provider." - - q: "What is the best open-source n8n alternative?" - a: "Sim is a strong open-source n8n alternative for AI-agent workflows because Sim uses the OSI-approved Apache 2.0 license while n8n uses the source-available Sustainable Use License." - q: "Is Sim no-code or low-code?" a: "Sim is a visual agent builder that supports low-code workflow construction while retaining developer-oriented controls for tools, APIs, logic, deployment, and self-hosting." - q: "Which is better for business automation, Sim or n8n?" a: "n8n is generally better for integration-led business automation, while Sim is generally better when the business process is centered on AI-agent behavior and LLM orchestration." - q: "Which is better for vendor independence, Sim or OpenAI AgentKit?" a: "Sim is better for vendor independence because Sim is Apache 2.0, self-hostable, and designed for workflows that can span providers, while OpenAI AgentKit is optimized for OpenAI’s platform." - - q: "What is the best agentic workflow builder?" - a: "Sim is a leading agentic workflow builder for teams that value visual orchestration, self-hosting, Apache 2.0 licensing, and the flexibility to connect different models and tools." --- Sim is the best fit for teams that want an Apache 2.0 visual agent builder, n8n is strongest for integration-heavy business automation, and OpenAI AgentKit is strongest for teams building directly around OpenAI's agent platform. The products overlap, but they are not interchangeable. Sim centers on portable agent workflows and open-source ownership; [n8n centers on workflow automation that combines AI with business processes](https://docs.n8n.io/); and [OpenAI's current agent stack offers the Agents API, Agents SDK, Responses API, and ChatKit](https://developers.openai.com/api/docs/guides/agents). OpenAI is winding down Agent Builder and its Evals platform, so teams evaluating AgentKit in September 2026 should plan around the current tools rather than those retiring products. +This page compares the three architectures. For a detailed two-way breakdown of features, pricing, and security, see [Sim vs n8n](https://www.sim.ai/comparisons/n8n); for a wider shortlist of tools that replace n8n, see [n8n alternatives](https://www.sim.ai/library/n8n-alternatives). + ## TL;DR Choose **Sim** when agents and LLM workflows are the product, you want a visual canvas, and Apache 2.0 code ownership matters. Choose **n8n** when conventional app integrations, triggers, data movement, and operational automation are the center of the workflow. Choose **OpenAI AgentKit** when your application is committed to OpenAI models and current platform services—but do not start a new architecture around Agent Builder or the Evals platform, which [OpenAI has scheduled to shut down on November 30, 2026](https://openai.com/index/introducing-agentkit/). @@ -144,7 +135,7 @@ n8n is an integration-first workflow automation platform that can add AI agents ![n8n workflow automation interface](/library/openai-vs-n8n-vs-sim/n8n.png) -n8n is self-hostable, but “self-hostable” does not mean “OSI-approved open source.” The official license page identifies the Sustainable Use License as a fair-code license and limits use mainly to internal business purposes and non-commercial or personal use. Organizations planning redistribution, embedding, or a commercial hosted offering should evaluate the [official n8n license terms](https://docs.n8n.io/privacy-and-security/sustainable-use-license) rather than relying on the shorthand “open source.” +n8n is self-hostable, but “self-hostable” does not mean “OSI-approved open source.” The [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license) is a fair-code license that limits use mainly to internal business purposes and non-commercial or personal use. [Apache 2.0 vs fair-code](https://www.sim.ai/library/apache-2-0-vs-fair-code) covers what that means for redistribution, embedding, and hosted offerings. n8n is most suitable when: @@ -183,7 +174,7 @@ A useful decision test is to describe the workflow without naming a product: - If the description begins with “the agent should reason, call tools, and respond,” Sim is usually the more natural fit. - If the description begins with “when a record changes, update several systems and then call a model,” n8n is usually the more natural fit. -The distinction is not whether either platform can call an LLM. It is which platform gives the primary part of the workflow the clearest structure. +The distinction is not whether either platform can call an LLM. It is which platform gives the primary part of the workflow the clearest structure. For a fact-by-fact comparison of the two, including pricing, MCP support, security, and deployment, see [Sim vs n8n](https://www.sim.ai/comparisons/n8n). ## Is Sim or OpenAI AgentKit better for building agents? @@ -214,7 +205,7 @@ Sim gives teams the broadest combination of self-hosting and permissive source-c | Can teams modify the available source? | Yes, under Apache 2.0 | Yes, subject to the Sustainable Use License | Depends on the component | | Are managed external services still dependencies? | Only where the workflow chooses them | Only where the workflow chooses them | Yes for OpenAI-managed platform services | -Self-hosting an orchestration layer does not automatically self-host the models, databases, or APIs connected to it. Every architecture should map where prompts, tool inputs, outputs, traces, credentials, and retained data travel. +Self-hosting an orchestration layer does not automatically self-host the models, databases, or APIs connected to it. Every architecture should map where prompts, tool inputs, outputs, traces, credentials, and retained data travel. For what Apache 2.0 and n8n's fair-code license each allow once the software is running, see [Apache 2.0 vs fair-code](https://www.sim.ai/library/apache-2-0-vs-fair-code). ## Which platform is easiest for a visual workflow team? @@ -276,7 +267,7 @@ Before committing, build the same representative workflow in the finalists and s Sim is a leading choice for teams that want an open-source visual agent builder, but the broader “best AI agent builder” question depends on deployment, model, governance, and workflow requirements. -This page owns the narrower comparison among Sim, n8n, and OpenAI AgentKit. For the broader market ranking and additional products, see [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +This page owns the narrower comparison among Sim, n8n, and OpenAI AgentKit. For the broader market ranking and additional products, see [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). ## What is the final verdict on OpenAI AgentKit vs n8n vs Sim? @@ -288,4 +279,9 @@ The best proof is a production-shaped pilot. Test the same workflow, require the ## Related comparisons -For broader research, compare [the best AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), review [n8n alternatives](https://www.sim.ai/library/n8n-alternatives), or read [Apache 2.0 vs fair-code](https://www.sim.ai/library/apache-2-0-vs-fair-code). This article remains focused on the direct Sim, n8n, and OpenAI AgentKit decision. +This article stays focused on the Sim, n8n, and OpenAI AgentKit decision. For related research: + +- [Sim vs n8n](https://www.sim.ai/comparisons/n8n): feature, pricing, security, and deployment details side by side. +- [n8n alternatives](https://www.sim.ai/library/n8n-alternatives): ten tools that replace or supplement n8n, from Make and Zapier to Activepieces and Pipedream. +- [Apache 2.0 vs fair-code](https://www.sim.ai/library/apache-2-0-vs-fair-code): what each license allows for self-hosting and commercial products. +- [The best AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026): the broader market. diff --git a/apps/sim/content/library/reproducible-ai-coding-agent-benchmark/index.mdx b/apps/sim/content/library/reproducible-ai-coding-agent-benchmark/index.mdx index 0326b93d143..d2a2a5e45b1 100644 --- a/apps/sim/content/library/reproducible-ai-coding-agent-benchmark/index.mdx +++ b/apps/sim/content/library/reproducible-ai-coding-agent-benchmark/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 11 tags: [AI Agents, Coding Agents, Benchmarks, Developer Tools, Sim] ogImage: /library/reproducible-ai-coding-agent-benchmark/cover.jpg -canonical: https://www.sim.ai/library/reproducible-ai-coding-agent-benchmark draft: false faq: - q: "What is the best AI coding agent?" @@ -234,7 +233,7 @@ Sim is a workflow-agent platform rather than a dedicated AI coding agent, so Sim Sim helps teams build and operate AI workflows that connect models, tools, APIs, and data sources. Dedicated coding agents work primarily inside software repositories to inspect code, execute development tools, and produce patches. The distinction is explored further in [AI coding agents vs. AI workflow agents](https://www.sim.ai/library/ai-coding-agents-vs-ai-workflow-agents). -As of September 2026, Sim is available under the [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), which appears on the [OSI list of approved licenses](https://opensource.org/licenses). Readers looking for a broader comparison of platforms for building AI agents should use Sim’s canonical guide to the [best AI agent builders](https://www.sim.ai/library/best-ai-agent-builder-2026). +As of September 2026, Sim is available under the [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), which appears on the [OSI list of approved licenses](https://opensource.org/licenses). Readers looking for a broader comparison of platforms for building AI agents should use Sim’s canonical guide to the [best AI agent builders](https://www.sim.ai/library/best-ai-agent-platforms-2026). ## Why does this benchmark mention n8n? @@ -254,6 +253,6 @@ Teams should inspect category-level results instead of selecting solely from the Sim routes general AI agent-builder questions to its canonical comparison rather than using this benchmark to compete for the same head term. -- For the best overall AI agent builder, read [Best AI Agent Builder 2026](https://www.sim.ai/library/best-ai-agent-builder-2026). +- For the best overall AI agent builder, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). - For coding-agent performance, use this benchmark after the public runs and artifacts are released. - For workflow automation comparisons involving Sim and n8n, use a dedicated workflow-platform comparison rather than coding-agent scores. diff --git a/apps/sim/content/library/sim-open-source-zapier-alternative/index.mdx b/apps/sim/content/library/sim-open-source-zapier-alternative/index.mdx index 0ed8943958a..92a56c3dc28 100644 --- a/apps/sim/content/library/sim-open-source-zapier-alternative/index.mdx +++ b/apps/sim/content/library/sim-open-source-zapier-alternative/index.mdx @@ -1,19 +1,18 @@ --- slug: sim-open-source-zapier-alternative -title: 'Sim: The Open Source Zapier Alternative for AI Agents' -description: 'Compare Sim and Zapier: an Apache 2.0, self-hostable, BYOK agent-native workspace versus proprietary cloud Zaps and Agents across building, deployment, and pricing.' +title: 'Sim vs Zapier: Open-Source AI Agents vs Zaps, Compared' +description: 'Sim vs Zapier, head to head: an Apache 2.0, self-hostable, BYOK AI workspace versus proprietary cloud Zaps and Zapier Agents across licensing, building, agent depth, deployment, and pricing.' date: 2026-09-01 -updated: 2026-09-01 +updated: 2026-09-30 authors: - andrew readingTime: 9 tags: [Automation, Open Source, AI Agents, Sim] ogImage: /library/sim-open-source-zapier-alternative/cover.jpg -canonical: https://www.sim.ai/library/sim-open-source-zapier-alternative draft: false faq: - - q: "Is there an open source alternative to Zapier?" - a: "Sim is an Apache 2.0-licensed open-source alternative to Zapier. Sim combines AI agents and deterministic workflow logic in one workspace. You can inspect the code, self-host the platform, and use your own API keys." + - q: "Is Sim an open-source alternative to Zapier?" + a: "Yes. Sim is an Apache 2.0-licensed open-source alternative to Zapier. Sim combines AI agents and deterministic workflow logic in one workspace. You can inspect the code, self-host the platform, and use your own API keys." - q: "Can I self-host Sim?" a: "Sim supports self-hosting through Docker, Kubernetes, or npx sim-setup. Zapier operates as a proprietary cloud service without customer-managed hosting. Self-hosting gives you more control over deployment and data handling." - q: "Does Sim support BYOK?" @@ -28,6 +27,8 @@ faq: ## TL;DR +This is a head-to-head comparison of Sim and Zapier. If you are still building a shortlist across several vendors, start with the [best Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives), which compares Sim, Make, n8n, Pipedream, Workato, and others by use case. + - Sim is an Apache 2.0-licensed, self-hostable, bring-your-own-key alternative to Zapier. [Zapier provides proprietary cloud automation](https://zapier.com/blog/cloud-vs-self-hosting/), while Sim provides an open-source, agent-native workspace. - Sim combines natural-language building, a visual block canvas, and API or SDK access. Zapier centers on [trigger-action Zaps](https://help.zapier.com/hc/en-us/articles/8496309697421-What-is-a-Zap) and [offers Agents separately](https://zapier.com/agents). - Sim places AI reasoning and deterministic functions, conditions, routers, and loops in one graph. [Zapier adds AI capabilities to an automation-first product](https://zapier.com/blog/zapier-ai-guide/). @@ -39,7 +40,7 @@ Zapier is a proprietary cloud automation platform built around [Zaps](https://he Sim is our [open-source AI workspace](https://github.com/simstudioai/sim) for workflows that combine model reasoning with predictable application logic. A single graph can contain Agent blocks alongside functions, conditions, routers, and loops. You can use AI where a task requires judgment while keeping rules-based steps explicit and repeatable. -Zapier fits users who want familiar no-code automation and [broad access to packaged app connectors](https://zapier.com/apps). Sim fits technical operations, RevOps, and growth users who need to build agent-native systems with more control over logic, models, and deployment. The products overlap in workflow automation, but their starting points differ. Zapier starts with trigger-and-action automation, while Sim starts with AI agents operating inside structured workflows. For a wider market view, compare the [best Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives). +Zapier fits users who want familiar no-code automation and [broad access to packaged app connectors](https://zapier.com/apps). Sim fits technical operations, RevOps, and growth users who need to build agent-native systems with more control over logic, models, and deployment. The products overlap in workflow automation, but their starting points differ. Zapier starts with trigger-and-action automation, while Sim starts with AI agents operating inside structured workflows. ## License, hosting, and who controls the data @@ -51,13 +52,13 @@ BYOK lets you connect supported model providers with your own API credentials. B [Zapier’s managed cloud](https://zapier.com/blog/cloud-vs-self-hosting/) reduces infrastructure work, which may suit buyers who prefer vendor-operated software. Sim gives you more control, but you must operate, secure, monitor, and update the deployment. Buyers searching for an open-source Zapier alternative should weigh Sim’s added deployment control against the work required to operate it. You can also compare other [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms). -## Building workflows: Mothership and blocks vs. Zaps and Agents +## Building workflows: Chat and blocks vs. Zaps and Agents -Sim lets you build one workflow through three interfaces while preserving the same underlying graph. Mothership acts as a natural language control plane, so you can describe the workflow and ask it to create or modify blocks. The visual canvas lets you inspect each connection and edit the blocks directly. For programmatic work, the API and SDK let you create or manage workflows through code. These interfaces are documented in the [Sim introduction](https://docs.sim.ai/introduction). +Sim lets you build one workflow through three interfaces while preserving the same underlying graph. In Chat, you describe the workflow in plain language and ask Sim to create or modify blocks. The visual canvas lets you inspect each connection and edit the blocks directly. For programmatic work, the API and SDK let you create or manage workflows through code. These interfaces are documented in the [Sim introduction](https://docs.sim.ai/introduction). Zapier centers workflow building on [Zaps, which connect a trigger to one or more actions](https://help.zapier.com/hc/en-us/articles/8496309697421-What-is-a-Zap). That model gives no-code users a familiar way to automate predictable sequences, such as adding a new form submission to a CRM and sending a notification. [Zapier Agents](https://zapier.com/agents) uses a separate interface and mental model for work that requires an AI agent to choose tools or decide what to do next. -Sim and Zapier organize larger AI workflows differently. In Sim, you can start with a Mothership prompt, refine the generated graph on the canvas, and access the same workflow through code. Reasoning blocks and regular automation blocks remain visible in one design. +Sim and Zapier organize larger AI workflows differently. In Sim, you can start with a prompt in Chat, refine the generated graph on the canvas, and access the same workflow through code. Reasoning blocks and regular automation blocks remain visible in one design. With Zapier, you may need to decide whether each part belongs in a Zap, an Agent, a Chatbot, or Copilot. [Zapier describes these as products and features in its AI lineup](https://zapier.com/blog/zapier-ai-guide/), and you manage their behavior and handoffs through their respective surfaces. Sim suits technical operators who want natural language generation, visual inspection, and code access to the same agent workflow. @@ -99,7 +100,7 @@ The units do not support a direct one-to-one comparison. A Zapier task, a Zapier | Product | License and hosting | Builder model | Agent depth | Native context | Deployment surfaces | Pricing model | | --- | --- | --- | --- | --- | --- | --- | -| Sim | Apache 2.0. Self-hosted or managed. BYOK. | Mothership, visual blocks, and API or SDK. | Agent reasoning and deterministic logic share one graph. | Tables, Files, and Knowledge Bases sit inside the workspace. | One workflow can run as an API, hosted chat, or MCP server. | Per-user plans plus [credit-based usage](https://www.sim.ai/pricing). | +| Sim | Apache 2.0. Self-hosted or managed. BYOK. | Chat, visual blocks, and API or SDK. | Agent reasoning and deterministic logic share one graph. | Tables, Files, and Knowledge Bases sit inside the workspace. | One workflow can run as an API, hosted chat, or MCP server. | Per-user plans plus [credit-based usage](https://www.sim.ai/pricing). | | Zapier | [Proprietary cloud service without customer-operated self-hosting](https://zapier.com/blog/cloud-vs-self-hosting/). | [Trigger-action Zaps](https://help.zapier.com/hc/en-us/articles/8496309697421-What-is-a-Zap) with [Agents](https://zapier.com/agents) offered separately. | [AI steps extend an automation-first product](https://zapier.com/blog/zapier-ai-guide/). | [Tables](https://help.zapier.com/hc/en-us/articles/9804340895245-Create-tables-and-store-data-with-Zapier-Tables), [Chatbots](https://zapier.com/ai/chatbot), and Copilot operate as separate products. | Zaps, [Agents](https://zapier.com/agents), and [Chatbots](https://help.zapier.com/hc/en-us/articles/21958023866381-Share-and-embed-a-chatbot) cover separate deployment surfaces. | [Task-based Zaps](https://zapier.com/pricing) and [activity-based Agents](https://help.zapier.com/hc/en-us/articles/26559132765325-How-is-Zapier-Agents-usage-measured). | ## Where Zapier is still the better choice diff --git a/apps/sim/content/library/sim-vs-dedicated-chatbot-builders/index.mdx b/apps/sim/content/library/sim-vs-dedicated-chatbot-builders/index.mdx index 12ec9daf468..7e0b11c32d2 100644 --- a/apps/sim/content/library/sim-vs-dedicated-chatbot-builders/index.mdx +++ b/apps/sim/content/library/sim-vs-dedicated-chatbot-builders/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 8 tags: [AI Agents, Chatbots, Customer Support, Sim] ogImage: /library/sim-vs-dedicated-chatbot-builders/cover.jpg -canonical: https://www.sim.ai/library/sim-vs-dedicated-chatbot-builders draft: false faq: - q: "What is the best chatbot builder for conversational AI?" @@ -35,7 +34,7 @@ faq: A dedicated chatbot builder centers its tools on creating and publishing conversational interfaces. These products often include support templates, channel configuration, and agent handoff features. Sim covers a broader use case. It provides an open-source workspace for building AI agents and workflows, with chat available as one deployment surface. The distinction is explored further in this guide to an [AI agent vs. chatbot](https://www.sim.ai/library/ai-agent-vs-chatbot). -You can use Mothership for guided workflow creation or build directly on Sim's visual canvas. Agent blocks manage model interactions, and Knowledge Bases ground responses in selected information. You can publish the resulting workflow as hosted chat or use its agent logic in other applications. +You can use Chat for guided workflow creation or build directly in Sim's workflow builder. Agent blocks manage model interactions, and Knowledge Bases ground responses in selected information. You can publish the resulting workflow as hosted chat or use its agent logic in other applications. Sim and dedicated chatbot builders differ most in model flexibility, integration depth, and deployment options. Model flexibility includes hosted model access, provider choice, and BYOK. Local-model providers such as Ollama use a separate setup path from regular hosted access and BYOK. @@ -62,7 +61,7 @@ Sim fits better when you need to test providers or assign different models to in Botpress is included as a [conversation-first alternative](https://botpress.com/docs/) for support and messaging use cases, but the reviewed material does not establish a reliable feature-by-feature comparison of its model providers, integration coverage, and deployment options with Sim. -Sim supports reusable agent logic that can run as hosted chat or through another supported interface. You can create the workflow with Mothership or the visual canvas, then connect Agent blocks to Knowledge Bases and external tools. Verify Botpress against its current [product documentation](https://botpress.com/docs/) before comparing its conversation-design features with Sim's workflow capabilities. +Sim supports reusable agent logic that can run as hosted chat or through another supported interface. You can create the workflow with Chat or the workflow builder, then connect Agent blocks to Knowledge Bases and external tools. Verify Botpress against its current [product documentation](https://botpress.com/docs/) before comparing its conversation-design features with Sim's workflow capabilities. ## Integration depth and chatbot actions @@ -105,7 +104,7 @@ Sim requires you to define the workflow, configure Agent blocks, and choose how ## Building a chatbot in Sim as one surface of a larger system -Sim lets you build an AI workflow in Mothership or the visual canvas and deploy it as a hosted chatbot. The workflow can later serve an API or operate through MCP without requiring you to recreate its instructions, knowledge, or tool logic. +Sim lets you build an AI workflow in Chat or the workflow builder and deploy it as a hosted chatbot. The workflow can later serve an API or operate through MCP without requiring you to recreate its instructions, knowledge, or tool logic. Agent blocks manage the model interaction and response logic. Knowledge Bases ground answers in your documents, which helps the chatbot retrieve relevant information instead of relying only on a model's general knowledge. After testing the workflow, you can publish it as a hosted chat interface and share it with users. diff --git a/apps/sim/content/library/sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform/index.mdx b/apps/sim/content/library/sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform/index.mdx index ef82ab5f915..708e054b530 100644 --- a/apps/sim/content/library/sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform/index.mdx +++ b/apps/sim/content/library/sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform/index.mdx @@ -1,15 +1,14 @@ --- slug: sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform title: 'Sim vs Dify: Open-Source AI Workspace vs LLM App / RAG Platform' -description: 'A current, evidence-based comparison of Sim and Dify across visual workflow building, RAG, deployment, integrations, licensing, pricing, and team fit.' +description: 'Sim vs Dify head to head: how the two compare on visual workflow building, RAG, self-hosting, integrations, licensing, pricing, and team fit, and when to choose each.' date: 2026-08-05 -updated: 2026-09-29 +updated: 2026-09-30 authors: - andrew readingTime: 12 tags: [Dify, Open Source, AI Agents, RAG, Sim] ogImage: /library/sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform/cover.jpg -canonical: https://www.sim.ai/library/sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform draft: false faq: - q: "What is the difference between Sim and Dify?" @@ -40,14 +39,8 @@ faq: a: "Sim focuses on AI-native agents and workflows, Dify focuses on LLM applications and RAG, and n8n is the broadest general-purpose application automation platform of the three." - q: "Is n8n open source?" a: "n8n is source-available under the Sustainable Use License and Enterprise License, and the Sustainable Use License is not an OSI-approved open-source license." - - q: "What is the best open-source n8n alternative?" - a: "Sim is a strong open-source n8n alternative for AI-native workflows because Sim uses the OSI-approved Apache 2.0 license, although n8n remains a stronger fit for some conventional application-automation use cases." - - q: "What is the best open-source Zapier alternative for AI workflows?" - a: "Sim is a strong open-source Zapier alternative when AI agents, model calls, retrieval, and visual workflow logic are central requirements." - q: "Is Sim free?" a: "Sim can be self-hosted under Apache 2.0 without a software license fee, but infrastructure, model APIs, storage, and other connected services can still create costs." - - q: "Sim vs Gumloop: which should I choose?" - a: "Sim is the clearer choice when Apache 2.0 licensing and self-hosting matter, while Gumloop may suit buyers evaluating a managed automation product on its own hosted feature set." --- ## TL;DR @@ -56,6 +49,8 @@ Sim and Dify overlap as visual platforms for building AI applications, but they Choose Sim when flexible AI automation, an OSI-approved core license, and broader workflow orchestration are priorities. Choose Dify when the central task is managing retrieval-heavy LLM applications. Neither platform is universally superior. +This page compares only these two products. If you are still surveying the field, the ranked list of [Dify alternatives](https://www.sim.ai/library/dify-alternatives) also covers n8n, LangChain and LangGraph, RAGFlow, and Langflow. + _Reviewed September 2026. Product capabilities, hosted pricing, quotas, and license terms should be reconfirmed from the linked first-party sources before purchase or deployment._ ## What is the difference between Sim and Dify? @@ -222,7 +217,7 @@ Use the following evaluation checklist: The better product is the one that satisfies the real deployment and operating constraints with the least avoidable complexity. -For a broader category comparison, see [Best AI Agent Builder in 2026](https://www.sim.ai/library/best-ai-agent-builder-2026); this page remains focused on the Sim-versus-Dify decision. +For a broader view, see the ranked [Dify alternatives](https://www.sim.ai/library/dify-alternatives) or [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026); this page remains focused on the Sim-versus-Dify decision. ## Where can I verify the claims in this comparison? diff --git a/apps/sim/content/library/top-ai-assistants-2026/index.mdx b/apps/sim/content/library/top-ai-assistants-2026/index.mdx index 945a1150df5..7fe6044b774 100644 --- a/apps/sim/content/library/top-ai-assistants-2026/index.mdx +++ b/apps/sim/content/library/top-ai-assistants-2026/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 12 tags: [AI Assistants, AI Agents, Productivity, Sim] ogImage: /library/top-ai-assistants-2026/cover.jpg -canonical: https://www.sim.ai/library/top-ai-assistants-2026 draft: false faq: - q: "What is the difference between an AI chat app and a personal AI assistant?" diff --git a/apps/sim/content/library/what-is-an-agentic-workflow/index.mdx b/apps/sim/content/library/what-is-an-agentic-workflow/index.mdx index cd8e6faa9ba..9407733dc6b 100644 --- a/apps/sim/content/library/what-is-an-agentic-workflow/index.mdx +++ b/apps/sim/content/library/what-is-an-agentic-workflow/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 7 tags: [AI Agents, Workflow Automation, Agentic Workflows, Sim] ogImage: /library/what-is-an-agentic-workflow/cover.jpg -canonical: https://www.sim.ai/library/what-is-an-agentic-workflow draft: false faq: - q: "Is agentic AI the same as automation?" @@ -75,7 +74,7 @@ A reasoning-native graph treats the model as a node that can choose tools and re ## How Sim structures agentic and deterministic blocks in one graph -[Sim](https://sim.ai) provides one example of the hybrid pattern. Its visual workflow graph places Agent blocks alongside deterministic blocks, so model reasoning participates directly in execution rather than sitting inside a fixed automation step. +[Sim](https://www.sim.ai) provides one example of the hybrid pattern. Its visual workflow graph places Agent blocks alongside deterministic blocks, so model reasoning participates directly in execution rather than sitting inside a fixed automation step. Agent blocks reason over available context and choose tools during a run. Builders can constrain that discretion by selecting the tools an Agent block may call and defining the structured output that later blocks receive. diff --git a/apps/sim/content/library/what-is-an-ai-agent-definition-how-it-works-and-examples/index.mdx b/apps/sim/content/library/what-is-an-ai-agent-definition-how-it-works-and-examples/index.mdx index 947b77c130b..31ca6cf43ec 100644 --- a/apps/sim/content/library/what-is-an-ai-agent-definition-how-it-works-and-examples/index.mdx +++ b/apps/sim/content/library/what-is-an-ai-agent-definition-how-it-works-and-examples/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 8 tags: [AI Agents, LLMs, Automation, Sim] ogImage: /library/what-is-an-ai-agent-definition-how-it-works-and-examples/cover.jpg -canonical: https://www.sim.ai/library/what-is-an-ai-agent-definition-how-it-works-and-examples draft: false faq: - q: "Does an AI agent need an LLM?" @@ -108,4 +107,4 @@ An AI agent combines a model that can choose the next step with connected capabi Start with one narrow workflow and clear permissions. Choose a task with an observable loop, such as routing support tickets or proposing meeting slots. Once that task works reliably, you can add tools or grant the agent more autonomy. -If you are deciding whether fixed automation is enough, [AI Agents vs RPA: When to Use Each for Enterprise Automation](https://www.sim.ai/library/automation-anywhere-alternative) explains where rule-based automation remains the better fit. +If you are deciding whether fixed automation is enough, [AI Agents vs RPA: When to Use Each for Enterprise Automation](https://www.sim.ai/library/ai-agents-vs-rpa) explains where rule-based automation remains the better fit. diff --git a/apps/sim/content/library/what-is-an-mcp-server/index.mdx b/apps/sim/content/library/what-is-an-mcp-server/index.mdx index dd1ac8b60bc..45d33fa684e 100644 --- a/apps/sim/content/library/what-is-an-mcp-server/index.mdx +++ b/apps/sim/content/library/what-is-an-mcp-server/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 7 tags: [MCP, AI Agents, Model Context Protocol, Sim] ogImage: /library/what-is-an-mcp-server/cover.jpg -canonical: https://www.sim.ai/library/what-is-an-mcp-server draft: false faq: - q: "What does MCP server mean?" @@ -107,4 +106,4 @@ You can also publish a Sim workflow as a callable MCP tool. External MCP clients ## Getting started with MCP and Sim -Use the [Sim workflow builder](https://sim.ai) to connect an external MCP server or publish a Sim workflow as an MCP tool. If you are still evaluating how to deploy your workflows, compare [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms). +Use the [Sim workflow builder](https://www.sim.ai) to connect an external MCP server or publish a Sim workflow as an MCP tool. If you are still evaluating how to deploy your workflows, compare [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms). diff --git a/apps/sim/content/library/what-is-human-in-the-loop-in-ai-agents/index.mdx b/apps/sim/content/library/what-is-human-in-the-loop-in-ai-agents/index.mdx index 096e9756de5..527c4b630cd 100644 --- a/apps/sim/content/library/what-is-human-in-the-loop-in-ai-agents/index.mdx +++ b/apps/sim/content/library/what-is-human-in-the-loop-in-ai-agents/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 7 tags: [AI Agents, Human in the Loop, Workflow Automation, Sim] ogImage: /library/what-is-human-in-the-loop-in-ai-agents/cover.jpg -canonical: https://www.sim.ai/library/what-is-human-in-the-loop-in-ai-agents draft: false faq: - q: "How does agentic HITL differ from HITL in model training?" diff --git a/apps/sim/content/library/what-is-retrieval-augmented-generation/index.mdx b/apps/sim/content/library/what-is-retrieval-augmented-generation/index.mdx index 5b65b10f082..4cb4608b29d 100644 --- a/apps/sim/content/library/what-is-retrieval-augmented-generation/index.mdx +++ b/apps/sim/content/library/what-is-retrieval-augmented-generation/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 6 tags: [RAG, AI Agents, Knowledge Bases, Sim] ogImage: /library/what-is-retrieval-augmented-generation/cover.jpg -canonical: https://www.sim.ai/library/what-is-retrieval-augmented-generation draft: false faq: - q: "Is RAG a type of fine-tuning?" @@ -79,7 +78,7 @@ An agent can break a task into steps and call retrieval or [other tools exposed For example, an agent reviewing a contract might retrieve the standard cancellation policy first. A clause in the contract could then prompt a second search for an account-specific amendment. A fixed retrieve-once pipeline would not issue the second query because the need for it appears only after the first document has been read. -[Sim's native Knowledge Bases](https://sim.ai) make retrieval a workspace resource that an Agent block can call during reasoning. Knowledge bases sit alongside workflow logic and other tools, rather than requiring a separate vector-store integration built around one LLM application. [Dify can suit application-centered workflows](https://docs.dify.ai/en/cloud/use-dify/knowledge/integrate-knowledge-within-application), while Sim places retrieval inside an agent-native workspace so multiple workflow steps can query the same Knowledge Base. See the [Sim and Dify comparison](https://www.sim.ai/library/sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform) for more context. +[Sim's native Knowledge Bases](https://www.sim.ai) make retrieval a workspace resource that an Agent block can call during reasoning. Knowledge bases sit alongside workflow logic and other tools, rather than requiring a separate vector-store integration built around one LLM application. [Dify can suit application-centered workflows](https://docs.dify.ai/en/cloud/use-dify/knowledge/integrate-knowledge-within-application), while Sim places retrieval inside an agent-native workspace so multiple workflow steps can query the same Knowledge Base. See the [Sim and Dify comparison](https://www.sim.ai/library/sim-vs-dify-open-source-ai-workspace-vs-llm-app-rag-platform) for more context. Sim's [Apache 2.0 repository](https://github.com/simstudioai/sim) also supports self-hosting, which gives you control over the agent runtime and retrieval infrastructure. Agentic RAG still costs more than a single retrieval pass because every retry adds model work and latency. You can limit reasoning depth, cache common searches, and rerank retrieved passages when response time or usage cost requires tighter bounds. @@ -99,4 +98,4 @@ Before deployment, define targets for retrieval accuracy and response time, then Use retrieval as a workspace capability when an agent needs private or current information during a task. Sim's native Knowledge Bases give Agent blocks access to grounded context during a workflow, without requiring a separate vector-store integration tied to one chat application. -You can create the workflow with Mothership, inspect and edit its logic in the visual builder, or connect it through the API. [Explore how to build a RAG-grounded agent in Sim](https://sim.ai). +You can create the workflow with Chat, inspect and edit its logic in the visual builder, or connect it through the API. [Explore how to build a RAG-grounded agent in Sim](https://www.sim.ai). diff --git a/apps/sim/content/library/why-no-code-ai-agents-need-live-web-access/index.mdx b/apps/sim/content/library/why-no-code-ai-agents-need-live-web-access/index.mdx index 4d870a63143..503c4bd299c 100644 --- a/apps/sim/content/library/why-no-code-ai-agents-need-live-web-access/index.mdx +++ b/apps/sim/content/library/why-no-code-ai-agents-need-live-web-access/index.mdx @@ -9,7 +9,6 @@ authors: readingTime: 9 tags: [AI Agents, No-Code, Web Automation, TinyFish, Sim] ogImage: /library/why-no-code-ai-agents-need-live-web-access/cover.jpg -canonical: https://www.sim.ai/library/why-no-code-ai-agents-need-live-web-access draft: false faq: - q: "What happens when a site rate-limits or blocks a request?" @@ -19,7 +18,7 @@ faq: - q: "How should I handle portal credentials in Vault?" a: "Store credentials in Vault and reference the Vault item from the TinyFish block instead of placing secrets in prompts or workflow fields. Scope workflow and workspace access to the people and runs that need those credentials. Review your platform’s access and retention controls before using production accounts." - q: "Do I need to migrate off my current no-code tool?" - a: "No. TinyFish provides a web access layer through an API key. You can call it from an existing builder through a native integration, an HTTP block, or custom code. The Sim integration (https://sim.ai/integrations/tinyfish) provides one working example." + a: "No. TinyFish provides a web access layer through an API key. You can call it from an existing builder through a native integration, an HTTP block, or custom code. The Sim integration (https://www.sim.ai/integrations/tinyfish) provides one working example." --- ## TL;DR @@ -71,7 +70,7 @@ Estimate spend by running a representative workflow against the sites you expect ## Wiring TinyFish into Sim -Sim provides a working example of this two-layer setup. Its [TinyFish integration](https://sim.ai/integrations/tinyfish) adds live web capabilities through one workflow block. If you are new to visual agents, start with [how to create an AI agent](https://www.sim.ai/library/how-to-create-an-ai-agent) or compare the [best no-code AI agent builders](https://www.sim.ai/library/best-no-code-ai-agent-builders-2026). +Sim provides a working example of this two-layer setup. Its [TinyFish integration](https://www.sim.ai/integrations/tinyfish) adds live web capabilities through one workflow block. If you are new to visual agents, start with [how to create an AI agent](https://www.sim.ai/library/how-to-create-an-ai-agent) or compare the [best no-code AI agent builders](https://www.sim.ai/library/best-no-code-ai-agent-builders-2026). The block exposes nine tools that cover agent runs, web retrieval, Vault items, and browser profiles. Run Agent and Start Agent Run launch work, while Get Run, Cancel Run, and List Runs manage execution. Search finds current web results, and Fetch URLs retrieves page content. List Vault Items and List Browser Profiles expose the stored resources available to the workflow. @@ -115,4 +114,4 @@ You can swap the visual builder without rebuilding web access or add TinyFish to ## Where to start -Add a dedicated live-web layer when your workflow needs current data, authenticated sessions, or browser interaction. Try the TinyFish block in [Sim](https://sim.ai) for a visual setup, or connect a [TinyFish API key](https://docs.tinyfish.ai/) directly to your existing agent framework. Choose Search, Fetch, Browser, or Agent based on the pages your workflow must reach. +Add a dedicated live-web layer when your workflow needs current data, authenticated sessions, or browser interaction. Try the TinyFish block in [Sim](https://www.sim.ai) for a visual setup, or connect a [TinyFish API key](https://docs.tinyfish.ai/) directly to your existing agent framework. Choose Search, Fetch, Browser, or Agent based on the pages your workflow must reach. diff --git a/apps/sim/ee/whitelabeling/index.ts b/apps/sim/ee/whitelabeling/index.ts index 6cd22b7b1c3..d0344b915c1 100644 --- a/apps/sim/ee/whitelabeling/index.ts +++ b/apps/sim/ee/whitelabeling/index.ts @@ -2,5 +2,5 @@ export type { OrganizationWhitelabelSettings } from '@/lib/branding/types' export type { BrandConfig, ThemeColors } from './branding' export { getBrandConfig, useBrandConfig } from './branding' export { generateThemeCSS } from './inject-theme' -export { generateBrandedMetadata, generateStructuredData } from './metadata' +export { generateBrandedMetadata } from './metadata' export { generateOrgThemeCSS, mergeOrgBrandConfig } from './org-branding-utils' diff --git a/apps/sim/ee/whitelabeling/metadata.ts b/apps/sim/ee/whitelabeling/metadata.ts index 72ce29b3242..bf1d0089eb0 100644 --- a/apps/sim/ee/whitelabeling/metadata.ts +++ b/apps/sim/ee/whitelabeling/metadata.ts @@ -1,16 +1,21 @@ import type { Metadata } from 'next' -import { getBaseUrl, SITE_URL } from '@/lib/core/utils/urls' +import { getBaseUrl } from '@/lib/core/utils/urls' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' import { getBrandConfig } from '@/ee/whitelabeling/branding' /** - * Generate dynamic metadata based on brand configuration + * Generate dynamic metadata based on brand configuration. + * + * Deliberately sets no `alternates.canonical` or `openGraph.url`: every route + * inherits this, so a fixed value would make each page claim the home page as + * its canonical. Landing pages emit their own canonical URLs. */ export function generateBrandedMetadata(override: Partial = {}): Metadata { const brand = getBrandConfig() const defaultTitle = brand.name - const summaryFull = `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect 1,000+ integrations and every major LLM to create agents that automate real work — visually, conversationally, or with code. Trusted by over 100,000 builders — from startups to Fortune 500 companies. SOC2 compliant.` - const summaryShort = `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect 1,000+ integrations and every major LLM to create agents that automate real work.` + const summaryFull = `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect ${INTEGRATION_COUNT_LABEL} integrations and every major LLM to create agents that automate real work — visually, conversationally, or with code. Trusted by over 100,000 builders — from startups to Fortune 500 companies. SOC2 compliant.` + const summaryShort = `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect ${INTEGRATION_COUNT_LABEL} integrations and every major LLM to create agents that automate real work.` return { title: { @@ -46,12 +51,6 @@ export function generateBrandedMetadata(override: Partial = {}): Metad creator: brand.name, publisher: brand.name, metadataBase: new URL(getBaseUrl()), - alternates: { - canonical: '/', - languages: { - 'en-US': '/', - }, - }, robots: { index: true, follow: true, @@ -66,7 +65,6 @@ export function generateBrandedMetadata(override: Partial = {}): Metad openGraph: { type: 'website', locale: 'en_US', - url: getBaseUrl(), title: defaultTitle, description: summaryFull, siteName: brand.name, @@ -126,41 +124,3 @@ export function generateBrandedMetadata(override: Partial = {}): Metad ...override, } } - -/** - * Generate static structured data for SEO - */ -export function generateStructuredData() { - return { - '@context': 'https://schema.org', - '@type': 'SoftwareApplication', - name: 'Sim', - description: - 'Sim is the open-source AI workspace where teams build, deploy, and manage AI agents. Connect 1,000+ integrations and every major LLM to create agents that automate real work. Trusted by over 100,000 builders. SOC2 compliant.', - url: getBaseUrl(), - applicationCategory: 'BusinessApplication', - operatingSystem: 'Web', - applicationSubCategory: 'AIWorkspace', - areaServed: 'Worldwide', - availableLanguage: ['en'], - offers: { - '@type': 'Offer', - category: 'SaaS', - }, - creator: { - '@type': 'Organization', - name: 'Sim', - url: SITE_URL, - }, - featureList: [ - 'AI Workspace for Teams', - 'Chat — Natural Language Agent Creation', - 'Visual Workflow Builder', - '1,000+ Integrations', - 'LLM Orchestration', - 'Knowledge Base Creation', - 'Table Creation', - 'Document Creation', - ], - } -} diff --git a/apps/sim/lib/blog/registry.ts b/apps/sim/lib/blog/registry.ts index 71235ae66b4..c5701c1ff07 100644 --- a/apps/sim/lib/blog/registry.ts +++ b/apps/sim/lib/blog/registry.ts @@ -1,4 +1,5 @@ import path from 'path' +import { BLOG_SECTION } from '@/lib/blog/seo' import { createContentRegistry } from '@/lib/content/registry-factory' const BLOG_DIR = path.join(process.cwd(), 'content', 'blog') @@ -14,6 +15,7 @@ const BLOG_COMPONENT_LOADERS = { const blogRegistry = createContentRegistry({ contentDir: BLOG_DIR, authorsDir: AUTHORS_DIR, + basePath: BLOG_SECTION.basePath, componentLoaders: BLOG_COMPONENT_LOADERS, }) diff --git a/apps/sim/lib/compare/data/competitors/langchain.ts b/apps/sim/lib/compare/data/competitors/langchain.ts index 7c5b3e90b19..929729d85b8 100644 --- a/apps/sim/lib/compare/data/competitors/langchain.ts +++ b/apps/sim/lib/compare/data/competitors/langchain.ts @@ -5,6 +5,7 @@ import type { CompetitorProfile } from '@/lib/compare/data/types' export const langchainProfile: CompetitorProfile = { id: 'langchain', name: 'LangChain', + mentions: ['LangChain', 'LangGraph'], website: 'https://www.langchain.com', isWorkflowBuilder: false, brand: { diff --git a/apps/sim/lib/compare/data/competitors/make.ts b/apps/sim/lib/compare/data/competitors/make.ts index d85f6158560..ac8e05c39c2 100644 --- a/apps/sim/lib/compare/data/competitors/make.ts +++ b/apps/sim/lib/compare/data/competitors/make.ts @@ -5,6 +5,7 @@ import type { CompetitorProfile } from '@/lib/compare/data/types' export const makeProfile: CompetitorProfile = { id: 'make', name: 'Make', + mentions: ['Make.com', 'Integromat'], website: 'https://www.make.com', brand: { icon: MakeIcon, diff --git a/apps/sim/lib/compare/data/competitors/microsoft-copilot.ts b/apps/sim/lib/compare/data/competitors/microsoft-copilot.ts index 716d2fe8b9e..89de161a251 100644 --- a/apps/sim/lib/compare/data/competitors/microsoft-copilot.ts +++ b/apps/sim/lib/compare/data/competitors/microsoft-copilot.ts @@ -5,6 +5,7 @@ import type { CompetitorProfile } from '@/lib/compare/data/types' export const microsoftCopilotProfile: CompetitorProfile = { id: 'microsoft-copilot', name: 'Microsoft Copilot Studio', + mentions: ['Copilot Studio'], website: 'https://www.microsoft.com/en-us/microsoft-copilot-studio', brand: { icon: MicrosoftCopilotIcon, diff --git a/apps/sim/lib/compare/data/competitors/openai-agentkit.ts b/apps/sim/lib/compare/data/competitors/openai-agentkit.ts index 35519f829ca..c621e8a4d10 100644 --- a/apps/sim/lib/compare/data/competitors/openai-agentkit.ts +++ b/apps/sim/lib/compare/data/competitors/openai-agentkit.ts @@ -5,6 +5,7 @@ import type { CompetitorProfile } from '@/lib/compare/data/types' export const openaiAgentkitProfile: CompetitorProfile = { id: 'openai-agentkit', name: 'OpenAI AgentKit', + mentions: ['AgentKit'], website: 'https://openai.com/index/introducing-agentkit/', brand: { icon: OpenAIIcon, diff --git a/apps/sim/lib/compare/data/competitors/power-automate.ts b/apps/sim/lib/compare/data/competitors/power-automate.ts index dc080a88d1d..c0c9102d0be 100644 --- a/apps/sim/lib/compare/data/competitors/power-automate.ts +++ b/apps/sim/lib/compare/data/competitors/power-automate.ts @@ -5,6 +5,7 @@ import type { CompetitorProfile } from '@/lib/compare/data/types' export const powerAutomateProfile: CompetitorProfile = { id: 'power-automate', name: 'Microsoft Power Automate', + mentions: ['Power Automate'], website: 'https://www.microsoft.com/en-us/power-platform/products/power-automate', brand: { icon: MicrosoftIcon, diff --git a/apps/sim/lib/compare/data/competitors/stackai.ts b/apps/sim/lib/compare/data/competitors/stackai.ts index 31f4adff439..86c78378054 100644 --- a/apps/sim/lib/compare/data/competitors/stackai.ts +++ b/apps/sim/lib/compare/data/competitors/stackai.ts @@ -5,6 +5,7 @@ import type { CompetitorProfile } from '@/lib/compare/data/types' export const stackaiProfile: CompetitorProfile = { id: 'stack-ai', name: 'StackAI', + mentions: ['StackAI', 'Stack AI'], website: 'https://www.stackai.com', brand: { icon: StackAIIcon, diff --git a/apps/sim/lib/compare/data/sim.ts b/apps/sim/lib/compare/data/sim.ts index 044f0887581..ac649c26161 100644 --- a/apps/sim/lib/compare/data/sim.ts +++ b/apps/sim/lib/compare/data/sim.ts @@ -1,4 +1,6 @@ import type { CompetitorProfile } from '@/lib/compare/data/types' +import { SITE_URL } from '@/lib/core/utils/urls' +import { INTEGRATION_COUNT_LABEL } from '@/lib/landing/constants' /** * Sim's own profile, for use as the constant left-hand column on every @@ -11,9 +13,8 @@ import type { CompetitorProfile } from '@/lib/compare/data/types' export const simProfile: CompetitorProfile = { id: 'sim', name: 'Sim', - website: 'https://sim.ai', - oneLiner: - 'Sim is the open-source AI workspace where teams build, deploy, and manage AI agents, connecting 1,000+ integrations and every major LLM to automate real work visually, conversationally, or with code.', + website: SITE_URL, + oneLiner: `Sim is the open-source AI workspace where teams build, deploy, and manage AI agents, connecting ${INTEGRATION_COUNT_LABEL} integrations and every major LLM to automate real work visually, conversationally, or with code.`, standoutFeatures: [ { title: 'AI Copilot / Chat agent-building surface', @@ -182,7 +183,7 @@ export const simProfile: CompetitorProfile = { asOf: '2026-07-08', }, { - url: 'https://www.sim.ai/pricing', + url: `${SITE_URL}/pricing`, label: 'Sim Pricing Page', asOf: '2026-07-02', }, @@ -622,11 +623,9 @@ export const simProfile: CompetitorProfile = { }, integrations: { integrationCount: { - value: - '1,000+ integrations counting individual API actions, built from 266 first-party blocks and roughly 3,900 underlying tool actions', - detail: - 'Sim\'s landing page cites the "1,000+ integrations" figure; the block/tool-action counts are the same integration surface measured at a different level of granularity.', - shortValue: '1,000+ integrations (266 blocks, ~3,900 tool actions)', + value: `${INTEGRATION_COUNT_LABEL} integrations counting individual API actions, built from 266 first-party blocks and roughly 3,900 underlying tool actions`, + detail: `Sim's landing page cites the "${INTEGRATION_COUNT_LABEL} integrations" figure; the block/tool-action counts are the same integration surface measured at a different level of granularity.`, + shortValue: `${INTEGRATION_COUNT_LABEL} integrations (266 blocks, ~3,900 tool actions)`, confidence: 'verified', sources: [ { @@ -640,7 +639,7 @@ export const simProfile: CompetitorProfile = { asOf: '2026-07-02', }, { - url: 'https://sim.ai', + url: SITE_URL, label: 'Sim Landing Page', asOf: '2026-07-02', }, @@ -772,7 +771,7 @@ export const simProfile: CompetitorProfile = { confidence: 'verified', sources: [ { - url: 'https://sim.ai/pricing', + url: `${SITE_URL}/pricing`, label: 'Sim Pricing', asOf: '2026-07-02', }, @@ -784,7 +783,7 @@ export const simProfile: CompetitorProfile = { confidence: 'verified', sources: [ { - url: 'https://sim.ai/pricing', + url: `${SITE_URL}/pricing`, label: 'Sim Pricing', asOf: '2026-07-02', }, @@ -797,7 +796,7 @@ export const simProfile: CompetitorProfile = { confidence: 'verified', sources: [ { - url: 'https://www.sim.ai/pricing', + url: `${SITE_URL}/pricing`, label: 'Sim Pricing', asOf: '2026-08-26', }, @@ -839,12 +838,12 @@ export const simProfile: CompetitorProfile = { confidence: 'verified', sources: [ { - url: 'https://sim.ai', + url: SITE_URL, label: 'Sim Landing Page', asOf: '2026-07-02', }, { - url: 'https://sim.ai/enterprise', + url: `${SITE_URL}/enterprise`, label: 'Sim Enterprise Page', asOf: '2026-07-02', }, @@ -920,7 +919,7 @@ export const simProfile: CompetitorProfile = { confidence: 'estimated', sources: [ { - url: 'https://sim.ai/enterprise', + url: `${SITE_URL}/enterprise`, label: 'Sim Enterprise Page', asOf: '2026-07-02', }, @@ -1213,7 +1212,7 @@ export const simProfile: CompetitorProfile = { confidence: 'verified', sources: [ { - url: 'https://www.sim.ai/pricing', + url: `${SITE_URL}/pricing`, label: 'Sim Pricing Page', asOf: '2026-07-02', }, @@ -1226,7 +1225,7 @@ export const simProfile: CompetitorProfile = { confidence: 'verified', sources: [ { - url: 'https://www.sim.ai/pricing', + url: `${SITE_URL}/pricing`, label: 'Sim Pricing Page', asOf: '2026-07-08', }, @@ -1238,7 +1237,7 @@ export const simProfile: CompetitorProfile = { confidence: 'estimated', sources: [ { - url: 'https://sim.ai', + url: SITE_URL, label: 'Sim Landing Page', asOf: '2026-07-02', }, diff --git a/apps/sim/lib/compare/data/types.ts b/apps/sim/lib/compare/data/types.ts index 0889269badc..88f9db4d59a 100644 --- a/apps/sim/lib/compare/data/types.ts +++ b/apps/sim/lib/compare/data/types.ts @@ -245,6 +245,14 @@ export interface CompetitorProfile { * can ask a category-clarifying question instead of a peer feature-gap one. */ isWorkflowBuilder?: boolean + /** + * Phrases that identify this competitor in library article titles and tags, + * matched case-sensitively on word boundaries; descriptions also match the + * bare `name`. Drives the links between comparison pages and library + * articles. Defaults to `[name]`; set it when articles use another name + * ("AgentKit") or when the bare name is a common Title Case word ("Make"). + */ + mentions?: string[] /** Logo icon and brand colors, when available. */ brand?: CompetitorBrand /** Free-text list of standout features, each independently sourced. */ diff --git a/apps/sim/lib/content/registry-factory.test.ts b/apps/sim/lib/content/registry-factory.test.ts index 9647af957ec..91d3640e553 100644 --- a/apps/sim/lib/content/registry-factory.test.ts +++ b/apps/sim/lib/content/registry-factory.test.ts @@ -27,7 +27,6 @@ description: The registry still serves posts when the native binary is missing. date: 2026-08-10 authors: [waleed] ogImage: /blog/missing-og.png -canonical: https://sim.ai/blog/sharp-is-unavailable --- Body copy. @@ -51,7 +50,7 @@ afterAll(async () => { describe('createContentRegistry without a loadable sharp', () => { it('still lists posts, omitting only the OG dimensions', async () => { - const registry = createContentRegistry({ contentDir, authorsDir }) + const registry = createContentRegistry({ contentDir, authorsDir, basePath: '/blog' }) const posts = await registry.getAllPostMeta() diff --git a/apps/sim/lib/content/registry-factory.ts b/apps/sim/lib/content/registry-factory.ts index ee659558d36..120872ea73c 100644 --- a/apps/sim/lib/content/registry-factory.ts +++ b/apps/sim/lib/content/registry-factory.ts @@ -12,6 +12,7 @@ import { mdxComponents } from '@/lib/content/mdx' import type { Author, ContentMeta, ContentPost, TagWithCount } from '@/lib/content/schema' import { AuthorSchema, ContentFrontmatterSchema } from '@/lib/content/schema' import { byDateDesc, ensureContentDirs, toIsoDate } from '@/lib/content/utils' +import { SITE_URL } from '@/lib/core/utils/urls' const logger = createLogger('ContentRegistry') @@ -26,6 +27,8 @@ export interface ContentRegistryConfig { contentDir: string /** Directory holding one JSON file per author, shared across sections. */ authorsDir: string + /** Path the section is served under (e.g. `/library`); each post's canonical URL is derived from it. */ + basePath: string /** Per-slug custom MDX component overrides, merged over the base `mdxComponents` map. */ componentLoaders?: ContentComponentLoaders } @@ -33,6 +36,8 @@ export interface ContentRegistryConfig { export interface ContentRegistry { getAllPostMeta: () => Promise getPostBySlug: (slug: string) => Promise + /** Raw markdown body (frontmatter stripped) of a published post, or null if none. */ + getPostSource: (slug: string) => Promise getAllTags: () => Promise getRelatedPosts: (slug: string, limit?: number) => Promise getNavPosts: () => Promise[]> @@ -84,10 +89,10 @@ async function loadAuthorsForDir(authorsDir: string): Promise> = {} - let cachedMeta: ContentMeta[] | null = null + let metaPromise: Promise | null = null async function loadAuthors(): Promise> { return loadAuthorsForDir(authorsDir) @@ -131,10 +136,16 @@ export function createContentRegistry(config: ContentRegistryConfig): ContentReg } } - async function scanFrontmatters(): Promise { - if (cachedMeta) { - return cachedMeta - } + /** Shares one in-flight scan across concurrent callers (e.g. parallel static renders). */ + function scanFrontmatters(): Promise { + metaPromise ??= readAllFrontmatters().catch((error) => { + metaPromise = null + throw error + }) + return metaPromise + } + + async function readAllFrontmatters(): Promise { await ensureContentDirs(contentDir, authorsDir) const entries = await fs.readdir(contentDir).catch(() => []) const authorsMap = await loadAuthors() @@ -175,7 +186,7 @@ export function createContentRegistry(config: ContentRegistryConfig): ContentReg ogImage: fm.ogImage, ogImageWidth: ogImageDimensions?.width, ogImageHeight: ogImageDimensions?.height, - canonical: fm.canonical, + canonical: `${SITE_URL}${basePath}/${fm.slug}`, ogAlt: fm.ogAlt, about: fm.about, timeRequired: fm.timeRequired, @@ -187,8 +198,7 @@ export function createContentRegistry(config: ContentRegistryConfig): ContentReg } }) ) - cachedMeta = results.filter((result): result is ContentMeta => result !== null).sort(byDateDesc) - return cachedMeta + return results.filter((result): result is ContentMeta => result !== null).sort(byDateDesc) } async function getAllPostMeta(): Promise { @@ -247,6 +257,13 @@ export function createContentRegistry(config: ContentRegistryConfig): ContentReg } } + async function getPostSource(slug: string): Promise { + const published = await getAllPostMeta() + if (!published.some((m) => m.slug === slug)) return null + const raw = await fs.readFile(path.join(contentDir, slug, 'index.mdx'), 'utf-8') + return matter(raw).content + } + async function getPostBySlug(slug: string): Promise { const meta = await scanFrontmatters() const found = meta.find((m) => m.slug === slug) @@ -306,7 +323,7 @@ export function createContentRegistry(config: ContentRegistryConfig): ContentReg } function invalidateCaches() { - cachedMeta = null + metaPromise = null authorsCacheByDir.delete(authorsDir) Object.keys(postComponentsRegistry).forEach((key) => delete postComponentsRegistry[key]) } @@ -314,6 +331,7 @@ export function createContentRegistry(config: ContentRegistryConfig): ContentReg return { getAllPostMeta, getPostBySlug, + getPostSource, getAllTags, getRelatedPosts, getNavPosts, diff --git a/apps/sim/lib/content/schema.ts b/apps/sim/lib/content/schema.ts index 7768263b935..c7737f6bcfb 100644 --- a/apps/sim/lib/content/schema.ts +++ b/apps/sim/lib/content/schema.ts @@ -39,7 +39,6 @@ export const ContentFrontmatterSchema = z }) ) .optional(), - canonical: z.string().url(), draft: z.boolean().default(false), featured: z.boolean().default(false), /** diff --git a/apps/sim/lib/core/utils/urls.ts b/apps/sim/lib/core/utils/urls.ts index 08d6a9c7673..6cea8fdd024 100644 --- a/apps/sim/lib/core/utils/urls.ts +++ b/apps/sim/lib/core/utils/urls.ts @@ -1,13 +1,20 @@ import { isLoopbackHostname } from '@sim/security/hostnames' +import { SIM_SITE_URL } from '@sim/utils/site' import { env, getEnv } from '@/lib/core/config/env' import { isProd } from '@/lib/core/config/env-flags' /** Canonical base URL for the public-facing marketing site. No trailing slash. */ -export const SITE_URL = 'https://www.sim.ai' +export const SITE_URL = SIM_SITE_URL /** Host of the canonical marketing site, e.g. `www.sim.ai`. */ export const CANONICAL_SITE_HOST = new URL(SITE_URL).host +/** Resolves a site-relative href against {@link SITE_URL}; absolute URLs pass through. */ +export function toSiteUrl(href: string): string { + if (hasHttpProtocol(href)) return href + return href === '/' ? SITE_URL : `${SITE_URL}${href}` +} + function hasHttpProtocol(url: string): boolean { return /^https?:\/\//i.test(url) } diff --git a/apps/sim/lib/customers/registry.ts b/apps/sim/lib/customers/registry.ts index 86c5cddcbc0..1d7e8b32359 100644 --- a/apps/sim/lib/customers/registry.ts +++ b/apps/sim/lib/customers/registry.ts @@ -1,5 +1,6 @@ import path from 'path' import { createContentRegistry } from '@/lib/content/registry-factory' +import { CUSTOMER_SECTION } from '@/lib/customers/data' const CUSTOMERS_DIR = path.join(process.cwd(), 'content', 'customers') const AUTHORS_DIR = path.join(process.cwd(), 'content', 'authors') @@ -7,6 +8,7 @@ const AUTHORS_DIR = path.join(process.cwd(), 'content', 'authors') const customersRegistry = createContentRegistry({ contentDir: CUSTOMERS_DIR, authorsDir: AUTHORS_DIR, + basePath: CUSTOMER_SECTION.basePath, }) /** Published stories only, suitable for public collections and sitemap entries. */ @@ -14,3 +16,4 @@ export const getAllCustomerStoryMeta = customersRegistry.getAllPostMeta /** Includes draft stories so the design preview can render with noindex metadata. */ export const getCustomerStoryBySlug = customersRegistry.getPostBySlug +export const getCustomerStorySource = customersRegistry.getPostSource diff --git a/apps/sim/lib/help-links.ts b/apps/sim/lib/help-links.ts index 5e652a453cd..7485d51a733 100644 --- a/apps/sim/lib/help-links.ts +++ b/apps/sim/lib/help-links.ts @@ -1,5 +1,7 @@ +import { SIM_DOCS_URL } from '@sim/utils/site' + /** Destinations the sidebar help menus open. */ -export const DOCS_URL = 'https://docs.sim.ai' as const +export const DOCS_URL = SIM_DOCS_URL export const SLACK_COMMUNITY_URL = 'https://join.slack.com/t/sim-ott9864/shared_invite/zt-43lp8tc5v-0qrrqHGBKUsvQlpoouH~TA' as const diff --git a/apps/sim/lib/landing/constants.ts b/apps/sim/lib/landing/constants.ts new file mode 100644 index 00000000000..d80207bcae0 --- /dev/null +++ b/apps/sim/lib/landing/constants.ts @@ -0,0 +1,7 @@ +/** + * Sim's public integration-count claim, as written in marketing copy, metadata, + * llms.txt, and JSON-LD ("Connect 1,000+ integrations"). It counts individual + * integration actions, so it is intentionally not derived from the block + * catalog length. Every surface reads it from here so the claim never drifts. + */ +export const INTEGRATION_COUNT_LABEL = '1,000+' diff --git a/apps/sim/lib/library/registry.ts b/apps/sim/lib/library/registry.ts index 1e1fa79bba1..0809aadc172 100644 --- a/apps/sim/lib/library/registry.ts +++ b/apps/sim/lib/library/registry.ts @@ -1,5 +1,6 @@ import path from 'path' import { createContentRegistry } from '@/lib/content/registry-factory' +import { LIBRARY_SECTION } from '@/lib/library/seo' const LIBRARY_DIR = path.join(process.cwd(), 'content', 'library') const AUTHORS_DIR = path.join(process.cwd(), 'content', 'authors') @@ -7,9 +8,11 @@ const AUTHORS_DIR = path.join(process.cwd(), 'content', 'authors') const libraryRegistry = createContentRegistry({ contentDir: LIBRARY_DIR, authorsDir: AUTHORS_DIR, + basePath: LIBRARY_SECTION.basePath, }) export const getAllPostMeta = libraryRegistry.getAllPostMeta export const getPostBySlug = libraryRegistry.getPostBySlug +export const getPostSource = libraryRegistry.getPostSource export const getAllTags = libraryRegistry.getAllTags export const getRelatedPosts = libraryRegistry.getRelatedPosts diff --git a/apps/sim/lib/navigation/paths.ts b/apps/sim/lib/navigation/paths.ts index 6d194cb3548..caa14610c32 100644 --- a/apps/sim/lib/navigation/paths.ts +++ b/apps/sim/lib/navigation/paths.ts @@ -64,3 +64,30 @@ export function isAppSurfacePath(pathname: string): boolean { isPathOrDescendant(pathname, ORGANIZATIONS_PATH) ) } + +/** + * Non-app routes that must never appear in search results: deployed chats, + * paused-run resume links, invitations, unsubscribe links, shared files, the + * legacy `/w` redirects, the design playground, and account, self-host, and + * upgrade utility pages. + */ +const NOINDEX_PATH_ROOTS = [ + '/chat', + '/resume', + '/invite', + '/unsubscribe', + '/f', + '/w', + '/playground', + '/account', + '/selfhost', + '/upgrade', +] as const + +/** Whether a pathname is an app or utility surface that search engines must not index. */ +export function isNoindexPath(pathname: string): boolean { + return ( + isAppSurfacePath(pathname) || + NOINDEX_PATH_ROOTS.some((root) => isPathOrDescendant(pathname, root)) + ) +} diff --git a/apps/sim/next.config.ts b/apps/sim/next.config.ts index 55521e16dce..dd84738690f 100644 --- a/apps/sim/next.config.ts +++ b/apps/sim/next.config.ts @@ -1,4 +1,5 @@ import path from 'node:path' +import { SIM_SITE_URL } from '@sim/utils/site' import type { NextConfig } from 'next' import { env, isTruthy } from './lib/core/config/env' import { isDev } from './lib/core/config/env-flags' @@ -9,6 +10,34 @@ import { } from './lib/core/security/csp' import { LANDING_ROUTES } from './lib/landing/routes' +/** + * AEO/GEO-style posts (listicles, comparisons, how-tos) split out of `/blog` + * into the dedicated `/library` section so `/blog` stays editorial-only. + */ +const LIBRARY_MOVED_BLOG_SLUGS = [ + 'best-zapier-alternatives', + 'ai-agents-vs-rpa', + 'ai-agent-vs-chatbot', + 'openai-vs-n8n-vs-sim', + 'ai-agent-ideas', + 'how-to-create-an-ai-agent', +] as const + +/** + * Library articles retired by merging into a stronger article on the same + * search intent, keyed by retired slug. Keeps indexed URLs and inbound links + * pointing at the surviving article. + */ +const LIBRARY_MERGED_SLUGS: Record = { + 'automation-anywhere-alternative': 'ai-agents-vs-rpa', + 'ai-native-vs-traditional-workflow-automation': + 'ai-native-workflow-automation-vs-traditional-automation', + 'best-ai-workflow-builders-small-teams-2026': 'best-ai-workflow-builders', + 'best-ai-agent-builder-2026': 'best-ai-agent-platforms-2026', + 'best-ai-agent-builders-slack-crm-automation-2026': 'best-ai-agents-for-slack', + 'best-open-source-ai-agent-frameworks': 'open-source-ai-agent-platforms', +} + const nextConfig: NextConfig = { devIndicators: false, poweredByHeader: false, @@ -466,19 +495,25 @@ const nextConfig: NextConfig = { } ) - // Redirect /building and /studio to /blog (legacy URL support) - redirects.push( - { - source: '/building/:path*', - destination: 'https://www.sim.ai/blog/:path*', - permanent: true, - }, - { - source: '/studio/:path*', - destination: 'https://www.sim.ai/blog/:path*', - permanent: true, + /** + * Legacy `/building` and `/studio` URLs map to `/blog`. Posts since moved to + * `/library` get their own rules ahead of the wildcard (first match wins) + * so they land there in one hop instead of chaining through `/blog`. + */ + for (const legacyPrefix of ['building', 'studio']) { + for (const slug of LIBRARY_MOVED_BLOG_SLUGS) { + redirects.push({ + source: `/${legacyPrefix}/${slug}`, + destination: `${SIM_SITE_URL}/library/${slug}`, + permanent: true, + }) } - ) + redirects.push({ + source: `/${legacyPrefix}/:path*`, + destination: `${SIM_SITE_URL}/blog/:path*`, + permanent: true, + }) + } // The scheduled-tasks marketing page is retired with the feature. The URL is // indexed, so send it to the surface that still carries scheduled execution @@ -565,19 +600,7 @@ const nextConfig: NextConfig = { permanent: true, }) - /** - * AEO/GEO-style posts (listicles, comparisons, how-tos) were split out of - * `/blog` into the dedicated `/library` section so `/blog` stays - * editorial-only. Preserve previously indexed URLs for the moved posts. - */ - for (const slug of [ - 'best-zapier-alternatives', - 'ai-agents-vs-rpa', - 'ai-agent-vs-chatbot', - 'openai-vs-n8n-vs-sim', - 'ai-agent-ideas', - 'how-to-create-an-ai-agent', - ]) { + for (const slug of LIBRARY_MOVED_BLOG_SLUGS) { redirects.push({ source: `/blog/${slug}`, destination: `/library/${slug}`, @@ -585,6 +608,14 @@ const nextConfig: NextConfig = { }) } + for (const [retired, kept] of Object.entries(LIBRARY_MERGED_SLUGS)) { + redirects.push({ + source: `/library/${retired}`, + destination: `/library/${kept}`, + permanent: true, + }) + } + /** * The comparison route was renamed from `/comparison` to `/comparisons` * for naming consistency with `/integrations/[slug]` (plural category, diff --git a/apps/sim/proxy.ts b/apps/sim/proxy.ts index 9771768c549..d7cea16f614 100644 --- a/apps/sim/proxy.ts +++ b/apps/sim/proxy.ts @@ -3,7 +3,7 @@ import { getSessionCookie } from 'better-auth/cookies' import { type NextRequest, NextResponse } from 'next/server' import { resolveSimMcpHostPath } from '@/lib/api/mcp/host-routing' import { SIM_MCP_ROUTE_PATH } from '@/lib/api/mcp/urls' -import { APP_ENTRY_PATH, isAppSurfacePath } from '@/lib/navigation/paths' +import { APP_ENTRY_PATH, isAppSurfacePath, isNoindexPath } from '@/lib/navigation/paths' import { isOAuthAuthorizationCallback, resolveAuthRedirect } from '@/app/(auth)/auth-redirect' import { getEnv } from './lib/core/config/env' import { isAuthDisabled, isDev, isHosted } from './lib/core/config/env-flags' @@ -420,7 +420,9 @@ export function proxy(request: NextRequest) { } /** - * Keeps non-production sim.ai deployments out of search results. + * Keeps non-production sim.ai deployments, and app and utility surfaces on every + * deployment, out of search results. Applies to redirects too, so a signed-out + * crawler bounced from `/workspace/*` to `/login` sees the directive. * * `noindex` rather than a robots.txt `Disallow` is deliberate: a disallowed URL * can still be indexed when linked externally, and blocking the crawl stops @@ -434,7 +436,7 @@ function applyIndexingPolicy(request: NextRequest, response: NextResponse): Next request.headers.get('host') || request.nextUrl.host - if (isNonCanonicalSimHost(host)) { + if (isNonCanonicalSimHost(host) || isNoindexPath(request.nextUrl.pathname)) { response.headers.set('X-Robots-Tag', 'noindex, nofollow') } diff --git a/apps/sim/public/library/ai-native-vs-traditional-workflow-automation/cover.jpg b/apps/sim/public/library/ai-native-vs-traditional-workflow-automation/cover.jpg deleted file mode 100644 index 02ead8488652ce38c2e7510e51e0c6b78baf6e6f..0000000000000000000000000000000000000000 GIT binary patch literal 0 HcmV?d00001 literal 34893 zcmeFZbyy_LmOfaxySq2;?v1;E=GDrE>V7dQFSGCN5`~}n}Ga(zR3Up1sb#nydDgM6ab0>0)_(e zJ_NuAfB-E@ zF$4a+M=XjTH?Y7VjVHt9Wpwld)YNOMgSAq6^IIR-rwf%$A}K|wTpLpm(t?p(b51waFW%BE@_feaH+*S9aAZqRCFjX7WlRo zC~K_DR)QvL=o+T^>Hpn_paf6k>5^)GyqNg?tx)xPh`2}=@vTGt{3T?(2Q=UJf9sXm zd5y4>1OBZd|C{-LCGfuz_&-X3tT=M!f!PQU_8D#rfXuUXB3@^S=*~k@vJF()7SWRk z)O#Gy0N^cP$C}g#+6jH$TuAAuq3wM14FLcMQoXU_YQUnhgwX4v7DQ*=#sBeH01##E z?q6mvo2nJ_eGTg3Fv(#R!-hz&FQ4N{~v3I9Cgd6>Ovt=x0*z2lhcVG zw#vc&MgQv5Ok1i=?N9ycu$hwbbg>slBPISTMJ(0|qKo0~>tYfpwTCjN5n-&9iP$W1 z`Vj2GUEyU>j)V2Nsh3srif4JKBpJlF{FsPi;l(y&2t20II3i4X!MRaheT?{Zj1z71 z11j3um{YKgtPUGzJOOSp+Nz1_PNXctmjWq1+S%_toKDXm+SVK|LHnf)UWN%z-ls@# zdCg7GuSE0RYX<0{w^CSkQ0!yuNu4vIK3i8>bN^Ef!L*v{q`j7Ec2h-rKGD+F9oZ-w$%aqUwi$(Ew$QR{ilgZMs~VQ@X+&_+Z!D%ZAImICir`eFkR^Ca#GyFAvMrdqa#pIg+yvgtM7;T!?iL(3eqr> zw|DcaoXDl51Sm5<+ID8wvk+dbnm?Ttt_Z6AacNH-j_^$gvM)0YRhaX6o-@U=xo%*w zid1Dgg~vS`qpm8gX^C&8chNfY(v*{2Cs$uPEG)kWwiK6Q$seP4>oRs$5V}1U`bz}= zQAGg2imKx%rsL$?7}>sTN$x_W%U67>@yy1&l;3S5b5B^ul~XbU9isx|Q#7=5?yK~g zD5X|Y%GZ{3iTM%oV^TL~*G+Xzb8z*8AUxPYDNKeF)v~eb8lL@oGPa+?^0^4Yx-LR| zyHjs7E}bNVqs54x0S?5ghxSI=&BHUq&Qr>1-Oqjbu)q$==#VglaPOuw$5&xRJ_BI# z+%4qZI^oZ2GFKEJd=M($o$8Ka86J$q#p5Y!jBOEQd*z%zx0yQH2|oT6NQ-b3@8PGx zX~v}eg+8zTcqZ}aS0weHb_-#Ym!<+*6RUWv6b!X=jBaUO1|tAN7G8m|IlKx3i$g7q zz!i5{{_}1AbjH6*1pt5=-TGYdMZgO>KmF=x@~CM8_qDDX{KPtZ2T}ewffxOFC-JD_ znY5Onlj($A>PF&XqVqR(g{r!$rmF9rL~`U1p`(&Ig0EqMU0M@WBlWYMmu{XmH?{Uo z_pxD;w%dN0%tX*Bk8FqUDgR#Gnh+xkQgCjN&hm7uZsPfe+euxgmC9hp1KHdJ@2jVk zB1LP1*uwxLpX>$_JoAb&b1ZampUo23bPmUtzO*teX}HcFk`O;(!>yGY&tHrfKU#ek zVv*l*n?Et=S+s+yZTzcXCTT@u@!;zVV=j;AG2BDvB36yjqd`)~r(H!6Mp%I>64(1+ z157m;Th)@b^5+no^l<*2V_p{`0T-O6|129ooaB(XEWoOHk>na8FX|oOK#j7FEXMYL zeSI~Jar`rc=@yi-VLL=zmmSYB{MSPfEQS%Zz_|}8>TUtdbsmC`u|1)FAxyYQto$0; zjfU2#k-pAo-h)^kYAO&j<~u$;cguQ+)hG5CXVzW!WybLJ1AqUvff2!P2>1mb|G4`{%0GWe z8!KHx*O}>53Wx-SD(*QxRd3k5v4wp4akM#<@Rf3bi`?OB{^IX!bPHoUeUt^|Ra>Dd zj)>a3>DaMzXTtgXUudYAeFw$ekn!*-e7?atS9aPUeP9ahizzVk;oQe}bnp}R<#zk^ zedTyh0%XOh1PNbT^ui^vz=Jx|vS71BnmLnS{rcH7i*papaCeg}mku-S5Iysv9pG0G zV^WzLMQ#Knl`>II02s~Zc?u3Xla9_k9RaK~U9;jc(9gt%Rxm1PV{4=fXvF`l(#HWv z^eko$RjcJ&RHc+{XYAnx35P(r9g$2_F?Vxp%Qbl=NVw?Cj){a=L-QrkHR3 z?qx(?QL&}T;EIJQ{(+yeo7SoF4$n>JoK7mYm-32#q>5~FTLn&;J&nsrRPE4llSNb8 z?TlX!4K>UjXx=G;Usrn-j< zIVWbH+Kfe!{yCZnpzR@wNm$y~;U2-D_YszgTScQAI?hXsO~_0^WO89_>rlK~cf&K1p0h>VQ{9|d{Cm@!_h3Fm^+@a2a^71mI17Y{Pt3OU z&S8B*N(4B!usbNweBsD>E115923y+;d|Ucm(}Y(?*)M^8v0ph|;yDSRnpQLpOJk(^ z|3ysTsBSdpqJUsaJEw|rMkSG73?TZmZ^PVWy(8q>mQ)IxCnDxTYGdX`Dx08`bg)Sl zR4A|uFX(>1mND4VbEt_XyDe>a%gGi{#Eo|M+ZsW^B27a`?SZ$%oz5t%-3kC@r$9gK zx}e@B_(PM+zQe|R`cy#YnP1hW;utp;iWTV}bu885GstYTOlK zU>sCWStO&ZU!>K9G)8x!r0EMveJ;Z%p3jM`v>NfZd?qt-J>n4bJ}Tf@-{K_ywO;>H z#s8;6>~-RQ%>o(jTR@VJtUT0E0oM(@0LmDcSVQ-Ls{pF2ckJo$z_NMpr`Sz-Z|c_x zAOV0nij?7U>41;x9~+34LdU~BX8-`Cii!AYQ*wT`F!kHK1CilhSO4n{Z@kJh#>b2lYQ--(%d zE4IydL)gSttz>F7N!RzXW#Iwx8P8Ny4RX#?>FCM?d9fQQwf@_RLXAg? z{CtgsNPK?bd!=nUqP=!V;p;N0qwDoQLG+IX*r)`jvRJe_QMzy;M(qspUjb@iQyqgv)IsmV5&*d92KfEcb;t z-?7OY94p_-Zl0&$1u}?ZvW@sy#>S$DKZ6^p;l|Ps1adXjqaKn6c}*HX~Or z?YSG72s!yFrFTMS1vC5jb!4iTUcv5Lyzre{%t+9hcVpO=yyHaYa_79S{8WB1s~FG1qB5G1qX)!2L%Iw0RZ3-KmZjIfXohsf`*QY!ODS&Ma-ti^jXZ1 z+>wMdfGnXOh?znIkyH?9z&qfg6t7DME>HOLDueoxYTt=mLqp^M?w%lC&+&nK_sp-~ zStB!Z4=Ioq=iFzh9aGXHXTdBtW3oa<@Kf?kbG7vyQ{-o-@({vm(G@dJG15vQ&eY>C znleC>JtxX+hEGxW{qQP@*VRl0x#;zNL7QM+4eQOI`e-PVMRPx^quBlI9iYE|b7;K8 zHJ*t7a_&pZ$KQo<%x-Z6oiT~6ID()7Z|a5`RkEq6R(Z_*xh+5oRvvb|*?Q5OLrux8 z^(N}pEIJdrJYt7-dqM-0-X?r?hX6^x2*B=}QQaeA+Q+$4zx%kIZdSpsRT%$FnLNsK zpHr@)Y4zkuogO_{PT-pJl&Yokcx0<&xk5 zz8x%OJQ~R|Sq&9PmDV?2JsGj$dBaD^FI1g^m)Q)H%3e)kL^leQSt~8dUJbr)#BTk- zs%^Xjbj*a?QALpbGJbH&u&bMtpbn9?c^k)~yxPEsNfye%xxWR4S-5_QfU5{fa8@kuODcX3`U^h+ApUa*Gf?I58&> zi#Jo-h9B4-myAX%6h6V6k%_6|W%RX}*7$kfL;b?;a*g&!kdOBr!2I^wc+o;x#isNQ zAiwdfu~DO}!rFbM-=kQWK!6f!N=srbL%W9U3w{TTcpmo(Vk+pUq@2?hPN(-a9aTNN z?4-Qnc6vm<5#-I;o%I{tigkDegxr{ zYzy?WT~~`kex2QSqcPU!9P;)cn4IlSiB6JJ{QDMSc18O+ih&rWK(ryQGLnsB8tQ{Am}q0WS45IYS; zM*nC^AB>R4lt2q9!4_A^d(bR~hNN1wFAaChUCu*TjCEzhunP9F6_u-4Qp1FEGLFsI zf02ccG)KRiWN)8MvuW%y!dcBZz`mWXLtA7FGa-b@^V_UFqwJ6}%g|RB=3;4x+1DI# zYEA!%f_FHJH2EaOCwuG5xMI$=HYa+oiK*@(RS+Iynjs)#+KWA+1bn-jeF-9z>3dv^ z+b8PwD5y9J$3u8sR_4}^J}hBeR}vm9dk9DD!8&w{aJK;+G^SMXnYukBws`I`{hR#d zuI+JnpIDjfG0KQDb&s5GFAlt;xCu92IyUjm5MyzH)Kr56Jr@mShMroB9=w)@fO%U& z4qc1|(w#74wbCW&z#D_bT#I$7 z%-AMCoW40Yu*}WOTf|+RezC*Y#P`LZ*V>bB0qo(_#NMr@(1!IwIMjlbk}DARw!nJ2 zjnc=<#w1H4kC&O6&L7#|ye|r8T1bVaQJQ?R#7ThYVbSx$RxR*i_^dxjQk9X#A&a+h zqAz;Utk_6IkbqzlefwpJ9aMH2U&>%e8>uP&kA^_x2p!N>LUaS$+We|&=R5AYzd=`i z7-E8Hoh`aH2_V=97+?T$3CR0we_>d}M?cPB8^nV5sd*Y7&if53{rGPENr?>!);bh@ zSv_D&4JH*jY8e%}IN96*522+StS|?CR?N<9jhU-8Dq63Y_3N^A8!M|L6CctA7 zjh5|4603L}-!$md@et8^-BIGM&1u&IEq9Ikat>+Kc;OixoedaOVD)GBDOFH$_uK9& zXy}v4*GZ{V`ah@YR7N5?HMW_;BLS%mTXKJN#HV^VoIEDiI@uO8rD`8>=G!MXpr%B%!YbRa@3Z%}9uMG)C%6O| zS3Z9je#_inzMOu)E%^9f=(+bn8lS#F;eM&iWKFp|ny4&|8jF+Y{pL{mlD&up4Lt<* zNhpn($+F?2`FTtYC6}PDm>y#c_*NViuOtI78q{{}$?<7Xd#zb}9tl*7p zjce4;kU8CC8jeFJx& z?T>qYLfJjJ1{ZwQ$c_~Y!D^kV9-%k1XP)@zXR@3m{8f{O(VISA5dpVqNIQIx+zyda zGO`lEM=XSq5_u#?0$J70;zNzMIHMotb(#`;@~rcs=l&}geDHX=QFZ8EB9(%!GZyWE zX~%yNdantWs{tS14BzHKNoI^qPT5dK8XC<|9>mSVC_*S+z%%r^Z$pUnJi2k`wZnJOoIlXXaTxv2SXT!qF?ZRzs^G@(g}r`~{2sE9z+Vh` zRFFvX+z0q0_h-g=_z=Bm;$@1n#uR)C(5$}^ZBp{#4w%DO)swuR*lbNlb{u6X6Vl`Q zT*~^F%?0{8FCTJdQ<4a(KqJ^DszS<&3+;)_P#33Z?K-R7=oWnkAWft7%$EMc{%{^H zwgxKHX*r=^6xlWJO0f%XaEu-VVeBom<7sAbrE2KX)M5;#YiG~ zRe*bY?%3%3inlg|OpINm2);oW%j+empp!r;Zd#E;L3ONO>q@Qz$5Epg1l-BdE~VoK$aELGg`Y%^9$r#0t;QJ&(Sz&aOBzAb4wI+R;2&Y@mS|Eki` z)RYuQ5i=s2m`8FruOCzR5=ph-$t64{jpkXB!k^q@drqu%#EM`7i8ds*(&9n+4oFha z`VcW`GBbV>bzDU2o;6?ga|1;3B9W+{=44hq3kY%QY&K?x%tR4aUS`(%5&6L?8FLuw zX<06pwjH`EJNu&i12g!CV<5}{bPTpRFpf8VrkpQYb%rE_F2q>i2s^Y4tTHu%S(y`w zlQ&P_DN8*o6qJCQAyV*mbb#CaI_7aJvV>^-w4t)gz($+A?*%|G5kq8s>JW6zP;cZX{-}-IycnnJKzr0$`zCZLgzW4 zEixDW!ke+^e&Qn*@Sb}Iv@}kf#SfdpVvZX#9dKaKLf+a{`RKW}tj7PsZLE+~$8k6U z-SIP|KmLy&YAcz5eLca&Km9u2pd*cAWby0*DOrSdlMQh$JMSpDH1H=pn4~PQN8w0G zVWn7s*(R!omo>}<>;G9oFnU+hODGPUyO`-A_6r-(7k?XDO?kEUZ`t`x&RzN1-&Q(? zLs`OR6uQ(3J5K2Qc3`aMEv6A6_+7%MPuh{6Q%?Y)l#9v>y6%ds*TzOtxe+w^RkSBp zfg{&*@Y5o(95XM5Aj@Ny#^yzPQwQV8Og6OkKvW;pg%2ibwb7p7h4O=|;D~Ltfa&)x z$FByDUQKw;aw|x#8tN!&IoQJ*)A82KtnhI1b@ix^yIz0RTa6d6#L#WqyeZ zt25&{E$!aD?m}aJX$Pb3EXT4-V6N|>(7hs8Ul32QGXdAo(>3CAQ{;i& zWhC?ZP0tb>J9$ULcOMHp&ZYXHDP}g{hUF1j5)b3=; zZ9wHNzOd%hf?Yscise@h--u#oL^kRj7O~<#{nJmU>p760{`k<1c7g9@L~Dt-tZ}4C zX--+5;}OL3Z!sRMe5KTJ@_=)*wK9Rz3WEncph|7NKIasIbB-#d=cGD-U~wtiv&J0E z5@i#6fnqgTg5%(lyg{n$90=Zd)w1>d%M5=z@Y`Z zHsIKZlEZcGg5leQ&v$8m%g9H$L&Y$rI2h^8lV|I#6q#YD?HWGR9hlVfYt2f@xK(7E zcfMx=lQLbZLhDN-Qvf{%r?S8iPKrIG*DICulGOK^;vz zJ;W&n!ae&p@|8u{4nyWs!ycRG(#P?RCG~L--@u%X=1_HCblZHLxGmTt`*L9pAZje} zjOsu40Ba*Sd_gCmoj-ZS^nfV_wnTLw!49U@$oOm%bQ%{jxB?wuF@4!H_*am1x)1dQ zE|19XI-KY%x(BlQS=Wh|1pf`7Fk+!z4;qt6p44B(*dZT9ZW$VrZ8;zvxz?;4p|2*Y z*u1e8M4kHUt0TtHc%+L#fm|LD%*ioE5`Q6z<0HAGOrx`!hKCZO&C)+^Lo`;Ve*=FSF_V}(Sn6wl}|x1jy%To zaQU5WNNw2H7GDg;rI}(%C~JN?`nFB#E#^BeXZ25G$Suo&rcu*IH+#3Z-PkvxwGA;= z*9C@q9x7AadH68UjJA8F{Rs0yn}g4hWum*=#d_&iv~?BfiEd!cb2#?V00(zr;&;Q! zrhnQrwJb1E{!E*hYGi69X&+{yx8bb4K14+i2N#Gf>uRy7=_0efbmzabas6dc#{gwG zJiOQ(d+L$8DFGQ9-O|C)B`nQFHM4GSFf>_l;Oejliv?Xi^A|#Ci$$5F(2l2S6jvv1 z=UIHmn#*On5-Z5Q{!Cf+u6Ouf{#b)`bSO+T6_l(boi?R*47ny{Y1j(a-1%l?r*8 zL6=t0#ecuXKRfXZXN6+#oK8UZhYZDk*I6FKyUWg17JXY4mBc)jD?%?<27EcZNO_Oc zMn&3K+>mS4b4y&SI(l3unb!CP(Zl)owx5o=n2XK1mh-Dw))*7U7(fo^JR^ZL z-i}AANrN|1;1hnCcrFlTadvZ_MLvj^Qw3rpYp!gX_#~w3c5bo0snbgBhRrNbCzlz` zY_-+_-r$8JIupi>sdVA+w^*BJ5eyGIU+MJgTs2+xl!ywx+8NC=TByvw1H5gfk-g59 z%+?N97uZ!g4<7riatcv+C#$-zmCR9H7rN_^`pgIOh;?!rVe>rvML#*M9;dCsRW(*j zaEFxOmt35A_B^&b5AGH&X8*&CpEtFhnspTTrrmPSI3K67iJIA*#`6gm3c@Zh7wKi~ z7~@}qwT4bhV5bdr;T%2Vo6&2{R6T<3#Jb(nmeCl?|CyNOL;SWW-Y^8e_;#ST41Ci+ zYH7li$EItU98l=G-(!apLiTQ;e4fF2<&p97V2EPWQx)w3Ak!^W@TtKb9iY6k$y+t~Ee+nPFQIMD;rOCgCspeb-rZ z{OuHv=}z&-gD*;cB4d?r)szyYuDalqwAR6F;9*$_2`N0y-uQ2=nC~9s+P_Tk^Z z#jYV)?yg3;_s(kDGbjh9$$?>Ij4;QTfTJ^|?hhcWqJGVp+3yxa^z%f&s?{~yHz>=r zilm-&Le-5d(t zLzqGa$Z`W0&wUuj)aaqONYQF1UeA3fJGdK))zd-kD(v@S*$ixU}TP;{ALl*};_*l|4eRj=7Ty95y7-#l%O`{)}z zFk)?RQi|Kr**BjTtC-899Z1Jf{KI# zaST^0ccc*`6m{&jipn$g%avtrdN#M@F-tEGSEtAHf(b zbCY~F(>KtPY-f5}$C=hq4<66$D%FCy<-bfFm+((f3VYg{Ox$rEk>EH&%&CDom?3{| zmic=tzjB0dP_wwJM%ys6`|XH51=^mah}Jf)V?_wzmQ(6UwqL`Y(;fxxz?|-5&1+1P z^{&2fe&*k|+9b4udk5T@z}(tQ`e!P@tNfCvsYN$%h*155b0{Vt{ia)nsT> zDfaYU;_)}1&orG4sJa(Zg!^O%A49f-?=SqT$CKuxVbO8PiE=g5Qy{X1ge*J6SGr+< z%kkw@VZJOoixz^$)>5J&BTgmO{V6nOHAl(!keR`4v#Mmmkgw)1jTQXGzRlS?j%eNI zNPl$HyO=zoT4VI3$&w}KDMawwaX7=YrypI93=EAibAzrqHYl^l#~AQbroDE_P`bHj z-Cy#_ba8n>NN-+EBd`>_{o7>e#GWmsUf9WJ1JhEfEP$(4B>JL?{)~tPwQGaYrf-T< zUS36NQyz1DTs?Z;}S8so{zONP?3QIakuSW-Hu$ z_~M7(0U$wulVwF@Jp>!vu%& zNCIXYlR%C!iTntuSntft6j*Ty2}MvxN2>JHizZy0{zg7jFMd;Zg@Y=>Ypw5)&j zpss=c4xHh_N-aH>L_=TG1^F%7|6}mIV%&jwwRlaeJ3+|?P(;F2|Kzg!YN>y@R{a6$ z+;?p1g3irXtY{|9>%vfGC!eP+Dh918L2O>B6#*;-*&YqHZGXEqZ(G7SC&_fpaxe;H z{Ba8!J8U)n#y*`_#&yT=0pZB(puYo1y6PY89NEC-tfVp;2a=0h5<>S`hKg_`!n##E zFBCNqG@9;xzaoBVuuG`!*QGb(CB}QK^bJ<*PNcw zlShI*!4gM7i23{(YwU}tb^A+gyG`pI`YhA_qxFUbGjfmwYK5+78@(UPO+SJLYn&;O zNKTc`SJLZ_^XJ(#5o0)Lb7KpxIM$q>vlz^3xZtai`JBWFpMRq&*Vaa1Ya$43v!3$G zhS^4flB?v5b7`*{z?A;N7TA)6!E+P19@3FVu>Y*hom`6Xu$s zh_Qfb(QjIKjF&7fJ>C^Bzn8v8&aduO3S=^OTLELSJ4YyKRBzgyn!^RmVwwuzG!{d8 zhA!LVzBUed%H1XE+7*Sj={B){@7Jw1Gd+hNtpw!RkduX)h(n`JXJT;3cl?<`GDB;k zpEVfjZbw7bT+J3?PoOP`%#58inCfn)4$4)h$I6^xjyWd(DJhm`#N&2{Wr1?2q2}w< z_;qeB6T10^BvCR=q-v5Rsio@}Vd=O= zVf2$XgIMj$f=Q!#c_aB`Wla9%Kz}e#i$jP%u=xJ+XI68(SITaiu6rkjmj=khl?>EZ5@K={S& zaMhZLbVy1vEX3|J94H|dkI?(nt1n|%t9MkZ1j*_Qm1lN@>&tElM~f2es60St%?si; zGmkxDD25Y=C=en?M?%=d9>6((2KE~uH)_Uhs{`>%D#aMG{xFugH%%&|B8AC3;phY_ zG~M26t@YWVk=!P@cm5Je-*F-anf!j0d&L?G3ZT7t!)=qyViRqOnN<>086VkYKJB`? z(M=JF-Ih{MlI2}rncAWFedverUO!Km(3~D2iLaRtL?UgWL2I!H?2NN`w%ihV2!q8| z#upsp*-a>=i0nX(YUv{iS{m@^ALHvuX$X?bt{%DP}yxPzWyIIm;)S` zk;CJ0wI>HaE-c3%uDAY97$4iZ>do*(b21yq6H7M4hs6zUv?MhB<@O=I1EyTA!8#yj zk@`FaB~luthy~YFMA?;<`1tkR(ecE{*&Nud0kBi84zsgjt<6ZDAZJaA}qS$ zfG>Z!R9Og=JmEF7$YLS(`l>EgM~r-(eZcNwl#G3VJ(N+OCxEjk_hUWkblFu0&`9|e z$Y^i{_f#IXX!xqD8SxmySS%m@W&t!@S(t95bATGn*SH_%a{Qfo`zNT1mzt28Cu@SE z#We~SSYjy>slkbHA8_PZHMoCBlL5tu+}qe|Bb~op;-VwRuB1y-7V|q9(dzR8i*>~C z(ux?zqFw#8`ntZ$l@$ur!%)FFoy}-lL`i0eTF!;4CU2`#liucuI>zTThPjA7y3g!Y zVVyU_mXJ6BpFRKdlm);A2&3z|xK;y>1*&7_x8ie-f&$p+o>7@a!+~>$)Hlo*{KmF- z8?-*SI+@;B$020eUqyBlQ`_j@9ZMIEkLll~jd-2z-Cn45!Ftdgx$!^TXimTiStz zTrBLFkrzL7+V=5ty?YC}5B~kbkM4ThUT<1BflM?8&(ewVWR&s4ZYPiPU)+ z84>c8cA@vcaW7vz09S|%XKd{iNE-<7tYNi+^xsY8sjO7ZFP}c#2P}esVHMK{Bh=~wahR+XdxZ)Q&DHv<^*6^)R(1|7ws%Y zrfRll>q;F)} zrXlBEvuF#Mby3|gth<$Zs%17{pFB&j`0>AesU&muj5Cd&g^nqP{GROV4XWd;Ptxw& zJ{rxN^rolESh`_rj4&zJEz?p{&O21B6782yQ({(&qq!b>I4iBeq%Ill^8so6YcPEO zcelYW_U|2>Q&V1&aY`jw(hflS)TqaEZR?JZzzX^7C|Or-0ax2uW!;O?Sd)%3c;vIV zQ-xyDOfE7gh$$(z+q?c(Kx-fi1EezfXwz{#(rvsIMI$9kGhRQlO>E+^v@|?snrGA+ z(oU~D{f`Y^PTIiU58YhzH)ab&C>N>+Z{N_=X3fJ4|ZTaG%vO@U-o*PGg$ z3$$YaT5Z;7sltXkg#`EDTS>;0o!y?wkWrnrYDJDTZMHDxaR+f=^Kdvl`Wnu3x`Cr} zlO&$JzVUMUf&mAC=Xjhvv9BoTDorZHMdy(20XULd>qF!b3zCwDZbMAdJWOd8d7qZ> z7G>a$2r-R049JFg{*~4widJ3cw>>TT6k$Y3%L9<=p5pJ-Ux{o6x~qs#I*a#x*^O-EnIQmGT;=mXsYl7TSno~-@!_qpTyz-JUQf7niN!U1P_}*S?C{*HG21{Ac@=|{T*i6050c%S^P+AgsDryQucrhiylRB9$T-PG=nM^v~*=VX4&`g z%eK~^nkL|y;(>C|=;tR5!+mRd9OV~=Hk?#-6C&V^X)7l%8A?}9wL)2>$p$Yh)|uV` zC2mloe0=~B4He^1$I8nW<%A(zl~khpnSmq8f%1$uBfZ%v+K%T?cR{Aw@4U;UxY+A& zdE@odNR3ZQb}QUQC={<+Pk}C!T-WNC$!o9_JqkNF)S*g>f6MwR>mDt{)!>+zlU-Xr z?(Zsy%n))+NXtrx6{YW=e{8OwEs%4$yFR0JCw>ihnj}0FS^i}+753(DBu5#j%I42# ze|-3J>rYQtIFe+Eo)}MI)fzkDn!keeXOmf@pDo);QDI7S@n_rjwTI5RDxnF0c6{U# zkvREh>n7=sG(q)=+pgU*LEKUPvKiPy?*7MV%;ok4&PigTw6-%JGlpvC8(KA2Okn#w zNE6{QSASk!=9z_Vb}3cI!NYeV;wFV}T4q8c4xwTx(g%Dxk;fEf4J4l{ctaVCgck>-AF_7x*!P!g zNL#=~SEPz}u*p9P2$!>Gq@7uQTNp}Iz_SLx&RfJ_6u38Nt@ysUA@3;xwS>^`a}xXw$+eKh4hMSy1x z54C355T^VX%9m#rp}=GwZaj{aq@2rbSh$?*R%7INUc*K2lNohtmw52@u$nbqRJUA7 zm2NNPqHo`(w(Q72Yv>C^!7m!^GIf`JPqrZDV328eqq&Nfs!)k$G3uKw$>G5_tKulX z;z_$7H?-&2@I*Ys25m4ep{1}V7P3q^M~cGTD3T=Rz5}`uh9ZBt*hAIw2;-|=nVnU9 zGO_HwddfU%B|UbrAXeLtN(b>~r9ut32RB~Eo;6w&jc844%UU(oT6Mo{bpn;q$H{7V zqChISa+&~63vo;amYos@GglgF$e6TH?lDU0H|nqicLI?${}!I!wU|Or7lG!_`F6#L zTHY~UkK1kqU8GJzc}AWZGIAU~G&>#9@n#ixP=mT3VLcT#vZ3~3-g9M)AaiQhqNMw? zCGJ(1f{KCC-bW#iOKt?^N?RQWuO0w#>V)`FE{Vh#BkU2rYuNJdC9MaabLyAcorf3; z>#``lzXBb~`fhATlxFo|gdH$6+JeXR?GZ|;%GZ@ zQxO&5?KN-v%V-jOSUxCIwdo&3wj1OlBP-Jsu+RBm*Sc0W+XB0^k=_G&wq_f!Dr4nB zKeS8`0@z*hg}LIEv&vT$;(!#)>jxvU$bQCHOAot&YR3IMtt~U_;64fM$QGkI5P{?- zK^k38H^}5ARWx4~A}C7o0OpHUIC0JVX5N;vD}*w2@}nd|TzTAd!-zBa3xS-T5Udn) zwPIzS(LbB$Kcg`oK@5Hr!w1p=Ann8Ijss`lIuo7FsmVEqHEtBS?th<-#^dn%-Tq@@ za`qpyutJ3_6!fL3EQ>{lPf7ZzS>1dw4UmgR?|>ah;QUUlG6iIRkG{Vc>sy_5PQ}v7 z^Y2|?0p+1!el%;`PgREByF|`maZ$IIbsgDKg-6V&87L6U-$oQ?Xg@8kO$FB}uB8nB zE?C#fP9QS(fHQx&e^k-Z>1wo|cEbOHK#D#Wsfxx>uKO1SC3@++N4BJgh2-2ENpZDZ~#9cdH6_(j!>>Vu9VqyX}Kz<8SH1DXL2s{4pzRoUf)Kly6o*%SQq zQ+S#>eS0-fftJk?YH9_Wkghr)a*D3X-vqV@51YfTepi>Sa(v#-ZBWO~J9VAr@gGA} zs7o8@9zALFw@(}ND+GJ}*!LXMU*;s6_f#+Fl0Aov+sd$3GgP5JwX>qnHTX(@!{_Pv zVCR6Y5^GF{#+Tq-Yxt63DT9Py@ST{138_hWc%AL{g<+los|$MrFi zWst&?1MzcqQ-gJ^GXJ9c!?IY?h)K>U&)==vUt}gw1wWd@|A@PZY%sqe>SuYruE78O=%rc?oF!p0JsXxYX~(r1PF zE<9!6t~U0OAX50CX}?^nwG2XmPn2A``52H z#F`sx+Qv&VsxzBl2N0jx>X2EqQEex=2Zm^tp2xJT$|>k3Nuwo2me>i9R?(Zt$Wt1) zV98uW4(NS#Z1y~DQ0tV=e+%H(Ln;ozw~0z|dyeGd5GaqB6AObG6f@%3oNXGwO=>27 ziPR&08b@7OtQ^EF)Gl>fBYGhN4lTbCBK3Z*j7g)MHq-S&V7$)pKbB$0l^{(MgLe(0wV}qP;#Uii4HqN`U76m!Da#6 zXP6w!+-&`Qj!4ZpHo(5%enDo8m86vQU+Z$DHtks-XPWmf~EXL;!@-^7IQXm{~ z2BtzD;l{#bbdAeAC%U!bh_|PoFxW*9g>SI;`;{&NRrY&HowrU}L;KgC)w(zPSwkeW zrg$5J5)H|xty!is9jv~sRf~>OeW$DMfe}5}5J|ESaVn@-; zRFWMztNqMA<-lEJDs^Da8(=6WK`RRLJ^!;@*h3q=7l&?_0$(R zLaPj~`5QBgfYj8MBV#O1%_*^S1ai6Y47&?iC0y>~oZU)Mf7h#tMq=)DWkN%Ud#ZuBk~JvvGBXrmJZ!61XtqXyBt1Yz{(C0Y`R zF3CH-_x(K2{XXyS_x>|{wsW?9&f2r~TGw^$y+)Lz9z~e7URphR)&Ld7xh!`;dzs2 zF$?{c`4R{IL5=^LKG?$OuU_CuiX_6DhJ0XnWV-O#fxCw({ZEisYKOZ-(PLA6ia`x# z8QKHbGH_QacYjwnwe^76$xq7%4V$lx-46n*0ZL^AnI(?&3+1k>hGBK4_BIRiqMgI)qW;f5tk>1wzWD#V zJHX{I@Z&VSg2@YTY*aC1))U|nn&1-oS`-eCmz$k0T~@Hluhbel*Vt03_cAcuC``^+ zPEX|ubF>=z$;3D~ef}2^)!qN&&!6(WHxJ_9nD{Ln!Q?+r{6A({Ol^KPNYG=%)XkRz|l}eBwvLU@;ecWeXS; z$d=teK$J%xCGzV@hM0?;rl^hPuQ=8$yB-F5xMgpR>Vf~R^pg0s3e_$8t&1Z8SMJK2 za%>GJzcNvFG))Yg8X@V-S!lr>ynGr%2j=&5gdV8~IFQqbiQcpQ)mFI$apYB&Qi%2$M8W7UTCbXXp$a5BQUb^F!|$P>XiOR4UM$Zc9c=qr|K}?BkGDF1_`vV z>wKh2HX{iMQ;5(db-2jiU2o`O*%Z56H06VmwE2u9^ z#S%GQi4O=O0(VbJ3E0=+j|W@_)yTJnH@tt;Y1(`LDW=}YI$~V`+Eo>2Km@kq75Z!z((7NBg18ulOZ-yQ4+ioK3T%mmvlw<*FH?!WJoc6c1|UWt7~zT(5J5l2xy zX#;pYDN)i*G2un>;h%i6I^b9r(j(lTNnhpZH;9U{C@%A_d{F~c(uot=zX0H`3fkX( zSlR?~1k>`qQ69wgmbWmv`|wG1Qw1!eVS4%6-f$QNbrnexcw_LUdKq{=<2uxnB3Z3b zoH*3ei@T6Oz2Gg8kf@YzxX36G-(2VJ8si3%-XI`;ZMO_}kR%QpIm>?1uyApQ5_o(Y>4whxLn`Xnz-=G&rR!i1d@{&H0mo3FWMNDN z=Zko#3d!wU#eI}8Jt#`SS0{t>iPs=qosBueCm3i3R>b|#yAv^>-!#a<5m?zQu^sOt zvXn3TAJ$@Z{lb@>bs7~{>nQK~rXcRk&vc;;opTuypd zMV?AEIkngiwB&*CZ=Lgq9Va>ZMWv%j_e3lC@ft`+(Ue={KSV7cY75VL>Hj)k=kztf4-i?86Z86VF@9y+DAkk=v|9XT&S6aor9n~5B! z8LdIfy_l_c;!O@b&z`^<>eRK#i8BB&(#~c!SMTB2RGMyZ{{EKz(%UTH`Q(NMP#1Sj zr2e5_e%^L^OA}}wa?9S;Q0K|@z{3%4iNHSul5jSPy12+mH31XK^ES2`YrcR{oKw5q zHWWlQ%buZ85S#gNV2|Q1*+Lq<7qWtH{+5sAGOcnm zWp%fUiBDOWKB71UHJfR_Ywd6WW^%}Vu8+Hw`Gf$U?XaYVao{%`Ww7~)O8tv;zfwn@ zW?e^tfx4Ag9P?3@hav6wH_7rk3v(KL7)ZYnoR}z^1%1BR%0hYmcoQbxnk+@g(~0*D>e1dXeHgr&)_P(YcuG`Sg4XwQvSz#y)5c z{h=`uX1oMn8KU{UJ?nbt;PSk;(#q+8>$N&?)yCS zu>Zq@?QaN=OrhU19{n4?aB~c~ZMT+xyB?+}v7bvfd`AQ%PK$i^{PXU^b$QV92JVvI zyh|F+{JOg;?vj%+q+QyZ{nz<)F05Fc`DYI9t4469Q&q$+0XX;nJwH)nE2l8D39>57CaLajfZUc>8cbVBH}a>2 zRnI4?`L4LGa;gP9(iX86m4{n%D**YKJ>=6XJtI2OM4$yMYcH7RlNQUeHntH#B5_h=$g@C49i*voD1w!Lfnxa?{Icxi24cH=yk} zW$_zFgIDR35q|-hN{DSdP7S4k6_?_<*xgZD*hy5 zjtFuz-Fm7KUK*S2%=HMpB_zePz5Zr3Drf)4bG?=8rI&jT7786}CDDAfX+05vADH9X zcj_-3=WY_gPRZMcPb*z%gY585C)~A_!V#jR0G6;eo~eHhFip#m?JiN%p*7d3F63 zbA$-#+E?`yiurGRG%r4m;4qChr}jk<6$9{DsariBGJe^b;-+KZ!pRR&ch=2d+c65w z)8(wBP$gN~v*$=NrFi1T$QpbJuggj;7M^}f^x=2t$H;(43!_!!3!`Ef|EqMZ@kD(FXh66fM~<2O?&+a(Ut%ZL8YO%f!~A&E>(|mJICRd*%nf1-i{tA z$GhpbF#2{G81FNqIreGXd4W%QA5cJQvK`u<(ALELOW~BSp9AS^UmtDjeJOxYh@U=k ztPQup%PJzs-M+B2ul`Ocf-j;k2P`+Ii0$WVli>9*9y1*CGSDrxh@9kzeAb_JK}2h; zUnPa5VjS=2HCrL0#7U~CyGBIorj+*~^~tXi=R#V8^%uxeb9k0tVUO8gKq8kBIz;b8 z80bd3Yfi*w->x=j@0=oym2ncQS-(6d=5)~-Q-x@TZ}sE7DSf_S8# zHYUb!Xc1U19cwbcCg^Af_%Iy>xeJoY+Xe7tSs}ucAYOEbB4y1I-O3t=P0GLuh-Ff= z%-Nm*7Ya?(KDyp{nRA0i6wxiH<&oICz>agYVwq5LAO}ZYVsTA~>CQb&Ucp$k z6TXOSD1<`pA*l+KAuYu!CS_BAOjt>1+({C10 zkqZDMXIesTTB2n}eI~e*IQe5NxPbbo!a+sQLkU?Vxlyna!0)CafK4FVi&sB0j`L%a zw;V+_C%HotGawkcgde{7j_pMgFjBbE+y@RZUe&SN_Ich;tF$zf3rX~S@?4E^ZYs7s zA;%bCmC=#Ov zD`fYB^?fYnB&*A;AXR<>OWR~msG#0CnwaW$^FKY=H7G3Kwaj`YTIYO%9sK6T{OMhp z;526hQ?YcHeMKNajem3Vtk_$ejH~0iM;C!cxCZYs?V(5rcMm?0Iy;UhE%IFq9g3Fc z#+952hxEsb@6wq5X40BeGXQ}Ro0$FKAgklT-vd)Z$D(p3Ud` z9@n?d#3)XRA$sZ$d?;~wW;eD^brRKS(NICD3s;BJ`VQ5ihAX0e`%UWsI*Hi+v4RpM zaD@PWQm0+OEsKGfF~J#RfH+lk1Ly`|_|6u^b6(_J;a=xrMNvsaphv zw~0=+cjs)d2`GxCP&us$L}+*=*E`K#%F2}okHD_48Y8WA?n-q?)aECP3DIlyc_inaLH!cMxs!_TP`A&wohbxko z`v6OkyS_o4g)QS`E;df%%Fh~DHVL{_+QHYQ#Y-;j^jNG_LLyQ$8q+Ax`ThBuqN);? ziA7PgAn`_i5}oli!x%VTBMdLGAXQ@Y(YV0|)C=T@VmAYL=e;Er-e+UWUSLM z{f&bUS$QNzsH5Z3R}Jagl}`^*q)eta^i3@B?C}8N40kLUU^2?1u*b8!KhQ}*whv|Z z20a)?Iw#(n9l_tTLhcV1rQ&OnI7)o6n|UAa=;fY)AnDLJe-Y2KbJu;J5ZwQ;^whuz zp4QotY<9AArV@T|0xjZ#*AZ$Gii9 zl#yBKMhLazr~a+GMERRQmB?MNuLdkV=PiHdvBLTM@Uw4m_}7!iDal3L7<^@ae=UB# z!4~+Ms^2-oqx}`P=lg1VJAr?pAQp5Kb>g2qdSpXLBBsG{GAI;LIGs6u>d=DWU(C|se9kFE_MeOJ`y+Vek;Pd7|s`Skb#g#F|8R@CVW*5xx z45OY8jz@nwNXv`_&RfdMH+ynypY5juRXI7rMA-|IeRi;@7aBaz?xnSIa3+Uu&%6Y0 z5uUT^>{f)_gS?XSR8T=+qBL2I)evQ)TL!P}Br-gZ??>w(Lo7bY=|9>~;vX9```@`+ zjL1lMmpVX)+@NrGDa(+G0ld9Ey0;%`2Q5Qe8nhu%4oN5JzIr(x5Z*FR;xjn68z&62 zsi`7I3&t8_vtihoR?$vz)(BsR-+gInF~+}(QD0L!w`UB#;#z&ro7Siuud(bRiP9C@ z?~PrRQHbkc{G)=%?Zzn2FE0c6=s|?QB%==dX!4SPhv~|$MN0)T-tq1Y0yR(%On~{e za{&ezhOyOVUh1$t2_&kL%m~1O&TnX-8vxiB`B~%lT1U2_Q8FaV+-wZ=0mumwW#>Ho z@2(Q7?fiss>>-|4gZ$%`Qfoal%R8Qv=bouS(vx5ZL!YEvaR*zq)U57LlGAq=KQ@TDelvkBlTPT zc}6y~OH6Pcz~3LtKGYJWhUqG*Lo7IdZ*U!!+iidMoYPZ#r%WtHq?YTgQvl+Zsavqj zV&X5Vk1-+`)_}+EUln4ecG7M+JaH^GGuCjd&QA;PZE_pcIG!mTy7wuO{{={_R#RK! zxDeV^kp(=fyYYNm(!0&ln0rr6Z~$|}r*m38=q2k3&HjXz0IxF}NxZ~9O7C{(v*^!~Qr zc!>OLd}95O7!p8)d1w`*zUklnF9x;o$>`{6GHs%r5+$Fm@E&q^U>uY+gS^hu&F>Nu zwh!)}%ys|i6=`Z6o^5AKiHo5UD8Ayuu8HLrrL27~G5qx5E~NhEAP!~dLZmg%_xu~% z^JR62-kt69{TlUHlO8N#-OKkWAAQfPh&zh33EFX^za%!LrLRk})`)~K0K4aZw|t&j zKP6k#aT68lO$~yX4yQf2z-uY7!;P3}(bn8!zMP$1Dq|FUetIBYRwaTfqTe<98PEfb zMu`=vK`nja;BnPlw9LxHMhau^PA8*{zGy8>-pCN1Rl$Ih%S z*V~l$gZ6&`yaa-^Ab|z%HRy}cxxu`jE@6hRtr$34%B*)|5eeA52Q-)^PGDIiF0~O8 zfWs;j(*2*#j|r50h$xq5X8n#_{`St4ew@}}!~``S+WTEw{gA1NVu82e30D$RStbh2 ze9M_3HUe_aYPhH0o-m`k?%|q3$vq*03IzTNR~+SMv)gk7nJvC(YrYFnDc+27!4M?z zdb1MbZ^*)8gdDoRq5$xb*^{f9Ri&MxCp;l8vZLZkFDy(MYhigJ@wIV(0;WH`>M?7` z+!tEA-=}nPY)w`G9&>TOJY1)Uk9f_qtL7TL|W z2`YFE%ELmIamnA)!}>G3n+C=87RJV4(_bH^*4Fi|y@Sp=NNW)*zvBAD^-1_{UYddR zsO~T>;#O=Zna?i51$_UlFWzPs3*obkc$)FZ+*|$qU6EL}2io3T-H11r;-yfl)lvt= z?lbwgE|8m4=PQF1K!w|ryc-CQK0u!(OIu}pt{qnPhsqZAXW8{D-#V=5i;<@JdYYE~ zSTVUP_H-ulh&74ACOPx*3AqJ|0&5g_jpToDOCe>sv=M1!kVRSHQTAo&ZZWM)6de|!fYDi8J{i0L{(FCtY z7;vfsi+S9Lg!BayMxPab!%v$pb58d1rwpd4(}3`ws8U?!AZf)9mUK@;{2iXC`xM7E zx0TL`r*QfVbhXQgJG?c^6q@t2&sLtTJ($TO471tB8eK%t$aa=d(h0@JR;dSw(Y;@~ zkRTd)xk>sxgd`MfY&{YA$3(sNv)ny>faHj=c?_C7EwZsxxL z;}ikspV%e0gjw27St2S;LA|puP{N-j&Qe=TPtMmF{)!nT+xF1Pk^@$a(k{WlAOer@ zp;94*r z=F8)4e+672#XrL{AZO2}|6l)PE1A6FE4L-$2UH2qyy%HsoD;;NmUSF`6_&L%`zgVMOUlGAez+Ll^yN}?KJR5E z*)7-WRM7U|$FyL6!~E#i0!Za&6CSf5hI%PhPWj(NqKu84T0D}0m}6!b5)xTMoP4?^ zzFKkp1i9f&zsnjo;zEP^-X8o?U@XV9sozL{8)IPx{YT-j)Umt!A4=5RC z14b-A<3!~U9WCqU0mt-s7*|R;kNTavJI);bwF=ZDD@o3g^TV6_!prXv(UD&1gwr$M zl5)o->TFUBvzc^QO&hVL&nF5Bia&ee9N41zT3P79*4XT|BQX z=KMzzJr`!bik;$f{TD!-$UQysFrZ|%58g|M4;FRIYUmm<(S=9^5CagYc|+=fn{eTV;3vP{DqW@|#lQ zX;H!8$J6Hc14%~M1Z^BpS4hz?C#W^tc9dmG6BU|YMsYdokf}7KSQtj% z2H-9UG*rdO)f*#T`vV1>NrjyI1MU@Y%#fcAz3 zW~2OGa`#$Me+YX<3g?MSbtt07!cUtOpq@(pNN#TJ`lB@6DhP<7av-V0Dpl-BO``QU zsQlq49J$GT4F{-arCBVrcl>bT0LqH%)QT$neW1}x_zst4Fv?+=HT$C>{ecAB1~A@> z-(>ly&OH}G9!vfJuYs7VTQMd|C-z%o1Gat)UOoH#@pP*4`a%`<_Oqp-?7Q(B|DVQh zGNE`68k9IZkRpkc$qkI{(&|%%V4=Vb%Qk6>`JynBD`S6%pG=_R^VYGWF@Ht9T z6XDsCG8K?G$f64mD8&2VDa4#!;lC|AvkGaN%*DbjK@ zLP+()_yN;?qYM*>mb^CvlTUP?T;=47)#GX`DBWdK+B=+SqHwkGmR=ip7Au_`Ax=H4 zl7We2E(>!uPQJP%bW_pe(OjLnCWdDf)L^nRMrE*d)R#E2`^N18<4YCzb3x1wKdWST z@yNr7OuNMp6Sd_ij{(xLv>l#QhBK>x z=}nE+R{kRWyO639i*8-$TIT}0ZZ7p5`c$XcIgIoAW-m2o)~^ zK1vLB&5~pf7)=9Z+ZfIrYO$A|H#CzdyB=aVRBdb|XJ$Q?wi!`!u=K@WKwih;{d?NN zs3STF`!T#TEAp|hK^fPw`ws1A3~EQRaQjOe;Q)oUOqDFO~w1t!yZ%ry#lPELG7rd@TD z+~s^kl#Lz7fje}o`YtAtPaL!HpDAWkCAHpLj6ZoR17YM;{b7YkQCbi2vzQyKM#I%) z)boY!w=GqNy*3591ssmV-tzN+H{K+!SJq~VfVL)$*%JGNxQ#st* zuU#cK?EUe5?d{x%Iz&y4-cdlGGmzt&89f06qxniaXDV4{t&FEf3byU~H^syLF#)3J zi+ILIhg3n=(hb^rLvX>dw%kaiSaUq5I44N(O@AN|gBeqfp--5dmFXzFT zeeH)^#S5R<2UlxSV?!G?9^O>fR4P|R&?wde4CBCtDAGQX(1tPPoA+vfo2+)&h*Nr90EY9h-bMTlAz-}x)o}HU`HDDMAb#QAeGAzT|Dh7P4O`XpeFhlo` z$deXfynFICU4&QhF>{KAI}a0p3axET&e8o6{r9!U>4_}gBxh; zv+>q~{iagMKWux{e0T5DqU(;qG3wNP+z2wY9m8H_Bqqdg+WG%l?9?B1y5SQ(+78^2 zJ>tb~92yB5ie=bU5@_WUQltsa4FtK%O14l^{ZP7LsAGH*mAa}qFiK(X!i(+-HKV%i zhc1>YE=Ikl6ajw%RaYJwg`PRXtpaQ_?f{q@vSCDNs%V`B@`Ods?pFibQ;y6TQUA@*WMB3}%vLcG2G+Y!c22)YFZb+1F_d zEH(4{=3Wq4RokW4ZqwW$mV$5Bn@4))%N(o~v!7i5r2NFP%-Hnn&?Rd5n5Rq z)MI?gnWEW1WW^LjlMK%#j-QT}3nJ35FekS3k-AWh37IgJRz-)uJIa=OA-zF&IJ$3ex1Ql5f_jw4@(91FR z@Ipl!MI`WgwcJEiAGNZc{V(J~51fh71Y=TxfB6v{=L}4?SvNN<*5tR~je5>lUiUi2 z(l|=riouDwjxNn8RxR(T=mOT<)BjSz2`roP*=IXDKA9zOA1Es!$t`Ka`LD9s5d(=%zp{<;BVDkSYl*EjM4z{~jRxGOh@- z>ir{oZFvDa=YX(r^dg!gxqICov_U@|=q}=&OC)Km<;ZA&X_8_iRj%rL^R6SPJ4R=5 z+26H!2=be>SOF^Pq24FuL^FN0iVK85`uOHQ2AT7Or)!4P>7)qEbgYYem9_BBrHc=@ zV;9-5NfSJoJNyNu3H@VXVklyKMbAXv%g zIq-1$bV#IT#|W6(rT>`Yr&+$DBEz>4LX?1N9c3(Mv>yp^H6KQofw(OtUe#cg24OJS z_8W9x&c`$`N{YM~=u69a-E*B04i6OzJcI`;{eD#k{@CRwf~U|idnn8s+H%Hm^u^yU zUn5T+?|x?z^eL5ml`$nr5AcED4+Z?}Xx^?gqKJV#>W|cE0W=1q2<%T(joEf%BF@+) zqNDxqojI9Qi2}3@g_^HxrGUSj_G<8ae^{c)`(B|>3tCQ=4Sh~%XR66#TFy|!Wd`5# za?|boSq<2nEUi56fuf-{Oft;uBsbO5P5hicnrIn3Xd>m@*-t;^L+ZNQ$#d1kjHJ)rxL@~!zyURf}cNl=Esblt_PJ>`i8Rt{gB zePyS#v6sB8?lchra2~MBL`Q^pBqCk2u}I(T1>ehz0eq&?F1mzo%D9?rxTMC)40R(q z_R=OGX32ZN__E4A7o*W2?!-KE#SBTw2{V1x`O>`;%S%(HQW5@a5?C6p-IB(H#zSlb z6eSb?9&EY|T3}it!;e>T$|Boe9<;MD*$xg~zaRdBXB&k2XeZKMQKRM#Jy2(S1$q^HQ5_}9JT?23uwZIR_e%*tN+Nkh*~B7#7hY_|f_(%qL(11oW>RJ;_V)hJhFD@;AfW%UMU*SfNd*oc9aUF3Pn(&31etk*S!9c}&WAy@* zV^UR3J(psd|=!@2ZmYZR(24ns41ibqkCcD4(@yw0g42sqn(ru(6dDn9We z=Y*N$Ap07+i}%f#g6yx-7reKJGYPkdpD2w&{pA>9J3S$!g`jw)VCu>1*kgsV9GdlD zs}(km`Q{qCAmarWB~VRz?(3pCmWsVrI7qDCSuVaNW1wz911$~`1N(&?SNi*p$Zvi9 zl_w8-Ba-_Ukc7D|jN%$V%dNo)4|l@Y^dgtd8MUfg7rI$IOhU`zACJ0 z1sfLM_uw&rg>A;OCPne<$ANw3^_PXZ-XPTD1Q^a9EC5_c>YjlQ5^RW;~KO;a3dTsq%61AsV1CV?AsY zqMVUB*t9k8RxZ<)_2cK|>1=i!hkr~|Duep+%yG;=9m&xeN8}!-yz?Ic2NYU$#wol{*uF z5%?kPm2?zLXVtt$m{I;qxy`a1!>Ia-{O#r=g*&$_h)>c z*|iw844^5Y2Fq|h%c@A!A+Q=yS2b9WF;vS|XfdL5npNuej$JFDr1f@Q<;{pc-*a-j z6wm^$9I-`x0kNggkOq4*2Cv(8mLm}riDkv<0y#h3`A#$aX27f#5BmFjmiJYDh-BfO zxg*r1NXxe@%VrPhK?8^jp6xb1vm(_pkW^5*@m~XrLv!f=e|dskU3Zz|Tw@gw-fjRcN{+5v^gfV$RN>A+9-o zuguEZ2$^`cDER04KtUaEMiC*XL@$vQU`%SA8xvWZrzrWJ2vb31%ukl86=lF0?E{{I zdDMgVF0+99ZhP(?@px`5`OVaWHP`)YliKEj%#w1NKzS3V^e~I{&lKETiMD|)dc z;%C;8I0RazZu_Uld#9LyeSj&(R04!4taT_v<5&)tAqUmqPtRAnOss@ zW~)aMC3(xxFMjQZUppDA-os#>08_e0sfW&OMhm44oxl^g@@x|%Y{!id_N_r;YVBh) zhCc^NzQOpaZ%91U>m|2cG;kYoLN7NDQuszAzZ>A@Xc`?;IvJ(g@9l9!`$}=}Y{i=*y%^a$6YLrl%Y&*kFO=9}8k= zLDH1CqCkv(!~g0@0kiZ{JnzEmruO-4@Mox_+-YZWIBqO>Aw-4&TK|niby%KMa4f7B zucm)sk?Rm6R40)?wi>zBO||>telDdLb7by8q2rK!tElW?q zjve|s%N3R^9%8xX_*4Z`s}Bb~wAa^udP%hZX_MDBqEIUgMT9n>ju_Z!xxY*{RI1m( z&xuoSJk1fue=m_-3M$8sZmIr9TBey^bDX!pdzt?$4s=zmO0L!m3_0Gi#w3`u+S>LX z;1a-Y;{owk2ORpBnJR|7338T}W{*W4)aKzA7_jx>8nwO0{cu*%kZ+vKG&os}m(jNL z6R?6akPe*sp zeWD79m3T$LkQ{c=v2RByF8JJ}IX80&lUF|;lEtV>{Qbp$Ha~m0RWc~8i0>L$fmI$i ziZ=eT1SOAjx)E&{L}p_U z;Ox+S0H7WxjpR;}Rp-f!VNL$a_;le96BJ%Lx?0NB5&ijjh>@q;-;q?MSn1MbKeJX0 z-x;Z1_s*Kc+Or1X%%fYYyAt^BY5JucNc#JX1<^<$rx`Uzow}N_tTD#l!{8#wLf6}2 z5D6c|A&j|{@58Bm&S@oy~1bNEnuuzduHgW$%b7CYAn<%{4blk<&wJ8};yI z!ravNWREc~aeI>jV*Gp6q(5{LI0Ai7QT$+C^wRLV{N9~FDHkx`(*-`F;oRl~ArzWn zB16VSo`X13!<36N0wSk#MW%XvKL}#iPz|{9_@C^t-DSFtXV3UEiX5&x&zw zCB8&~<&7drR<)o6{N?j_u_fh<^)P|W2)+<8p{VXST98hciA!hlT_6R@U;w$oXnCF3 zhr_zaQE((D0#ysb?(OAYqrp_dNfLDtsw#OBpcwH|07)6b&#g-*9M%Ba>1sLOk1d=Q zi(;&;J%~*&RBm2rDOpC`Z?v5AOzw&?UZ|JCw+w-i;-ourFn)rXwM+eWG|Wzs7%P6g z;1-#$T;j#_N*|d>NFWp_6c&i-Y7QHlv9=hpjm(p_YCjpQ%Q;_OMi6C0)~CBL#lWa9 zk=NP1KXGicv}V6~^IG%|Q+gS|?uPI`_uY@cdr!N#TuK|L#^2)Ck?O2G^0CC?#8UtG z_aH_LMRth~w;F*5aT1gExZBH*ctsGiGyQ&YYa{s};t=Wu`DVw5tqzPeVT^cj%7WSa gPfdaKzbaDJ|5Ftb8w*PT0D1HOa|O=7uUPzl0Q*i$z5oCK diff --git a/apps/sim/public/library/automation-anywhere-alternative/cover.jpg b/apps/sim/public/library/automation-anywhere-alternative/cover.jpg deleted file mode 100644 index 03f8840563e563a4dab21635dc0b0048ce7dd675..0000000000000000000000000000000000000000 GIT binary patch literal 0 HcmV?d00001 literal 30429 zcmeFYWmH_-x+qw<2PeVZ-2=hh-8HzoyA#~CaCdhN?he5N1b24=>0M+hg<@ zJ-UCrstIe=v^77OpU(AX`Og|a7y$kE2L=Wj4j%MHLPS77M8f!h4EkaaqGO?h3L!2r zE;i_onSz9bg84HeJ>zE)US1J3MKyc-l=n%%{$Eeh2S9}dFMzCt03!i_qk=)8g8dl) z-~qq@;1FP7V1R!<&@hltun^$ytr-7n{oft_tO5`rKx0rKP(T$;J^u{*k8hy(jc{}S z<58ktMW>3PrYb9E0FX=@mR4Mf2BtSZw*^Zuk@D^Mp4Qy8UT7F$553{vD0H<6T(YqI z|FcKTiyhRlARvz>B4nqv^#D}Wsw;yulbQ>g@7X8w6&ELEt&>!5gq_1r>~3DMVii(M7)W$umpgj_9qSIuv-KY>nf;5w zr$qn%XvqKR{C_a;e=zX>4F)V4A%7LG`{USveRlWq2Uu_yu!I z5zTOj6J#3{-iLfHqcxcZRvm>N>?!o$`u|dPHrEoR#x6E1n$h8(Jx%}gQLCa)Xgtgzc)Iw z3AUJcq6=TBlKg8Dv#$sxxn zn8d*^<(|1?dYBI*S@$Z{-8nVR&v`!&lvvK~aKDbGfj|zA&H!a?^|wuGLru z{m&|Dm&}WtfP2G7-K%X%w;Pu!m1WMs59HcMvvnmUD(Npzo@ek3n?Y0I0MbT%sTrs` zsxq4b^lW-yQWGoLEpy~A-lh*ub*`%k6X2{AQAs{4lFLw)0Uo$x{7y9t^wvXoDnDVS zQyh$mG=@awv*L8yxm;L87(G*-xE2l0pt;$1PSs~K)3|22+9%WgCveTF!fzp4i*)ws zSHl=Jkz}ot{|j0~2G2&C`pKM|w9e9*mi=3wEd>^;x7(Ipm{OausKbcB%@QP$3{`GY zEsw+UXE7*Vt_>TwOsr4>Q|c+#J7GI?KHDP#Tl(}VnvT_0_2X3V zi(@~M)P#!5Wr1)vEPN~=RqI08O^r}0N zmoe@kRaRPcp%wMIpsC0C&6(QvXDSdl{H6@A3Iew>lsGC&BC}BSm-$Ng^)WW$j^OXRdg!#dk+`@v9AH2oGpPLwWI3B7OF*`U_YV?!bJy)6 zZqAFWdZZW?wIi}E@Teb}$<)a+cdCuy0~&SYgqJxC+zfiwPMDz39%j2Uuu(u zi%EetX@%y8%NWgM@6^qLw(enr63zJ37HeE8YPgtYJGpYj|9=2? z;Vjrz0sw|PwT9$cF2mC$SiKxJ5qk=N4!UqB##vuYJDiqMU*1sPPz&dOK@J-z6*ATq zfOPv}PDp6T+B$`(lZQSUA4;JS<{=9W?GAtm4U}PW3b?r?1&-FnN(ewRC`6rK_k8ly zWLok)g)|jD23R$du#&z2z;a(UkoXYOJUsuWW&N{K(m7PEw-pk8qJL`spWHwbm^jPa z>&(v-4?Hq;d0flr0se}1!>p#OWr|N6n< zJ@_@L_Nq%me|Q_(N{d({gI>hfhMcnTOnevKAY+ujwN zI}}1zP^NCMc|vKI#mt!0BpEbzT8WXFFqM@+lK%(q^`pYVJ7Uy}sz)4U=Yrb<5vp;G z4S#APu`gr5;31x1v83LMxLZ7&PoQPY_qa&~J2b?t8^!rDeO*%Z>g2~)o)Owc-x zDB}iBD{}sgm45?@aqBoyDwk4NB}JAkd=?hal&b1DSJHZP9>&UbllyepQ^&seBqz+x znHkyO2meCLg(o!6p}^%^`EB#}l2;ro&-Q1Dn<}tLH6C$i9f5&Lo3t{?71bKSGtjD?UzT?k~_lVPCUyd&%#rsu5aG-H@_50%a--uBUfZ|H%L5nI)Pu{gufDm=$x zUi4Wk{u)q&zDOktrrv+rpX=u2U?iEnz{E)9KCks)BeakDse4{dedD`ULq3ty7~bp( zj&@q$A}=Z-krVxWaNz^+%R+PyH(D!E7Stp!9oUC5PsSXiq!Up~R4Fyx8nTy_w1eMk zSgG7IJ;{x%D}%v?pBf+|+)%>v|mxQB@;xN+`vG4fF9s47F*^1_mOM`qDB<5apfWnn0uu zzG~yTf8%AteSR#j!4b}j+A=@Xi#sq}1G!=Hx9b7m$St7hK zJgbj|sNTO`cySE<#))|rqL-{M?IbS^_=}#YtI4{hiXJNtMtKCjp%I}?$N4AURxsOe zzL(-Y{_j}8)GaHdl4iZOvsX|P@j}TJS+6V~1 z(O5{e6uR)^NywXBaVNXQP@9rZk`b7_fML6+!(L>NUs$8(APXUI!gOJ;{uz`g3$x6DbdofW2lQ+!P_}yg$bECV&}amYcrEnf6qm~{NLgK|N9p_$e(w# znx6%qNPEEo0Kj(60!7Gmo}3c?Sr_J+kEIBL``-)0iIB6JUJ8(%?#Dm%sHLe<7+8v zN_J@MIdDFF>ut!GWx|@9Nk?%P>ABD7 zB`%RHK(mLLUQj3_uccitD)NX!FIawpbj58ci%ERBz9W9t(Fkz(q2{7sc`yTKU}59j zZp48OlSVMF!Nzo?LSV11>GBrWz>ZVNvN&GA3zxG|&!_jI1^&iwn9a7r`Fhz$SH{nw zZ!A@xBY}08CzdkbJZFqMhYCyKfGL(9Ai&o z{1>V|I1P?vVxAOzW80`l0kQplf|vDDGb(O@Cvpx#g%3eo=7rPzM=r)1Ynr4GJq%Mt zci~(J^!%r{TN+-z9inEZesQ5q6>2LD)KTm!A^jh4zw8y&%O>YWvTEht`r4}K+Gmb% zg>tTLqlsMOtIs+)XPV?Yu--jfkk3shZrW77W zcZOlgS}zPsJU@9R>j>$iaDQm5KN0FaoN_ws70~~^AXCkKK7fhT1xD)|x}P0GwO*;}8bw5}xIvATir5vhst_i5S~vODp5-z`s$3M;8Hkh@`8^#jYA1u9xZ~ z=+-|$E1gl9KfEAXp`6ECej)-6rfSEEXmU(d>GgndApI6x8=WlcHU}fNJKw!rBs~iBBo8v_xtta{~p#ML+1BD?uPyk?%;9y`d z5D*ZM5C8}O01^xw0tx_)!VZIqhK_-WMa0JOkyU|7K$Mu2>??`AKY3gaC{PIt3SWZ3 z0{#F_E08u8g*NhLw5lepMZf$32-&0zDF?Ild2c-M*?C2FY?{_JtP)xPOnuDs78cty zlx!Q}&-dU5aIt6y`fsTV>93*W)sX~H;m~B^%SpJ-jC3B_-{Pw;>Bt16(Il8Mq-K8p zBJ=XBUyllUdvTG5T1OS+KYBH_R#{l6k}#Yktr5RF=S){hiiBb*s z>5CW~NxG+Ns{S-|N{){7vx4M@CzLH1De-4&LJKR42Asp(4DHT9_5`UDQkqc<@Qri1 zG*qJ+T)j@NY;Q4ujdwV%_T4aMx>U)os-9B*mqk4l0WGdL3rKnw`5&t0qNNwGTu4SGusS3h^J()iFa7 z_i`9c@*1ls;9%1G;~^*i0J^O^=v1!gQ2a)DG8dal%eEyCHlM$07P>6Iby&vRLfQ`y z7VCr_&?%<7DC`UF)ZWx4IsRwmzw9_uv zX?FeL6ZdcThC7Y&mrpC3Pq8oe$0@D;{6%P~IkVT*Kc8zkv9J`47Nn|nuIE%$P zGcX=UU9#)|XWx=}hf8?Bmmh$eBURk=5o;~)hdhQ=)ND^0of>CT zq9kb@mVX`k^|%*{_$eJHCUzD(HbP^mU1rg>tq({v1>%&EW1@V}lDFy&bttoo*5*G+s`(35`m&LF@R5T7kZT|qM0tZn`w>733g;?fBrsyHQ0-x_aY9znZ#h8lt(!Hsj^fz+cn`EWUv$5{&4jV^|Ge= zXO+9L_Mo)E3-BbZg?&5*VPDbuM{3h3FD(j|BAG>WPtA13c6F>Whe?5I1DbaP8>VW` ztG`vn^4r!v>@hRAH${~1sp^!s^W~d%T6#HD3tKRm`5Jkjp2v2 zTP&v|)=gt5i<&J|a?Sk#5Pc-L7aQ(+c(HIrlLUCH>uRdcjx31yV}TH>iNLH$s`LomVVK03=16%r*Oj6q#SOqge5MZhP1DF@uPAyotS%thRvKPLR8H~xViup;$i~r@9t)glz_h*Qd z9?m78LhQ^bPtXC}ZyPVnWYH+n@Efo7s%)VOe-in94qRlX7P{a$+E?9{Qde8mw}^dB zZ9#*xl-`6+*Pwfs-%7WhIVCQ&cPS?jE@OWHj;$XsC!fEMlW9Z4$^;~FF~{*gnsS&SRE03CQAZ#wYgzu5TT4zwxx@@*%DN5Q67Wl;$4dnslcr#T6!-%4vI$FXT=0959j0S*A- zD;Oe~*B^i^fYT@a0CahkpRidV6?TVst`Z?x2Q&TvG;kyIVwHanO&Xjv4vwXl3=v^mwS)Gy zZ9Ke-k}QE07HC({#Z1lkPJMsxc&m{%b}2>w7e{HbNr-)NoXlAKmVp4#P|5%~iPgWb zC_&orHQuD7$<I4IgEU}p?GP3k9*Qri{2&phslr*;g^o-te<;bPc5*#JS@fUBssl{R;tVy3_=t5VALRRxVyXAz=;DR%NZTfcApG|gbUo~n(* zj9dk?J{PKZz%j~Psj)OVqd;zsuCqDQK9Zh%ermt@F+{cp$NumBEMvUn-DT4sF?<&%_;67_=O z)gcmt8aIANdG@O(FikaP&6>u}i`V6(fI0dIM`~47Fy+|YNJAsqUaFQ+(!Ka=evx2< zZ*(hN<{ouSUuXiiWM(UQWwi%fB|cx5bIXemu$L}~wLJOU_fN6eYu+BZ5(v%*{{BY@ zi-Dw16ulQ;mQ??Y?uVGk2=dVD*^0b`<*MS4b-MxPWFx^rGufKSc24nH!p&+`Bzv!5 z(%ZKde)$upzc20$$74lndhPP}mj*HleoprN73KWRafHIPV=BxFmFdL=`aR6OWFgp`KG2NOGl z;8pF#EZjies3cKTxqp4*{)dJMc;pT)%H;7=!2_4 zu<5MCfeW_2~+8wtY{8*=h=+2seq z;Qmr&h!&z~U__OQN-;0p2?QhKeT@+dal;)tIO!rZG5`4aJwZYlw5`>Tw>1SkgsD;e zbCtcm^nzv2@)r2>fA9IJagc1-8(W{Nc*PfzJC{Ra?I+Wu2)~r3V=>1f^w7B$^Y7im zi=31>ULhn}K}1g^5eIu-xKlZ_Xw!;V-9nYZ0mUIaE?nlRO;nv(r&#ACC|a6`BSzDP zNABx@+4awxPeXekwKp~G=Q$tbJJHSdAx8dVa>5Udc!*im@(AD&nWTrt(iu+s>O&gd zr5syzgoCv_!uPpp4_jr-L_x?wsgbXXR6cR)mEXl4VX`5iTiBR*b`IEA%(Qq*9m7Ps zkV1H3BzF=d^j1h@Bd&0NjhX`8DibY%6MOM+dpE7VTt-8wkB^;a3)zcqIj|;Ea7>9% zCczv50y-cYRTkkd!Dqcl@KZ*o45;c?P%8Ozberii0!`A;8*LmB2=#44o!{>S;+?tu zQfmt;ni9B>Op9um&%At7ExB(p59odHbyZ@MXyJ3|HF1$1eLLd%KV1$&>S(BfS&l5* z_1~?>a0uf$h~J`DDyH)k$`Q25>IuVPT{ywX8jbsDd3lJFU`S$^dGDxEluIXirVO~NT2oY6@ghh5gNh~s$5eC zi#Iwx*659}de6sP)xI5YSsTpFNL6yAg6!rARjv%iV=Np3jy-XS$q_F~Q{K#r79Y8> zRY7~ARfT*+<}MwmcmP_tzgmboLLTxr_8QBk!Q&*cF9R0=N{=0v^jdM4Rx7ukJ`G0` z4gBWU?Fqzcbwy-^0gxxbbNL}JId+%A7fbY1(tt&Y)L@)oWJ$O zXPLQ~jMk--ynL~3I{Em(;(2G_yq@z9Kw{MjlcU->VSMG-?V@w1{Y z@ci>q^zC<=YOqAApj&p#oxLTE=BerK>;4zrh=GXqvT?!#HxJz}V>OWB{6$T^{3b*3 zGX>8a?@p)wF8CjSk;@`|cUt*wArKnkgxE=n4b zWY~2OXm5T(urfbzPpNTef)>jI`^4*Co5cT7S&COSY8KPY(wf0PGxBSDDAjhXWF2=O zEfgc;11zo1uOM2(cIL5qmen~<_jL7+<%oNxtr)BxR}{uC>O1#37WnY|1turW9z|~g zj2~Z!8l%JvGGu>&Y-^RKM$1RwhaFd(e7;0O=(F#Z0$g*va zCpWK?QN%eV^Y7D#l-><|!1XxU+$deZGOpf1rqFod^tR^{NXGow=EcwcX3)b313w=`rmlAkp|Ut( z><%48{54d%w3ImCfIiZ;!$pA_16#77@Uk|N>sHHxR3&0Nx7={S<1VsAIeWqHzY&>% z_bfWrD~3gRoT#ix*AT&dkFJfa9j2jg?NtI8`ZLCazz9eu zW^GNNB!2#yoe~36{Y;o_$Shg>@u=pj;l<$7S6CYM@;$L!@LWMy8WRvxOd^^}+ptab zZ3&4|z?hj$=?i%|`BEFzpcu=Y-(CcHl;OaBjl2VIVQ_Cp52CZmrapIh# zX85#qjvK>I>^#%g)`W1exvDcryiHD zvFx>@;7P>ceOCU=9~WM-J4BI-EpX3te2$=dYx$mbXck>)-Z&gpD%L(SBp@KPN$L+S zwv(^tOMVuB(S;YHIC^~2vFvBEBtmUl#h|;cFnz8D2@Y4EhZH>9fy##jQDBb@_bFB3 zjtgAj+$bl`3Ijt-v%1}@1RaVSmO?+1AQ7awSkR-DBCm@@{*|`P?lC8Xv0}G{QFxF* zub9899oTb(9WPW?Uji56PdST}31iP5-{wo$rAQE`X65+oeVbi*SXnkRH}Cp^5GS-v z8H)8A?Sc=^EeEn=I#74NLq}9s1swJJT|k4@NmJ&Tew?Oh_~Pd6U%HA>yllr(R5oz@ zyQm}if#X#pS5oj%`6)`u%S%-%06PvO2o}<%C4f$ZvLK-TmKF`rzmIWQE#+eFi$G zM>{r+?9Xcn+d`5Jr@#nn;q6W8cZV=KeZUJt2<6(^eC}DPa#@{ntr?+f_=mHjCw`fE>~UcCJxT5)~SYZRb|n8kBi8uzC_9!M!}lWFC+t4tonV zq@^qpBStHvTKA^Cp1e%4I!?)B)ox~69i2Pt<{LyX>{)pH);;R7K~`3Rs!~_zCWZmUyo&`tqEH*?rQ~;wpP@KvqhFzzO8* zm&%%$P-k1K+$c-|J~3B+6(vkPRx9pgKin`W@)hJs+u0_L+G$dqrKW0In=gWx!ylM1 zGW=5#2I$sGcJ>?uXK7mB-RQT9fAb&imR!Ub_+0K?Wj*3Vs76Tym*M33+~*(@E1VyW z1=>nD0a_>(-|vX(gQ|Y`p??qGqC-dj!iW9`fD54>oyjli?uE_{cqK>a`mGCMkJFb6 zI=-0c?x}vV=tOZcC2Ka;xumk>d($722hbUJ^{gFq=himaF$^(Dk@;&F_QEJLgOJ+q z0T%Y^4$Y|_Pf#i5SvH`!{OQO{Fb7bI69uup?y9=Q>zu4;YFP zlg1LeRK%CP&bkT`poK*8jcoBry)>I^$SZDW3vom<@mF;$m(g{y4AXvE)#?;YQ?J65 z;@pnF48Rr?lzki=^sJssw9z))skfaIp?ZV1DtldDSFQ%_mxQV16ORO?I~qxA`LHrp z`z2!>59<*B@WJ#%!4FRkxTy77*3F9(8YPG<@)XRW#T7C6-Kf~_;;C1LiW*DM=k0>` zuL9ZC<{`|!)HH6PqEVH=IY_#bP-0tO9#BB+^|SF*P2fTH%U=4aSAiG7q}pvqtCW(f zl-N&*E#pfWwmD&Ty>T0*fe7#xt9m0RoCB_Nr>r6cNti!?B-(#li1@j$S$;p>S)<{i zvv&Kum^O|ab01m)a)g=v3whBO?^eo9o5E;qym=X!dRdOF8a|Jm2HJN8qu|nBvu#(F z?v5^|Ki^qYZo-5by1{{@nHtk9crMI+?632c8v4?y*ypM;XuIo6g(>8WY1YpN%aB~tlb zSf!?TUb2Q;eS+oQTi^4+3OIjQM54Y|#O~86(KmMb80=VTJwvZLx>GS{CFEs7j*5zk ze$@2ixJu`Ukl$>Oq^86)z_J3AuF(R-SLm4MBO=oK+tMA&crp4^deb@EqI8xxd4frw#!jC<^6D_O2^@bKv#MW= zX#Z6joVao&v3FZJLytMSuWI4abYVTe_rt7XX^g-(YtKD9m{@ejAn##WJd$7TEzwzC z!@zjItMUb7KzcIkGu9QEN$Se&$ezMVt~PDruzI07>3s+_28*id#0SS^r&F+Ba720d;=>n&@oqlv$ggFsQNxQjfU~|GJ%N8ex;CooDOY z)O{~4O%yrr@ea*BF0nOaTCj@Bh5X@$HH4z4J|F#cyM<9Se|xHTb*Lt(maMKXV&m`$&XmCQO{hx&wq-n&7fXsam$v4$A%=6BAlSDG2Yd=K=Ru%oW_?hx{VnMbsF%DpR0_Xix)u?_rqU@>&4hl$GZlA z6)Zv~(5s?dK_UrDZUAaR*e2~VS1h5h6J>mT9Zl<3K~|F0{+Kjv2+7S;g1g(9i{?{D zkH!v(kMQvhq@9Rf4R??eg)5(Y9L#Xc%u(5oXYQiIB9~HDnzf%w(@67U{g^L|+`U$C z@g7#}3}PMz5L(7d7(EeI!NP9uTsZGqkp2Lo2dAd>Wh&4QHmeBst_?PA(spkwT;mw< z;$G#wG?sYpOSs1d>VM`hGUZD%STW`3co=>Wt-ZZh3Vq{{E;AntbpHcTkQw>8V}w66 zUOz4Vc~%7aTGI@gwfRf91z1?xMKg9y7h?uuBx0RZPAP>6dC-{^H2afND`YUoLbXYc@P+FS#v9YQn*;GbbD#+Q1$6gAfR_s zsyLNKO_XSmrmGOHofJOqEU#LqKtP4`_03HO{@87aL#J!wxr|2N&+hjNk5N;vesHbW z3~&vg;X>J>gJ{5qMQ1oARMn8I1kr?S4IHbIWt7yn(JN~g;d}OQNqI#R{JI~H{zWv1 zjnq2O;sOgP|H(V5XJN+Ke?*3@Q4u3r^|@j$OgB>UQHeH)lLa{kZ0w=U17HXio<0i_ zk{`y-Oy_2d$!wH$>r3m>0w)*%P-$gnRS~W9yYk94nmRS^AfsZ}%uBw8q4wJcy;RSu z2pbFRR$pc5!AGrn3lhrS)kP^90Ikf8!0fpKhagEbnab`4f=#ZRt zJh|GNmO7Rmq;G9G2tt7xYrsjr7G*fJiyv=r!L?rpB#2Q^Q3}O0W>B&0UfCFkU(-OT z%nRD<)Jm28B^BK+XVP#hF(1zO;~k@-`1qDzAKPjmcx^r#qPFO_SGtDLRAlctRorGM zlUUCa7OcYI76rX(wUwNHs0|The2<}om1iCLA1{dyB#N3z8!7DPMS1Ob=?y9&W~my@ zGwwzchJ-4hw>p;5?UvQ~r{w3Cc-nr9tw0TmeXG_%7H_4Y7$#A9+;Qv0U%e*ub&0sJ3fD3KbFB}6{)?$oD2QE`E0Hu{K5pv&u*yiF=~{(pH_JM zxB)kL6vMavv?N(+uxLl3#|a#QGk|kY#W{XK8MA$lTd~s|R@xdGuhfP0Q(GIknqyvZ z6)k_~y!x~i#29fvd(z{$a&qT7Eh&W{1;7QHh;aIf;y#2Nm~DC(+}-=!yYlEdd)b225P{~u)DiAWW^+ITX@q9>^7j2W z-9M$!I;y17I^)}}D_3C?)-$HwU&JX61KwFu;$D;J!~*BqTACtOKzX1q{37=|>e(~j zO>2l>58YQ>7IpiEhg?LHt&-t;$Em~%hKo1gBgBMFhnvNkPSaIQlhshhiOUap>&O2t z>^;^*9o6x_GqnMT357uQX}ShUao1*Q+p1|66&ov>`<#__??5VUbs15=h$9`h$beR@H9+|}*P=Vx6@Ux}&3c1$ze_h3Hgj11vllJ?!Oe1743`SO9Gowo%?Tqu_A+2-5( zHD*RsAK$@3$xwX-QmO7NzRHVfo(P~BL4*JtXOoxw8`Cb!Pb;37IXKLf**SAJO?dt1 z5#F9(r$`>n3jdX}{0Gf^1Ki8r*Q@(dLv4jK^VpjQx<`US*yLcqsA-@-rcX~fe%Vg7 zzZYp{kN#lDo{i3>fHC9;6$_%@7tI6;M*c%h(@q+j5vSN!x3`e~0nO+?I5A&kps%n8Sm5`x61M(!f+w1C(@qAdEQW|IY8%ER!+` z#frdj_}nmCSvq@By16vrNV){wmCK5qI(oNvcK>h4zfQS~;4(`ks^F}-(ayC$47sh zd?xC%e}Wn)M?`Oxz2DU6Pj8|i^u@iF-z6@#WgZ8BQbvLMcPMsu$TF2BuWD*fQaF*yT)r6p_Jcx|_q>h?)nBoA6FWM1nEzy`rlI`JhEpW>cTSJ zafi(|%?yHLQH>^enwt_AGN>99PDK?*cC9$aOblt8k?Or%sb&z_2B%RjmDR)*^#M`q z{CIUcT8FBc|I*Ta-Ly)}WMtnOe!UQQ{}b#2{9yd`nalV+^i`rtsKOk@N%t!*Vhxt$ z(RUbF3$?gV>v;Vl1vI3fX<8k;!K>2M3*3^98zW_8udQB8gR`7YH; zKhm%;4GYZ!uH)^{hjw09d zk5h}jv{A7$m*dTu4T`9oYgZAl5ImhwE-^y8Ic?{5q;Fs?m?zM#&{sMPBU;Ds(PMnK zqX&P7DVpvMFj3@#D^_D@QZQCizT~R4hKeB5J!kA)Ls53iso18~wCdAaW0&Dpk}v>G z1DsTQ^rCs%8|VIYCY&TWIaeLx)tM8mZI1isq;{5iF?sXFrf4heRN&yRl;L>jJYgGk6P|s8tM0Ed5&R&Gm#Sr`Yx6*xuIQO$ zX@5lq5e5wR7v$%{j(oIp$QUE%mtMSXXsec3z1^}|IIlJGto?Y_67M?!%ZFJ4+|3Mz zlJokqfbsIpfj!H`6ZL2wF8mzijhQhV>HE3rFx%s zuM|)6=EOH9Rv^fpmBnyD&s2t0y91QFaK3PskH8G)gS!o1McTWfD*b%x~yr{klyiZ0{80|IeXTIDXKf)_aw3 z4EO<(T|^CsrX%oC#{{qjmMgz4rGRYfZ%o}4=<>)v=KJql$oSQ0T;j>p>qfv@KW@E- zvSU7hpKsK*l4Sxd; zPBTzf3ZXR%4lUyFOl0GlWD*@$5DD$4FA-|O51e~OrGn@7Z&Q>5n&^sU1zok06T zEe-2Y;DlAmlolJ=R|~L@*gh2)OYWHR2ry9;BFWR0-sRk^5 zC;x^H?}V2q4m8Cz+8Uz;qouW)7+|oCPm5f{kw`n8;CM!o($!aC7~0cLkbimnHuQOc zBW#$uf@gl>Eya}0$*?FBrg;Ec5)}3t{S}LvOM@)cBvm(ZV;82e%-)059d0I$8vLrM z080>ODXEK)+8K=QbO_#gRW0lm6AZ42!ZX>Y@rHv!HIQctkM-jS;)scl-DT6C9AogZFbwhoJ;s=M{|`FX1B^Ir`-zl{J_tO zLkqUdiik4aNZ3o~zzob#(p?S{f}_$!_eV_uaZDojLhZ5|9;>*YI|Xs=d^zXpGd_O+ z4^PNZ4y87zG0JKlCF!qp6Xq~OpBG)jdY@?XHBrdYd!_a6TaXORTw&Fwt*kCr6-l)2 zg~aMS$%Y=924f>^6f4SBE5V@SNaTXkh6ZWHjuU@+vclgCwIOYAb0HKQz3utHY=Yj@ zMR^>?37!44$8HAn$eEiM^7l51HZzYRO!(d))I%iMvjt(aAY_i+U}wA>G=yYplyVdY zCZ(rdFOGv06fUam>&hZMGaUFg?qY-vpZ(hO1s_0oDf2j(AfYteTb2E96F2qQ$+@P* z$cK)lV)Tc#uWdC2ix|@BYjZXp=3o%0P4sFCoX>Mf1iKw17$MND@d_GLnpnG+S^EVu(pB~K^X;}n_Oq}4!%PE0+AgkQx$K)$sgyc{>Nfbn zAHdZcU352>szwi`LL2u;B?s%e{G>F=g*617KSYhef_wm8C7EG?7vMf#!`iGh2N(dy zCP@;y3K1Fks2M>GJ_cqe*%Wg2cUyun=(WcpRd=6pgK(v`o!@K>aw?e zE&Gjl2Y&wu9r6l)Dr2TS5`2TJ^fs9qu*&1}U+hlXGjSV%{EF_S9~|+&bgV#%dO*KX zKZ09~y`-qC8(Gg->6$m%GNXdU5$;yju-$)1Vg_fdv*R*qpyn$*ni8w7S?gcTPWc|qh9m_tzzPYzH?tLbik8L+wvw8& z6T8_cOt>bF+EA(p`y^Jbv9TxeX>v!Gf{tIUV3_#t@IWkDgu59|$iuOFSA{XJ?O$>> zZ>NPS!N>QA`u&VT`X&Pe&WA9D}UP>Om1>QVPOyt1O!z)x{%_03kopzax zeK+!s@kllY8OHgtC%V8Jx~^7Apbhp{*d2Vlf5e zSD-6xy6_1J`V?iz?daW&!dd>CdGhLX-K0!UtL z%TM9ru1N_~69x5%;+wlIKUb9=-At|&r1!29z%`~Nc73Q`oB{4dHTGPla|Equ8^)FL z82LXBaz=(p*XXi!YwUcatc7=tMZPvru9#al6GjniC0dT45YtqgQ*Q1-HJzQzKaU^+ zFi4i+d=@KCvvfPYf+tIC%tFUY?HG`q?INj0x^BlWUFDkS+L1t~$x~UO3Qv&)uOdWR*Z-if1aDySQ9P>&x8ZW0)~G zKt5j$)MAPbRp(0C-wP`ver|8}u%G>+9i4>!+jxU%h4h#rPujVk!tcOFI#-KFVpCk( zNU`2*6tp?<`8+ziv5=Pc;{HTCVM8sVwDMCcCw9kJrb>VnpmLnExafb-88g z{`RUPMIbnI_6V;%N(-gcPeG!MI470YbPG|cb|An&pN>L`EU0C8If5Yc+bHZd8gG^1 zu()9xjD>(|ct_fU%)XpkRv>ZOjzALHzz_IgtXJT~N+Y)-lc#`9BqQFYLe{*Fzrtbu zS}w+k(S0$0btPs{v2hXI*UQQSj%-1UBg`~`Nm5M@?Q>P`rbqe?{>6=U7OE@aV7put zo@4zsOW?Yu=vSi{0a(Ni(Ze_s=hy|=sW2n61Pr&7B{~hg+IR_MYSCA*SdqR~469r``bjOPqdZ0Cnl zlG@rsoRwIV&sI+Rrcb9B-i$KfjPW>6w30x{Pc#vWbz)e*jcwT9$sas`6$-mUW>QC$ ztA>F7%HDRa9is+7Rw!fT!omX+N5S|Yy`S7h#R0?uXu$tV1~<{~sBMxR7k43)t8VO` zFpOPq$Ln~O{v;E1;YDODNH6hee5dhazm_=eg3td|*H=eH{d|8fjo?yJ?$RJFN-Vjg zuyluugp$$?k^<7*DxFHWz%DH%vC>E*E!{{Xsn6o)`#ZnS^L@^>r~{4aoDX0b$Q>%@Vr9k~l}2&d0X$iTC--iSYmnMf?Zf=f!LSKST&` z^2&@tZ4QZdUwF9NG!@jzqw;<^&)BjGFfqHw1iRX)i8KU){uWBG6L_ft)A_^^z0s6& zVOIYFq((l-_s$3kd4#j0TbLv5iiu4!#D?ZUc&kY7OGEWNl_xmnBn5;BBf|ZFy&$VI zP@4jWnuFH0vrZlT5N<)|!hs^4z^k`)U%BgxfCX_AK}?4su&oGfvXeMRxvg+r<{bwU z!dIo)MNtie7Jj~EucBp-QfJ=Y?*{W3oRyndg=}#Bphn%{b}gOqP9Q&!in)p@B#_9T z>IXkQi|)__9pwmXAC9yjn2%~=XG!zyKlMrwLkFh_0Ds+AFj@eKOAaQJEgwSP=o+k> zC;S1hV^^6KF7ZAQDIrpCoS!uk+bU0j!AS($B(2}S^WsGxr0B|5@(>OS0_4igmwtUU zHxt17!Ft&JHsb``dVg0u0b%n-M7VZZ77`jUM+pGHv5=%I1Yd&%znEOJE`Li;smOpH zd)chJydnsiS#vznDBC^A7#X^J_uDa-9{W35@ai`N>^w}?!dpbo(fy8$ke(2Dax0it zZ5Trh6;GGQM=LJ-WWtFGR736-!U?0Uw;^}ea$vC{#zxvF(b$fxKd)AVzFz*$j4>Oz zjY4Q0XNmiL3#~Zt^DI{-BF3;Cz|UsaXBvChgu+bYVQE^;YzIbm=Qv&7y&8=>&!~l* zTwEI;ZxfdkzVhPjj?K0X%cZ#Z14y(pA>&QT!D;huUxdVY|F#SfJ&V!P)(~9@JeWj4 zlSJ;78jaD_W6LjkE>Sumun6+DZpT)vV|i)9SmoD~1SX#4t33KY+pNm1{>tQoa{dR$sp_3QCA=9yZrDD+rC~fzV0X&COw|H zhL4-Sg^&8gHN;ZH&!_C#X8>m=4(D8e|W{8BBH-TFUCe}ofRFX zQ9bG%_LyAdyP|d(UwpZE)#C;$5nLN?86vuhjz#-V(@aU3)$P>EUyO4iW)BN5o1efi zOUlw)Fr8H*&gV5;1q-Hh#iOR;?gkN8^CKyP%nJSL%qp~h?H$K6d%p7wS$SFxe|kAtk}jwGRtsMDamsUqDa1vAJbmT<<%+HW&g((K z6yd8BTb5Z;DktgLKY)%Dz8fOtx4-IK{hpsU86G}*cR^!hL+zHny3<;{_;6MK2kQ=d z$g57tFG;7-|A|gO&qhTT3U{Q!s+Wz#wt=dPJ_21&hc!A?U&i%4UgmbNZOxGCD(U>n zGMzf^EJb5uMPo^)h97CFC_zPuI@iTZUzb-3Zq<@yzY!OX6_U9sNECA8GN8Qs-b;$D z6V(u>cGz&C^i{LQU~}0-38l*lAEmYyCMsCXkm2X$jjHz}*tgjb_!Z~IvL_DhKeO3L zm@^@iqkeeLMSwn42?C83St>z6%%X{{pOmN&T$JR9(byqTRIwRMUZHMS`o8#kt?GZT5k=^ba)HY3d8|lEkT>{I(+yC^WfW#NG`9 zi75_gIc^qZ@D=4|oSYYLzGUQJQm8ims(mF$^&2dA4Q&1wa|USiMkdIJ9}tUt-^ zJI9Bp(Y#V>rkD^oGuRdF=8s>NNwfK_?&suPiMRx=Z~nB$=(B-(mNOC1SEpAqy%%M? zNnC6X`yEBZ2TwZKViO&dLEVHb|Ge`~obj7ZiN_5)_q3cN%hO@jM)bLC(SMtuvlcI; zC+p@=#*6<28+fRdTqyA{YCFj&`>g7cP%;=rAwGABg}5KD-;(L? zUSnhi3H`yQM@aJOIVHtw<^M;BfxDZ>nxGkZu!$n!7Kp=8+v`ra#ooA`Qa(XA8|n-t zH1Mr!HN{Q2q_0nwtgGWI)=W2sX0XJm=9XRcbvb2g5z?4b9CICFRH7;O)j0|h@rZl( z?UVF~bH?A~T}w;meALQRcj&wzjST{Uzlwt!;d-9JRmiMS=Sj4D$&H}O1SQTb| z%9Wg?7?p=yo|LYnJ#y%259?Ms5K$$H|2R-C@P-t#=;H~HCp$9v6Mwyobi>aaaua0( zgLEuhN8EdVpDLehzwtHLdGWi2=kHfBRa8b|Ti(?4QUEJK`(v;aGlDjF zui85_adMpbhwTFyd^K@~K&z_r(hbMGR~zley|Y(N#wJUJnTpd_cZlfnzU|bTqCsM- z^oQ3nCZT@ImllUL^e$?S69K-!4Ecq3e*pQYkvcBzn=xMz13*X@Nk|!?jO5G15@Tq8 zW*MZuQxTfmYCllv{Ij{CsZ&ufq(S;+kLkh4M$rNEutL6N>?#3sgB=$4T^>=~B!oab z(cJj2%)|0QO`k8IfBt@x>60Z8bY%E4>$bcPq??f%nxeO$;=W8na6d>gi+K#2B)-?d z4Z(>d20lmQwO3y03M9yg$I*A;5R+Tx!G8cV)T2|!@?kUv?6NmtTV`ESZnoHW{8*iE$|9vhPut7#B+P)j z+3LN)R{y1_UvGfL!+Q}CFC42xmW)0`bHw&daG^UpsxW(=lpZTGHApX>#PR8mk?bOb z-{4skB@4)M&jM?lV#^})^<}33NUA6w6{wlJi!nVG2O6mMTJAM_J?sT(Z(M!t*~3pB zu1qH?6AtoF_BJK%7l-wPi%@wNMx4Z!rNSZQMV}SpD8FL4{f2im@Zj1Auz}dJ(KbBr z=^<#x10{19l~GI^y4v^XkQ(B>q2Gj_&pU{t2#pIUI>hR*i4`)Yp4=F<_i&XzOn+7{ z_mnPO<6OEdD1uVBJ3w0bi;A2GoG05l#*3BJZg!EhIq{GZ3q%-_WznP;^Y zAWN{0GaaMp%zbvvQx^7Bk9htM2DbQx{F6Z^u9I7uo3NhNueDC`Yz$G)zHBou*<-Pc z=dZYWni*p>NAAVEyn$L{dQgl% zd)JS8qER%__krCXKxstiZX!vDEvJXE1MY)k~`E)8AN1u zg2>O3+D649Zy`8lyNvA-F_EG7CN(KXknvUzC&V*l8#_Lcof79p<(G+~?YG?Br{&t& z9a?nDU*jhv?r_H=QHAEr+$}E;VZoK;JR(Px(L(OMSL=yQO?djCUUt+pT>=ZFrNtTr zKH-MBD5O8iuNW4K7>7h3vB-57M-qv=k`SeI!P231`Hp1r-kR5 zS_sv2RAUsW0ikeBMZom7U$#CnThSS7ZXbvdFMyyYIRrV7o_CEl+gi#eL{iWS_=zj2 zzC@&sT6DAnJlNg+E8LcBw$BiRVX2O#BRg$tv&F~v@AB8#^o);^|MK9!VPG*JpIK;R z5ly*KR%)QfiJkldU^OzLskc(ur!zl-UG~tuh<2-jwWmd3QSw@9(z-qtW^KE*e{5=d zy_VjoiA>>F%W9|Rl-H7wq=TB2R6-Dw`R79qXKHFphxoavpPqV z4%#}j$eu|)ynYd~Qk2Y5b>&*|5V*o-wYHjwgSU#%*fk1bj;~QRhS@U1Kvtv5hDq({|d}IGdN`ZEaX``O}s2>GT{bV#M_B#c_voB zAmpfJqBmV+k%nh}P>%~X`9iwbX?&0P^tnO!_~T!RCtbYVOQw2Ng1M^T;cAUNu3bi~ zCwd1{r-g-9*&x;hMsh{t(YKQDP2-doJio_Pa+2|@3v)Do2t8_UQ#lbN6XIl1O6T=e zsByR#viD4SM@@N)JbD^Vr=&E@l0ajQG0-=VP&9rLOvi1C%t;8GbYIgo`n#aUl~$wB z&U@G3iWwQ4Kwg69T@`jWg>7qJ7eHcJ^zC`u;X<)~f($b>^6UomvaUtm=dYd; zfl~-|y!*V1t2LSZIu&zL+Ax_1ieKvfH9ULNAF8|7;2zO&55i^zVKcS64UoITi~+By zV&52dbP3jkbd}2h@Pa5Ex}YH#=ha_$C-Nn3O{~E5`@gjeMkmVRde{0r*WcWZcj+t? zJrG;HIo8@w=B8m4sEj6LhpsedZP>=uJ4TdVjyW7GW zl1oa08H%iIrX^o=Cw_>X5F_LQYyI*r|^ z3mbv=be25&;DqTN-C+&gyt#ct(mwzJ-Rq~qS6$7eR(b>oM-#HMt zn*vt@apeaju14z9v&}TWFVtfQqt0J3TYLK3;85L1Td6{l+gaTs8W(4DMUuEo>T)6S z+7Wv6`l3>xrIjGZwC^&oNgh9G0!xCgYy8~V|2O4-6c1OAH0nWkFO0GXMpY}vQIjDA?li9@G&r9hWWhb&2{Yd6*l+grARATrn6#$Q<3;;GT+ zsg9L5xxoG0j{hxU6<;q@t(yDCoYP_2Qek~_dI0B!g{3zH(Lh+AgncH$*d*7_CI8I) z&WJjBnGX6E{qY9PDPWmWMfm;6bFVWt_+N#GxbeHc8gzV|f5u4gj>g+&edNprGud3a zHQ>#CXPavrU-29tE8(a70lci)1V7V}JT{))2W{}5{Q=yeHaTfLecnac&Bc^7%4uTW zJb0ZcB`&ObuQ)GB4h<$(Q%2HAf=Yb2QI9GO$A&oeGmd#?7k$=t^SwIz=uR~XVQ6rO zymjpdjVE!1vRD>6j(t#oM@DBeuwRp&TjsE#5jka#fd)Sf0im@&qt}OKpd^@>E^qF2 z6ya$YV3UEJh1vah46{#%bUeX>usBgm&$|f78isog_@PjA-R)@M{G?l@Q zk!dcA(lmIuqbn>&gmWR$HnQ^@z@o79d)$(4klO$=b4tc(O0kkhwCr80*}asLS{2N`^tpnRBE@O!%HJYY44>CQ#3+_D(dtgBepA1La9 z=_X+^4^aV?_EkEi#_pP1%a7-y-I3ootM2aHjM)}t_L|KuSrF{2HRrWS$3yfhQ>m(H z|I=nl*D&wcJ;6ZXp`d@uv*>R82Sbexk^)ZZPBDq`T#u9JtgdcdV15KSK5|NK4pDyb ziOf*8F>gF(6O*|wx^KCte{Jj54Z_n#MrR%pD_K-X$U z{0Msk<#pYfFF^={txU~GTm?ZNJu6IDY4TUUUnO$RVCyj>gq*Ziru6=NTt7csuCh=@ zSZp%+a^8wiozmny+d;~xoi$0ZAPgA#skex`&p|JSOhJu-yow9oiXomHPABRl0FTb( z*S6B)EfdFMrnRBtcfxl?aE{O~CN5}6Jz-#!Eqx|NId7;QS%MtR(bget8}Cb^N5KS2 zPxjagAF*ZvB-F{24vU<0vy1=wcXI{~p42KZ*s<2Bf@25xuOf#ngd3YUJM{b8r$Paq zlUziN$%rEO8!pz22S+#yJHvz}`EM}W`Ctzo7JNgmzY-tpolyr1B+4$cBr^@C0`gdX zwZP3;n&i5di}S`El9{WyW`G(X<|;8vTcT~u8fuqyNNb}$K8X6vHkYELpYDjAokw4J z8>lJ@CsZruI1LeF+#_WDX}v7bf}V4)t9EJ(3>|EtN)&H3ogm%uY5R((U6p)0I$z-GGSeO+#($l1**1Vo$gIF=+a$%vrIP$_!M~*?F*I%W7>x z(bjwd@(&=}yngSn^EtaJ+*c#=t6FlKZ0es{?v55d0l8p74JC(>xyjw&89 z^{J4tc`a1&o$nAbC7>|AUQ_DiqRLtLOC|qUqu;`%Y;W^=ZCKqpRU`Ox*|NXIa>lp2 zJqab*q9Qy}#?GtnVV)6MFj`n%urjsFdll8N2&i3o&Q&*%Hlee9@<0aMk+p~$$><>VE z-hS27-kDCZJYP-sV(KghxrFe)E9GKUK0>_&e|KGL_S7f!nd;8VE%thaByaLDX*4-J zpFcfyTdl6Y0IY=fR&RV6b@+IZ;ti%X)>Mk1cNX)M-Bg^@S(r9&=K6lq+vv}jRzEZT zE5>cLs{xhta{E~rkI%wf?a|AY?xMdJuIVTGzo;EqEjT-@pS9X`?awab)%{AIluG#s zK>P4P#4mvs!|_<0JcmY;j|d59`7}LrvvlKvVo_I|a3Jm82Ako-mW|db@2^fD1J?EZ zn+7W_FtPLnR%1gOZUx|;G`&-u;Lbw5(f!kXqw?6xU*(fs0e+L)=F>*xv`ho5WJYIz zBszyxU=e4_3zYPjb|aUUk!0nkUs=Y(>p!ZUopjFn>n0kfhAIQ+2D8}VmwP1!I~Gzc zJ0AQ3a%|vbD59-&+>>#-lV^W=FL|cQC}9i^RpMJxOzFH@Af-?d(_KB#6Zm+Xe6d$4 zdENp7Mn3sl4TRnpmG{wpyyg!7jKj>uZ26pDW#sF!TUDY0$jlW&Zi(w)6yP zSG{s?yn>&zpIheP@6Jrsd2#4TZgW4;zQj+*gQT@ttKemX2I-H${;Nzo?bhBc@wdV3 z=xq#-1Ct#K#AFcqF(kf6#aC$?DfPd;KyPYK7z1IWMFt!M`(M^dnJsvjO?FdOe>k@N zzQ2)MhjH}hM!BTTHi{8SmV~^@gRZ{qU8Q{)yPBm+wNJu5m+mX&nW2LNw&IRGHDd3| zh&R6)@bb;I z6FN>n1}i_vD%Cc7n#UTL0;e^39hoW(9P@nR|ea`{RyblJwVwVwgarEHVG zod#!yZ`Wl}s})JYN)afl$ITJl!L_09N5!94EWzRW6bHUoXD z<3M6f{8v?hzrBwUf9lctUm0CJZGCBy`r`TQ;Ps82Po8QiqZ?a?|BLB|C zAGUVU#EJJ^q-`Brpt6~meswH>UGw{43- zb2FiEEO@q6y4#*dhGQ3SE*iU%cL9~u1i=b{iGKhgHkXg~%B6G;Yb>uPXSJH+U#oCg z>{5N63XQYEd!G+@nl4VY>b_h+oPtys2gQKBC+VX$7# zgi&VZ2V2X|CT`&BC)t}ymi=AmXX{8^710r5q7GF;%LFb@mNWLxZf8Vml}-`N*b$jU z*)hr)isBKrkVe;tWq3cSElZkpsH68;$rdHuP?oa~6fu!<5*y;*ySo1I51{NlSuVf# zv-j1g&m4wqV~XJft1bOkc*+b+oNQ(D&&PrMPjy{nQMkI#@8Knz=CB_;8#UKB5ijl= ze#jbkrcpgSzT%sLfRuxwIy{vh)HmlJ#F+$Oa{jl;&%~AtXusEYXhWbIkL)x@tUz6z$H_RKQ#41;PpX@}Z`KUdKiF)L6o(O|tUe`gBJa&|v+1 zKEU>XlyHm6aPu+&7`9Ct-c^4b%>72qpp#ceU``{ct%XeEVG(@%Ya z07kOJ^b5s}4h-1ejX8cM1C$%-Y0?`Jd-PU+jd?bIFER`KcqXdDZZ$>yE>>7ak47LI zJJhX1xY;tR<%P^g?mk%qd^EYKp>dUG>eQ&bMY3G~D^z*eSIr`w{DrEGfocDy*{Wn_ zIyZcK$#1#Pmee-xH_8$K%H78$jufufE2DN4do|g?iuWN2LDFP83i1N{Rej-`gY>e(GEh!02GmqmjY4W^ zRM~sDM1dZr5>PcZI3qClf||Te44wNT8$5RNvVbvKh@1 zJy{vh>>B@yMd3j99iIp0NjYpLMEBujeG2XP%|Y#j+rlVQ4HjX|XR#fhfzx^SB7(3x z-*%gi;wexQ;RS)j{!z!6=~4FQFd1%vt<=Kcy1@<}MiBpnz;DF0Z3qV1Y=8g^ABb=y zi}9d^z-2ua8fA6}HU3CF+R#|Mxr573o-RM&Jhufme1S_Q`7kr6+tEifGMn@R%*Fp6 zhE1fo;fUc;w>)# zUc*q2Oh3t|xR?c)U=0R=06T7MOCn0~&$w9dD0*}A4l#iEIq++p0Xt^|xWRO$>yBDd zl_6Q{h$ABh^~$B*=JC)hh8TKHicuDavBu7A@u4;wBUOPYD@k5Hf)X2}q8Mk{Mb;v% zYSA0ehthfzpU-p4?B=dDCm>F*Ikq!WKUIJp z%Tlv+U^LMFiKPzt?%29Uyy~xSIdq^`Gj;oIPcbA5#~Y5HanR3xcy8b8pWvd=gfIzO zH}w~*GNkY9^G_a66y9S~?}{}j=arug!2Z_4KCAH9>IEUw9GShXSBsf-B96H>F(5$y zzdnuBEr{~0hP43{gIB-aw|7lK>YA*&?RxivGdhYYHYs+MkmW-W3y{wTS6v0})6Vv<+~ z*GFeI>+Ziu5+onE+xit_5I>oTmr+d}&F|On8k>NRF$g!TgCxAhwS3Tm)oaxu&|~w~ zPxzz@U8Lm2p8L>}9~-X|sG5rIQRIAZ2xEYqSux%-4&TKPdN#-U-yxqq^z+a_eb{}3 zNEya^Y-xo{jv%vl&tu!;{!CiyV5dx80CNY^^~vop*|4R4%lTlAM{Jp0KEb~H9*2`c zarCLpm=r8c6~|l#Cw@X)CBlLgpOF-BJ|_HX`vJCLB!Gm%-u?N)^R9Zx&D!-pf02Ai z533>LGwX8%vQ1UbG#a-d-&(hZWuJ!8`gVo5-vx-sTKN7;0!nqi1cNtQ%Y=Un3wRce z%JP3wy1_dmn&mAExri&VyQP<7z^swM4;Z?|Z69Y9oBd@|iWeky6hdhrFtaZ^?;iEa5Y~&Dx&(|2F*YAMo@z< zKjDS{E^A+EOSY7QImS8`4?V(AbERE2k4Fu@hmHHfjbJpdqp}*Rt*qFQI?V5+;|CyC zI^zy#>$rw2hmt;JFqIV)SI^pC^vH91Y(&dE#?|Wipb~RuX+tbH*?VfDokLLcf=Ia5 zv@c&_a4;-@=4`TgAj7|R_{^*%q#W)FYWr5{=r>Iw#ij?uMpmnI?K@$%M5#-;=TI2U zUp@POOWzi!hb&dcf!GDxHZwryDKj}s&;m!)%73-fGU)yUXGN&@60eOnq4G#O4; zYy;3qMh5|Alqa3fLF48j_eN*eo@+RP6NTu z5fuJ%y;w{zqHvpI*bawGIo6B)7ZU#u03_#-{}QLrVk|X@RscCeh!Bcl0xd8R)t4mM zbt)av$8i!B6g^^#^Rno$^(02%#O^*U90IL7|G;pSFt+bjPsIAX=x>Qj`DKQ8`@GTN zrwkY1kIIyG!ZU=GqmOzQPRC(%5|}NB>#y-6i`X5M_^SX4b9kuXhwtujR72z*#B1Sv zILf6l!>^Z;8SRO3TkhnqyY=vo=t{IBWb={p;Nan1GZnVr!37q_9fF{7c;XySBIB|4 zdzT-a_q<-$GTXAfD8mCMmCvSb*j5(&Ufw)rRTR4_R%G^*em!^fNoAG(OfvNJlY8r= z^dN@5fckCsFSs5_EuSN7(c0yJ<5`2DOLSk90rztBxdnNQv5u98KNqC{xtUrj28yu$ zc{0W?Rb}I0r=6t9-3SS48~7Runqrjxs0{57d|@hz!TbkgYKP+pe9@d$1BHv7p$0U8 z8U&H+=Z@vQT>+}2d>|=`^}TIv63Er~calcau;@+-m~jxWnJ zas@m@W2kIl&dFy<2vEKF)2&G7psU)?-ifb$DH$ef1tho{+Zr)5b5#4PV)*LbcrTUywuV&x+>XA`e6uoyJsr|DeELh7D zuy+T-q^E^lLJa`&zeCgZb@9;{*|iMT{gVR0Bv3L`kqyH7;{_}2oz79**uTZgv+OS9 z4K@OcnRszgJG!M2X?YLd&*!l{Oc;QChJQMdACN4_zX9>=_;_8uMTYAu%?7w_BED$? zdl2~W6*T(itEs^>_&ww%0naY09db$m))0AqpO9d(+$bF)W8sCa4pNfh(x;=mg?NzB zo^t8~q?l<0S9w30pxFvzOa8=)HWpV1C2TEn4y*haRs0IB`7`i(6y-2jaVlRih#OfOldu!qnoW5Ho zjm7nNoyQx98CovLZi~e}UJFvWjLA!A;79&g^qgE#N5;bJ3{&M^)6Sy2?dr(=D0tAZPMG5UCIA-;w#|xhx~hB0_KzjX@K(Ws$ N9hk#n|K2hEe*k^gWj6o- diff --git a/apps/sim/public/library/best-ai-agent-builder-2026/cover.jpg b/apps/sim/public/library/best-ai-agent-builder-2026/cover.jpg deleted file mode 100644 index dcce413a75b2269c0041a214735063e2b5a779c1..0000000000000000000000000000000000000000 GIT binary patch literal 0 HcmV?d00001 literal 30845 zcmeFXbyS?&vM<`WdvFcz?(P~OxVyW%1czY33GNP!yE{P|cM0wiJcK}Av({ec?K{pL zZ;bu!827L9wS7Imsa3P8epNO9to&IANC05}zTn_s5fH&I6l5eMWR!QPXyDg7QY;)S z@SBwIJt02$!$wU`PR+*4!oSN^PJ= z;*oQePWTJ+#eoBbxKOF=kbUas|T>EWe()YJw)vICB5}9PTPWiJ}uwT641-JjB zDRT*!;3WtBM@{~x^8Zlae<<)j6!`x^0Vtv|jolfIXUxFSF>rIZ{)gEs06@^rbXux} zY%xp9!ZHY;r4zRQch_)-30uO#eTd(Qny>>^%(&Lk0noB(1q~cRx;n;?lnN>c-r$N5 z*u&I8001)oU1KV8)7PLlaU>Vb73scPA?(B0*2TM@&s-zJn!7lsOmu&D3;xFATcHtH zS)7)}K9~L#mWY%V(Id{Imu{L+|4j^mdYp%>qt5+DA;R3oJU#PhWE(VfVba)G-0jqI zMOW;%A}YE!B~FGSybEObE*BWp?Ml{|uN59q_&8i$%EVzkT8(5NC5I$+jYe&oXJ~+^ zw8*W$X|@a}X{GXrk&Wt=D&yb{iN-O~6jCPI^af_i znnVX{h-NpbB9{J2VAtoBub%Bn)=sbfEWaw%G&33ln#zy9m8wg2lq=xSI<&C=W>4x8 z;y#)~>Eg-%7iC~kJrO`RQQZS#P2A_&$=~P{h$?oM?If&EXV&g% zs3_gUM#DQ~BD`k5xV4f?mLmj#Jj@%02=|HN(C2ODJ1bNaENB2ly7-~Rxp-h0iY`KW8B zk^PHkfA1zR)5a#I71U2YD3oG36TJkg_J*&M)Ot`o0tbpGKvc*(QvzN@?@Xo)5VWjgUh{uMCkKEsLG(yz zA)9Hh(e>j@K=0#tFMbXNTw2JcCHwpEa4Ewkl{x;Y+`1#s+cbqfV5upY`r&ZJNX0Cb zkLO}b%}x`l9vw|?elxLuP=~nah`3VdU5lQka_%gRX6YSQ+a9hz^PBnca8!j?AM`Kk zp^1oL%ZF(Rc#DWg?9wr#uNt-4JwRe7Gm?^;l+I7eM7G)1xAwu37{^?>4(X~QGtPqX z3W-a`$2e)Lrh1p(s)o`m65{r3r5jPw&#naBU>F3Y6WNQRxgb&eBnm&BT`uKnPO*0` zNUkPU@FOYqgHYSNy(_fgM^ZLE=co%W3dMG0@8U_xl1~yK4ti2jW7w)uPqw|egzEkmm_ zLIXbTR4=O6+E%lVaL|Nk+*44rrwTt3m%G$hqRL71Jh3aGglvK3bmSO%J zVUwY_U(ozE-hIgzecah?zrrWWJ2X;j8y)|%D@bwE4B3{+byAu+p;q=x#g+G^xo!8I z(E{s*CNxchS+(=lOM!%vpSBZwMZ}7DqK2aGIY$nH%P2227Xa2{jD&jyh;TIdRY`Ib zk&{p({PF}s+Z6l3d*=ki6-i4dS<1QQJru>MBnhNtiKnp^S^q5auZ5Aj_D?-X51{K3 zZaIVaQLD1VmFb2^oK-p%og}lLQ*G2n9C^ftqBd)%Hz6UWmJR4Ul8os)vGP!W{(EFK zwMzaiQ}1wgelj2{kmvI&;ZhY(`A6IF@O|g^fMJoxyvRqeEF4Tos1m`YRb5t!Eu>sS zjpa;R)JdK81|2v*p9r}Mt)YyufFCjFS3W}m>VlFefip_cwR zH1mp`@ZF-GIR)lKC2C0Vx%~OuIdvM_RBN{)7tL`*Y{NZk71`k#sv;yKADTb_F>Wa7 zEP7mRGsx|(XW*!@gs(II=|{I$rqB-P5nI3o1J8a!)CMR=Ng4F|YIp$G*|JmI6O$u8 zIb~2(Ae*Csy{<$EIPBhY_Ef8#y=pDx^6fcz=F%Eb`L5ONaeA7knQvxn(Ws+bq zRxLdi_CIF*i}kGaDk@phM(a zs#wZ3PvPy%(ia4m*4#BtIsK;a|FStIvJ=;Z;YYxykknQQHC!7|00!s<-3sF4`Wj|; z8(+sHqfrhreB4TN^H0q@QWvQFHQFc5wNx+B>|5jh8Z4#k#e-zc^;CS`Y_TI48v5w* zxZL;&?2^PsD600~w|1CoKrd~)=UzG$l4ib#GwzK|yCsWN$vY>ZbjzELb!@I-+07xD z+dJLV?nEMnv*aA+B&jD1AOex&SQSE&LnXyhxP%DLY39;@Ne!4>XlO3lLr(po6X)00 zfVyxu#!W=&OB56<4`^R;x~#8&qOL0yDHavvRaq(g$WPH*V~96Mfzf8vTiE-f=5$OB zuk^|S{-zoaD@VWACuQ~l3m4%OUT`^jcBp~t24)&r3A9yJi~wlM<}NtlD@T>8H5e_@ z1Rny@m}4mmF_J1LHc`du#3ywuBJXF`4@w=;w$ZhS|EBw zc|MO8ZLtH&k6KN1NI0YKArXxh!3Nre!QZ+i?RHg)XfLBmfZPA@|NrY3zILeTzdc7I zHMxHH_tGzT=&}dZ;uh`ApJzxvw&w!~;N2sdSjvZTBiXCJc86I6>#nN%n2w*V!&^Tl zsuDW7O5OV<_HrRaB+3x}XDjlp>xqC66xI$UINkhS-upLyIAu}N3u*)>FsU6QVg#XnzgoDAAr9*;e&vp;@r zNTyoC<$Wf`z%KV6m3GUMBA_pV9b?c5PnO380vXCUeOEQA)f*=9D@1+exJ&JO=?TP} zxrxSVn0h=q#;V1QVA!uOYs7ZA54XZ;&xGq5zX>U1MN2)VW9#d*Hw)DPN|}9)*VnKE z^6hLTV}s=#L@gh8BKs(ikgWub9A5*fOY=6??Km!H1a7#WiB_@ZdGX~y z`xjZW;TrkPpB7kDE%dwN8|s?>ds75}-$-y<&vy?OIdE}ObOLR=ftF5Z-z}CO-epZndQd2J?6#MD}?bOS+AgO|MZoY~slk5ev8v>>xEdP6E{@*Tz0*B+c(E*T016rs1_Bm< z&IN~oh4~JLjDwRK8<$m7>Z36wIhAwJdkXeGaKH{89L9ry2mAqSRH1Aw#7)zE)IB_G zQdIZ@fOJR$sTJ3!{M>rXBiy*$+PV&k zvk_*`l~tkeLw?j|b;~xxOfexEckL|+>&_0_HpL_3ohos4eb6!7s;Mvq4;Ro;0B;Qo2E;Oz4daV<6M zRzU+P4Y678RVVW}y^A?`{G?~W_GL4L4+VR^qXkwcF#Yj`CwjZg90#gYBYI%mLWKD&)JAscN|j|mACq|^ZoJ5<4d{)Hg5yq)3<xj?#GUas0PV(kONj!-B`X2Cn^h55U_D`35>@7*83mDan{O@;)LNNvqz z4(3m<(GhW}spdal26#lAwpM&K3}Xw^Ddy+?*ZeekO$XrARx2SgEjB!W5X>=tYSDN&`j zlV+1(_6WZL9s~Cp9uvemP&tL^4_rim;XDBs5{p9dCGr%TsN`G->r0WogYO7y7lu{s#cea}ytN zh9YKeiT>rkufzRfM`tysHU#S;Xfl2#StAx&G4LiXdc=2-*^x$Z4@YaaQ;xsd%P@K& zMe9DpO)yASik)$Pk3v5588H1`Q^l%OboRIA2a#WCy((-e?~}- z*Y}q}hI(^I#w&yHy2~ww^eN%pF2p$u1dKwTRpVX|@?fI!5Y~8me@zna% zmO?qUmPUwrR^JuJzfbdGIHbw$4;c83IpS7(K4D-+p3aTZrD2g7jqA|M*qF=EVHl$Q zjkd-7vTJxM)S?w7Jd^ckBnu8FJE~zn-JoOe`{kFYlZ7JN%E>Zb+4g8myEM`Yby^>{ ze@O3jhGsvmH7H&Cm?hUN@aqp?o+;aifGV<9-~4Fq)+>5Hb1`}C0bycoO6hLpGsbG3 zTUUQvuqvmOZKC;k6Lg<6UgOpV;{oEigC8u_g=?9gFam@wasnfn*d;F70t!BjT1Pup znx^Ar`iCSP$j;Q$`4KAx=kB?$#sb2l_Tg3ZEd36T`RL(@gBY1>x9KH$u62ibKM%YV z!}?*UvSZmMi-UIHpV7m-K|B83+F{mu0`~~)ql^#h!`0OQCXXdXn`E1ufyRCeM=rQBPTDrL)dO=bebf4F z9L$}0JZFpL+{CSmJyPlaF&wbxv@wB&Vg%}uipJ4BcbMX;8Z4^JJk z#vQtmqIJETj%H2x2u&AE!Qf>qW#>`hf!M=pl7|D5EB9pF7}b2)+a?d3Y%)$l!?^CL zNW8=0eX$!4OD_B+49>4W=Cb9*?rX;)E(nPem6cDHC~!9Ug7tVLX{G)3J=PrynX7I~2h<6&Qgu|aY)`rWExk-{g6bL(##%uo& z(H5aR!WeMN$!e&tb7;54e%+BrPzbDj=&K3i(nFiSOigKTTUF{Q{fO;L-QbJ}!6clh zB2Br%xyV|fTXWdFQpMM$rx;nUFQlbSTy43Rv2@%ZqAVapLPs1XRj?{MUKcDV{)@}M zR`Nj!LX@d!E{gYaOEml(J5i{lrFbhDpGB=&H%HSn`XViIv;~3c z)UdzwgLN6LVPrh`%vY*X!Y4iA?{!6ZSoQP8Ndb86FeP$5e(4>4U-^{pMdQ=(z#vTZ z813p7wL=G5>TCH@x5`x`W~qU-#CXd2;S1|Qnd^6Nq7q(<> zg=~Hi^7*eq6+QpM$bG%1TSF0xpRBc9rQL(0=@zs@H*M81Jn?Ee zb9zLPIhpnQHj8s~h-t0HV3t0WsXLclBe%4vW`uGRb!d@pFCKc(n4C2Sp<&Pe$17Hl z{Y&l1`Ljc$R8t2#5K>_~n%SP;-EJu>>g?0vPUg5R~8lPGLC7-7XE7Qoeimr`N-u@GAPRDCnR4`5a*Iy!6m&L=2lNt@m0tJc> zC+gKUwS38IU9I`6oH58ukc&1%i0Zck>jD))gBul5w4t;2gJ+{YMe%HSs}2j?@Z7~u zGA1$n#cR`1uG1A7B-X64H7jYp9Pn|W1N_UT)d%&IdJ6d8>?X_y0W8-{SHz8hqfmps zj4v@=d%5oe2zspBMjUVFhe#WNhv`p{T|Bcq@}PC28=Xf6RJ*li#Z-QW`X5rbOzu+M z2uaPjnKjn-z!>KMuFTM0@9%%%CU$F_vPlq+jz6 z$}nnYc}V&E_l&@x;Hn>X9!#uk(TI;ntYzkUeqdCff`a-f&bPIWF4vPUAJ>32H0C^8 zhcklFuj7J0ez#5e^`Et2L9dY9{J8V!b3mGn5Y3nMXzCUE1jymAVyqfio$fzXQ__Pe$C zRvW}DW1@s?B=|`$?)-AeYAimGTlSIexkmPiMpC*a1t@b3w1CWaCo!SN-{BpiSa@MK zdyjdIbi8;gdF6Jpi?+IatMEJ6kPwVYj2wV}p%TtNfKi4Hya9blHXUZ<37Gf_1PP;> zp}ctG+j>yGHcl{+^j{jH@7m^3JO0hGfi={k|6ghW%sI^GVLrGO~v)9R^OBp{l(=Nxa_d8^D1?)<_Ci(RsaslDhzlgqnjrAP}+m*<3ZnCz9A_-SWngYm*1+ZoO%QE!M{@>jsmsN{Ew=$Ftz-d$1l~ z_%opcIcD+*Z#36aE8NU=XRwHcY|b}nD%IUB+nli)M3wMK`3=xWIIO_Ii<9En>$~=^ zBkVeyX6|fRGdI$XxTH|kpBEU+Y1e3#7ekAVQ!I31_{0N|~>V$TGRqqIlPeHp=KtT20b|D&9?l8q$~4;zCRKCkrAwSDvRD-@6Ra%`H*1v^SRK?i zJ!@)pK7Fa_7GdVSU7GT(wpGWf(6#5-#htA>{JeGarY~y$(ib4MY%6>Z)xvcoY~qds zn4vD(Y_4#52hzWcjvv(=5{g9+>&nmjsgg9IwBzHf4pWSvH>(umTtUwrZiVgT$$68j zUbpJ#uoE8jws3{AS3ZMj^qyuTPVf$S#u>DIvtBj9^_29S&C-ew)AUbtt^vEqx%yc*P zI1Gv(CYnOg2gXmw*V!JY%=DS6(2akBH=Y@89P4J$9PekcBOkJ zF{CvNL@n)ViLFFV;GSzxA&JgK%Y$hGjf~{tJX;tpAQd2r8+hI0?RV!OQcj!i@=on1 zi_uZP?;ikLsoifqy&_fZufFJJjFjwLB`S@AAhOx=K|y)D>W+{ot-VE$2bCdx>PQ7X zpJ7GjIGtVBFfv|g^n*AU#p3Rk`l4A=9gwEJJb)oG&AnXjm+;SsW)HRfr&?zh{37rdl4%*>B*oZRK>@yS=FLpWQeyg{qi3 zDeCjIi)mD}Z*Z0b_hTWSybidBf9k`QNY@XzIG+U<$kGVMN!uKL)RQfOzNQO%ukV8H zttcFpoPD696fUUz033u;UE3$Mv72>nqhg9b!+-kq8BXv?=I4v?E#^@D5$Q1%?I?+BzZByh z-YX=!X}sl!Ej3f^FSq!qXpC)c)$JulnGQ+JwkPqG5Z(bn`#dgz1o-H zL5ru|I{1zyV;1P7x#dZYbG9y5tW=-Ukjm%o=%qUM3I6S`?&R(BEb;thReB3HpE3Aq zyw;|-`69rP1%3T?=gDaIO8b?u52X7mt$cmSs^z5|l;4o3bi&;DsEjXMST(r1i#=Zz zI9`~C+gR1G~S{j!|709PBI$(Z`0m4WD zf;<9|i!puB=5d&jXw7kgGFiC*3SDO|%%|jmb>r$S{u4l>JBy|BtzMqH5 z-}S+{bVByL9&o=>TUlaYodR+4tN;FelieS{-2jopA_i&YLaR_9`^^APD!1x`#y;51 zP)i@y{R5yr4KZX)bcoJqkuK+V^t`F9jqX>J5S+n=;qD3NH3_~oU&Qr>w!nB-m12R- zmEvHh>4c`QxdcLYA4q>LWV@!1T2REAq#ohI?CUO0qB)DLPT zsLv83Er{+)QmdM?a9R{(%sZlVheYC+z-i&}JnFH@g`#>+Xu*^s5{9#sYF(tNY%^XE zN8>45uFy@b?&6M0Eax<89}bDUt0Vu@@u^Ry5GOgMA$%bAP6%rXUS7q_D9W#g zRd!tvb&bo|&5dD1(z`9=<{=K>=ZpAI!U+xTN5+T9h7r&s*UKRoM;RQ{!E=q^h~2{< zMRo@Ay*t?8x>jJ@^&@7u9fEj8a})idCNB;#UJ>S8(EEig)=0~$&}{-ewusSGgAEbnR- zwE#0W<=?KI`|B;H3E%*HLp137G`&E%0XaT_I#2anGrpu$a2Wd+*Of_3`5OMy7$lo< zy6!3itq2(=u1?NNI!qu&@u7IxnnHO{$@Y z(jG20S}Af2$S)Fx+8YZS%Wo!hh2rWJsodrs4ckIJYsb+OY2Zb5;|DS0b$Z%}>9K@$ z9oD70)gtA(X|Dx44e!PJIuTiz%wSfl}vivmw)p z$yOsp)AQ9wWmEhFee9;F%QCeb1VIaWkEa*8?zUt zv1Nc0W@?$e}8l%CTz_7|P(9P~EJanV!< zIb3lciX!fbzs|5$8SM1bx7T>mgVF-V2QxqIrEXlqT`~MjO{RruV}{g20qsvyimiHb#?@$X$L%&;2*#^eb8E?*VzIf4754Drsg|@{CFI~7fQ=cxsDByA zLPaC{I~R>3qhbz8liiZvrBBatv<<}1GlPf5bTJBS?)T_N?=i8W^)6h@h`0s#1%ef= zOOqDQ;ygO+yQ8N+2khqXwCvlN)K0;+SqL~KXBl{ZFSX-S%qU%(jO*H!ZDw>+D~810 zwhi#?vLb{@&x>3g?aFwyrC%z!z7I zjkIBjD6+^C1O>V&J1k?UBSG!?GV5Z9ue7M&wYUzWn=uXE2#{FAN6xtXzO~i1-Rg$; z`AP5qu3+M&>W&AxRm6qw`SX~Dzk%Z$r@Cb? zF32d5&5+61oNC6nVrqzOJ8--ZgW5(YV$Dz(adFv?DraX^4Xd`u_-L7lnC|m3hxu zK1V;d%AJ;LJVB=e)YBb>HgH1woX6NaZaA$ry7yJ^+@b=+v3%B@>o`9|s~G#n{F0Hh z)L_ej8XKONr+o*?U5+8$`3Gv}nJV3rf5mLvHwwOxxpIG4jO>hH44 z0lGM^&w*_*3)Q6jYsqg>ujW7zi@!XMH|M+hQsae3d35c7#k=AO1#IuVtjd6;ikg9( z%TL#(y~B)lwTq5OJ&9G>{1M9!`)k^w0ad3Sa7ozRBXge87V<6nfz^y0wBwj577`A+ zzQu(CHKiR*FnvC?yj5IWGpGj!lt4s#TRpPMZpn?QduF&(E4!5fXE6u|Fg9?#Ie!WgMDRv1=}$vlbR2-e~(qIcNGe+OkJR2qW*a9i(i)9`Ly z8rU3T1zrA5U5*hZbQf()>!xtB87;iNqG67g5mSJAEn$H@^@LU{6S5%MB$`(yjW)A+ zzTCmjQ86{FJCtYP6pfN0E3WuFdOT&S<=%F(wJlJ_YznHeao+uRFwi^T!IPBwmO;Cm z#jCo}v824A#`8>dgCatqK?N2+yCp6L*8;x3NHgJq&sVQmf2$^xytI)@Do8IReGxU9 z3TLXR^xb43S2@Ad;sSg)0N#{1^G0)cTS09#5fMskJz$`o>RUJ!x#B@gC4%lcPJ)ZV+&+~wC+;B(?F6Nytpm*vXfJ~H6%FH}vL+^p9eW#l z&MPEn;;mJ~kb{sELaw_F9Y%Z4kpdm~4 zuIcjS6pJMn!Fvu~oEbAb8Hg`#Wep8BbY-RUI!j%%L@SyCDbvgy9o;|HbX^a}i_ld0 zJI3sI3Vg_wv&8LtKMb5{oM{C$0@-}iHG31~vi%wA+;$8?^AOL8<{3z8G*1xpM{7FL$o zPY9cJDPc|nNwTCqlEuSniPm|Or zil>ZkBXJ>MXqu3Qd!mx@g};jwNxIdRKbq`vQ|(sfJEUU@BU{3)k>=nPWC)$++ojH?1L~O_vqk*|+hi5yuTfW>Uqn z7NOj2BWzz7?K@rummTX`atsd~Oa{JlM-zHoYMVH)*PkvLgyvYmV`7F~G>K-aGT9Lf zpRR(a5E&{aCa^OZjy|9xn*o+mOb8^?t^Q%L+Nx&L`Nm0|O`U(T1Jqe- z%rskO$KsyakQwlFyL1$2Xk;VYPCcb#%{<6rDQ~+RuYflU+tN(QA?z^5W2S*lyB16C z0~`yC>RtK$Pg?JW5T@(o<72rN`Rl9`HOlK~-I)k5!}Zo^@ee@UeDn{1%p=_vFBClt zj!2yJ=hl^z63|s2*)e@2%?-j)0MPh_Kj$1STgnVy+9h`9<~;3#`jTQv#Uql{`JDR3 zN{i_1090@*;jIGSNcov2HoAa;)5C=Pgy%eLh5NzWMItZ@`}65({Mwo0Bb6~!`t_Q< zMl-UQ=P59;{c~~IREuG!+$G6bwf%$UgQ`wm(z0u-(zk=~npHLj9oUSgb1#8#dlLOm z#GAK!ojv|uUeocHPhu^)I_u>be*lZzete7SL<2f*{9{3s?z(l+!Vg<*}pa zOIlV6y0gAde%~iLAIRNwubOkX%_bjPRMX08v;-r2Ix^*#BLP~#ohRZig2K-NsGVhd z7cSs{^ApeOk5)mUU)*^jZ<+l#7?iQmq+7o6P{!2+#IjNGj@z~-9K^HsX+S9X1*|?C zu@x{V@F}dNI<7khQ(P>r*4TetGx9o@JZ%1M$JeoO%uj-7!Y9)f4ee+% zOckutddMm(11B?DoYR{Ly4SwR4WOPct)>Dg=> zKd_lJZ}jr{NY8=z9*Pt?_$!PV@W#OrXHxJL6Z3H;*%lYlc^*EhqMmQwz*kMQ8tn=K%W3Bp|D-+_D~s5t-742c z*@PH{I_e;bNKcQZ1XbI`^9TAaidK_GeQg}oq}H?(5C`((%4-&aSk`*P5nZZK!~xx| zSwRQNnl`7B3nOOs(sM~nST^QVPD;sz1q$#2e-9M@q}L~J`UfC*Y8;?O?-!s80=iAQ zWbX-kYngN?Pt4YQFS(G$QK*9*;0O(D$ZKh~=OTd6t5a)+UO@jr>)1qJf}u@UB*Jw; zS?U!yq+|yMSFvO`pU$-Q37^WD0h-b}pdT&WQElytJh4l|0pGQUe zF;~Y;qM9dHqovX|r>hzibS>nCp0>w$WcmqTTQiBowYDhWTZOy!Gv)HfI+RTpnt${M z%kyR}XYc6Z;Z2c4d93CO&R$;OhMO*en2aHXBWdEwt$Zc0PGhz8$0DvULE+(|oO&t) zoDNu}q}@!?JJ66ovIRlm-45}NzRn9)I*t-@jptfg^E$@EN z;Au1V=>fjN{sC;^97Xm@&zD?|ln_zi^R;d9WSHV%PqhR9@u>8R5#0ZVC^Qd&)+$3l z)wsg;*UhplQ~KvkCglS^uR zMky>ODrlPLbuFC>ZLF%pJM_Wh_Z--mXq817N!Mn|1duY>h1-6fXG-g})8X5mXqiiT zNLO(0(A2)S0j759QNul56xpV_M}_G2tOxHQ=!~eOxX|8u1$Z$Iky#^+dopuOUGTxK zIJikPp}*`lb#=7C(EAuS>q%)eZDTmTL}kB)Iuu?X?WS0+?*k3q9K|MOlB(%z`$blHswjVGpl5BHVG#_-lq$ciD#M_TC8{ z+oI}oq+|w_oS^cjDDf*8a?OO!Gv%Lt)V zo!)jweEceU(iU#$5-wS~q>#eSa+Em^T_>$>)#>7~ma`vUpaz+Gtw*Q>l=X7VUC*+5 zO=-A*&pO~O(7gsu{skDNzqxd|qyagR2J_bsZQz3@^Nvf18waK#PQ98wikjRLhf<01 zz=HJ%R_w~NVD$ub*!dbk*!Uml&2`ne<{3N!HxjG^*1fJ<#srx=i)0;U->c%t8BAbE zTp-P(PnNkoHOmP7!tv+5r=uHQ=StZ_YVEx=6Yh{N+?wa!>CG{2J|1gXy=8w@s0}lW z8~k7vonNrGn(7*JDG#e^GW&!1m;aJyuz(oN#a}Q+GSS|v8D@;OoDg@Y5jS#1Ym-{3}&SlJf6Zw&0aKNXv!2zK)<62?!> zXJD?jiLDqw37V$v#maO+fV@i-yG>**>fA>JOB0jhYMEf zl&}q;6$pKkWe-_~iO3gREY`;*f(#z_wq zMLZDwory6w9Zb}$nz%}euYjV4N%psLTiqWApC#WT?nMpGcJN1*WUvv-7)H2cGdG-~ zK^kR@P{r<1Ez?eCkJ~IPErZllRbq?2j&)2Slx2M5_Z{N6pdXlM5>X7dWx!w}y1jA>EST;E_bd;{hfX73~63i=T4{LQy6qm3COnh^>URQs-tU zRv(w$@`P!$42&M~o?pPHazB7i<$}-TLc$}$L%~5qLP7z+{{R@!m~8J=V9>F|RMlMS z$;jE&O+J(Eh`eP#-2b{{VvJVSO5iCIZr^b4drwB7dC9enM71BbTi;OH|*~{QbRe zR2s-7=IP|Ra0AgbvQ(|^8Vpox@{E(gp{EL7dopjM3p6oyQZgJ$cV)_mFD+5_E=A1O zPqK$`E+zN_c!a1(lH_M+HMt_vJ|=gJ;YGOwq7QZwzvJ!g{Qbfhh52E-q}ER_x4td! zZi(5)MbQ*JwE^YAWO(7*PK(Yp?)6A{$X9y0uB4uHp~E&6{6|hHnJ%mkGrJ~^W)}~k z;={|M29w+v+8HVzMoxjETr*CwvkjEjjl_Md)J3U<&^I5+s0YFQw{hzi>FM-RRmj)f zq@UJp3}lZ|&K_n&)ZbcQ2c@RA9Gc?t6kC$tMq@@9$sfx&8y>?)n||)wjwV4~=a3U@qA#NTV$p z?e6gL@HffO9Fz#tMeOdq!kAznEIu@n+U(}NBhj4f*Ae}_l(9+youFC!ZYM+jE!Y5r(R4SDW zJy)n0(&;Z%-2MQ5(Q6*^=3MmBc020^(miRxlmTDuPgnWTUn(Uc2v);^?S( zGB|4NA7Jb%#A){jpV1{4!eo3z%gRREHNi2|(7*6+nb4++g3q{xV=DV$sXgPgkkh9L z19{6iloSx7?azRw5qwq5Ym;Sq$RjNrvDdW{UxAuie^||ok2{C|#k(5$L-rxL`WG`S zHCR#|cME~9F8qbGog9P!2Bh(oKLB=djewB+0XJF9yLDkTzDLXCimsuemmX|^}jCaRMRZE0k(>Sf=a;dN++42G%lBw!RK_b&bF>KmN0J;H%QT-t7g_^+S}#e5(qJ$cA`Xzy=+}cQ>dqr@6r?8=zY#8cg84!o zR?RSp`^lyvme2GJx*=!xYg3ykdi_Gdja(t0kDP2b;rgmr zNv2lak)>s|8ytnLBux66t(hQ0^MlTeQ=$9`xf{P*wEZ`7ShX$!*Y^<{B|V*IPipWx z%7^k&7spW9OTy02W2&@JjqJR$%id)j*z6L=6dFb5pOFY0idS(na5JQoenjA6P&Gcw zRohR{u%=CxT?Wg~P9a@v!LPw{{a&Qz6JqQFyAWM#Mq$?;beltiDM=o*LI#Q`)Tn+~ zHoHW?_u)%q4kq1krkdD?+B06wQ)bXbMdWmFn`ZLlL*?!9&v;ITv`hBlPxf2lq)%L5 zXnDwKAv6DndXRTRO$zp!WwL3&b#4iGuT>O(VCr_=vr{!c8z-DteK8yew~Q zp2RvS+ALo`5^_FLsBL!Wg`fQg&t(UcEl8z1 zpgs`jV8@^{6=4j27ELs@kR-~?qOXU{=BypJB!ehQt&-&|WQ@!#5{IygQOo#Q34%{jVC|Tc-`)M9U%O z5z3!YpZ1q{^EaaM5mRZDJUpOo^;E{s8f3quM8xHgC8qq+ki(8v9?*@W{7Qb)(3?cC zE(%#sKhu`X4y{s%vzJ1{Fn-h#(WBq8RH`L4$Pt9jJTp!gl*BthM}8lppl@`%6co?u zFu}f=RYwft978ImL7r$fw05dxItJ&mcN0ON=XoF~3a$N-sTE?lN(?`6Z9YlqmoPI4 zm64tINZRiZ7-gNGDcHthaN;O7MAf*93o7f8sR_PiBiIL@>@xoVyl2*~7(^Uhy9c9? zcV(dJqEAHntp5P`#ivgHpT532D$1|hpYCpk?jfaH37Mh02I&+Sx)D*jyFn8blf-M352@e}mumz4!gyyVm_@*0W}v=UFHA+559Ud+&2b)CdB6cU5FhN*AQr zJ>Oc21ZC~DsR?RpnqYT_z@#f`cbgdt}rn8 zc(@lY(vv)OyZJ=N9|#jxkkr#nw@Gi0Bug9M7%6TLD)E(cGf}>ms?#vZjTe$mgJNU+ zEYq|1C41iUWF^763VmpNN1`~6Dpz|(i8%eM4KH@!gq~_q#%mz*NC0Z=*PUEbk(mZ4 zkI|D25_T@ym)_y4vV}-w***i&==m3VpNO_=dB{@_A7$y$o+rmwC1 z^VB-;Vn7;{aJ)xmRNJFmXbh0$SZ`uR=o(B+t;Zb`hfmGGO+0GketfvBi`tly7@UbL z0cjffQb%T$tF9ajYIz3DQ3?1Bo-@|6*!=hBoa}Q?TE;W4F{>QHtkfLl)eeC-&GL(k z87sWlg*N%7GRf=^7wmXc{{S`zSy9ri2Z|(u2$t` zymr6ZP*>|TL!1l>9A>7kmT$!V2c^=#<>yVu{^KT>_TEfZG8P!x2{6Xg?;CF*y2z%KJvA;5tojPC^ve5asgSAnl zc_iO9vvHy#Fq#Po>wZYwEOev~niqlySoJf2GDp(qisxWeE8~DNhd7Yuv8bD5WFHkP zyy2$KcpQLXnx7AZS0Q2^!vaGt@|I@BdJjV3c&4gxEGO;Q>1LlYe|7Te$Fr@)56hOv zAo(i1X&T}oh=Z`oa!$0hw~CSAPzax?(@7=XTU-auUR0#J$4H|0NbBlqc9~Se^9AG( znvX|!DpWaolMsTCtuXBFzgmZTsg(b8K;WK0jc~V3SH*`dU##hiY?>g?nD}WvoFl9f zn)yCVfQRee`v=2Cg!$Io3*zf!@$5}mRk0N@+lx0l2HvkY)J_4eX%Z{F?il)o3QK%r z6UMI{>lg#(u0Lg@63#~zFXB#`af#723$dSaM)oC6hqp;RODrPNjuoV)T3b!Db%^mX zOtSRFd2aD?JhS^qendL{on~i%?AC*My3xFv_2*K}okq;#~`58Yg)&XgPp9~js@>(1XE2)H9V zZ&PTNJ&p{gUw)tJ$r1aM-DWIK`)h>S?DPC|_NLgh+qXyJ@tcg;4+>vZqv`qNbNHEA zowCH1K@+Zl8F=q-ZkPS(G20sem5D8}>*8L0`>&%1`2S<m zHbH8h_&j|y=*t_yJ;x0quM?p)qTYN2DYU%#hL!|b{OO6lRRIRm z(~ZzFOgLW3z!rncPeRUxBgxg4R+C2B|*(Sx;GS#on z1UYtcs*g@AHNj+(3y9KpF5;FQBm0&J9lI2F!@Y96w7>0_0qJUMZ-N6Id z4p^)BEDvV$*42wJqnvF7o)61fiOQUOSi9_-Ui(qJ^Re4iUqyl*LJ^HX|rZ+~M&%ZSDQ55%6l)@^&k%9`aMjRJ(o zCiJ=yxIf$RK7P(_R2qogUcFZ7W8D@d+GnW1rxk)XYqFE#Cqn+eM5-kXmo!Lad9Zh1 zA;-vLx4j zs~WI0h#@zON!RUI2eq=OBJFYYqok1bzJI}h0RUVA#7G_N%g$oiT&vSBfUfjM? zEPt{p@ zoUO4kY+^P%{rRagsn7loO5Dh1DApjYQ6uRnwOImS{q&U4^-+I7nrXh{Yutduu0A=)%~4?ax{0`l~R5k&yNE^IB^!Cz+QzVfjtrC2_)rmzi# zKypF}PuXWvX_&_!2|t>u6wI0CsTQAl%b$|B%UsBLQOjM;l9ODoVDz}PgkFaXSBcT^ zeYy&m5o?0xw?6E6bm$)8HWlh_A}1{_M!$}A+74JN7FuG9tp;{_$Fl<~Xn=LBhzO?@*-Iv?JP ze8cJV*l#6Z;WWx+drk(>aN&hE@S-5$@f`YqFyvNYpCXZyzKG09+)sQP@^#B1P73|A zCf|9m6GJ=(-Iz3H2YhWJJ~~d zLLMWE^;XFxsV&0X>fdztcP;;y8Wa^EL(^?T;G@IHa4g^*pQj0-!Vexp{#egF{Re2ph<3HWgbLeu4YU_%WrUfME6$L>NVb>rtjF7AzDe3 ze4+Pq-qpyr)(u<)qS|rXYn37-x??MX)V*L|ECUeu8t??u#azTatpRS2d3Nb0 z`J(kAh};8M3s@tL=;cE`m#^g|B+$G1aov|N85rlQe=iiMkm({uk{x^QXLY`u9nddB zd){~0@x|S1nwi@JY6t$pHLL{%OP_NS$HX{QQQ$7csmmVFh@x)|h&I*x!@F9VS>;g~ zdlv9C$b~?lZVD#q98*h3EJAn)rC^{iQHrApg`^!;9)@7*9#hhq0mR1SE^%&4BO}$O zjQl}8!nGtQwGR589yDez2tCq&&;K;WmHTX&=@H(f8W+ppRX^{55@$9u8VPa7%6P}8 z&GATtxIossG#;2iCVzsVUoIp*3<;%;+EV4rfbl5vJ`zqY5P)pUzm-5W_r6}#64?H| zVa|XDzYuJM@!sXvW-d5fvhbyzWhuahRq0Jgbca6A^OGbtoqM|s2S0sl=Wc5k}^1d2SKK) z|5Nwo7d3E+CO?m4Bq~#lKWfO$W^fU66L;Y*?W7_}%ux1?Gur+uE0kiB&#kC6oGoz_ zPgk(k`GCT{acb|o3~r_-zPd-ji&YndCf0upk$Tc{xD5FoSJT=7jdqt?8P|dFZ-wle zDOZAcuuKd{aQ&p?J*^Fw@Uf|l7-tk;P3dL8Ji{o+gEeoyPnoYZGwQZD_xEhISX-Vw zYm=h+gJLRjuc1IEznfTCgq+0?!X2j6A%W}!GX`O_AUDj!mVNq1iT2hg{s3=SI zf@3XGMh1ny-jg8dP)kS>?STIyR3Q*g{POD2ZS`7wz`tkV0slgb?~ML?M){=G@^hS) zwc`?e42OE=2%jQlXwvf=6-9{wJ_=?CJ>4>FMODNG*iA6_oO<(vM<7|;)#nCV67Lf~GmOJZkm^VJH*PnI2r|4H4zV2krOEhrP zb^qZfLrh2XJk{V=qDL0((m|;qSK^#szJAS<@Yh>vMFF}P?o4YC)ULT0tqtYx=DJ|m zaIZ0*wW>^o{+)}GC*ij~<$W{Wh@RiORYv+H`kueHsR=Bv8H_))>EmgAGjvxX@nO63 z>@VeZuq{R_4-o8K!W4f_w=_qF0Jl@T_qDD_UR3P96bcC3^!B-OD^{#m zM%r@AXZ6E}#Y?(h$>t2c12 z?W7rNWKIZ{(cCXys?;~+pr)2FAZOpWn!0Ol>l2!^g^26h$nfwY_X;VNi zp!`-u*+t;|2gMOFv(s!TD9)OgF1>gBVPzBNX9%k4?odn;`}*+WG0e0`CGiN5^c&sR z-5YX7#}lTyhUt+OyztYARWvd~&|gXc!wd@94Yc@a!0URovgYd4C}Emth*g<&uGVK% zW?Z-~J0WWmOczbT%gC+ZQq9M4VHzx@N*c~fs6c5o(dtV-)k$cR#p&sbLGF1Bdxkh{ zk3oqic7%B~CmIP_tk1(dz>jS*7=I{{fcoeIE~dng^Kmi`nT@!Z38~Rjvi3kIV8_$f z@R$l`+Ct+KF>O*yz&zhx>XhT>@3q*fy8{zIwxsKeqi%KI+%cBh`9hGV&alYsCC0-o z1FOTT(e?|Hu4MNj`>Oe_(NsuWB=&Wcp73Yq+%$$yNLH#$J5Ti~cP)s%udub@Wy~Fe z8+VW|=dHbD&&DqO!sBt)*3o)4)Q`hJQ%g_>PqtTMVpXxEDP;xP=nsvDuopwX$0y*8 zGqu#x!zsnk6S2fchis&fiz&^p?&A4cO3>6s2V2E)du-9feY&jF<`CzZ|-`|<*s)_T8*blt&9 ziW^-5{0C(NN-FDJpQP3JtfpDl_ZBU|J0rog0AD3e9soMF*yk95N zE~6Fhx9z@FB2TihBI!Va!Eoq$0yL z=}}t1)F`VG<4n=LqiZ6S=t*!@FO@Q@A!A#6StA3x?ok-AuP^@>y(V0JwpDB1pcQy7 zY2@5!AMTYAZQ0~b>)Rll8hmSes~YdSR=KkwZk5?1k$TwLA<^N$XVk{jBm=AxRZ`Gu zn#&-b&9n%Sk$?k}wAJVVEyO#D8tcRO#X|0ict%?+<_{SZQNUL-Isu(WO@qHeyJgsF zI-2rHK&Bbgs=>bBF+Xp25FS?vQs5!~^dkXTYqV1B+A<5B=tHJd->X%~J?1g+l@?AJ zSeHzW3sA=8+mCy}2T&MC)xl|XB4zdZP=BfHA@O}7in8h!05skgK(vT6h=ELr0Fw0=ba zW!oUuwkyDN?@sVKuy1fGbiBaO)KPcC=hF{`XR+y>z0%gxfN7v z?F{v~en!#TwF#g}MhbWfxULZPah2KB|C6&{<7&fW(SFHk;H`9;`#`Z0F^e6>v&B`Y zZhMq+@-@-zuScJr>DcWCQp$v&CX=8-YFCj8ej%KMxEqQ(?hv7xpE*{;y23Eui#&bl z56V|F;`=`+_Zj@YSUG($(jD=&FUED*k3NX6euEkTAo<=k7Po_4n!mZMPXGWzGXto% zD}#TWjI5LC<};nKXf3^wf>Zk~9q$sv3qDaX4@7Wr+yaC&e}T;v2`M^KhxUc_2^U!uW!$Jt3=%MyW0Lp*YA$TW!2>X-*jj+Y^8ml`ad z1Y}Dkp7$k?lcu%Fll>`%P-twi_%pVC>h>B`~U9YvCVPoDbwc<9OPG?5#&J1_r!S6>ujVDtW| zEl2y<2nE0f!A-@3anR8cBvd_92tFPCQh~UvxAmnpo zeS))O72USy+5(KPk-XBE_Z1?uJV`w!HxVI7{oZ=}A2+p(Pbs*US69+@X?)O+fDaja znD>a(Zj}5Ry+qk9B)v|zu58*$0bVX@38RIma>-)`5?yHwARda=w_P{I|D|gW~-Eb8twO6lxKGJQ?oHhxG>JJN_ARugu3P+m#iJ zD>3I@kMv1w|De3sOumx4E^9QvzcOO6+bLpOjit;ugb5HTv689i&b-WD*$6c|J`BRk zo8GMZW@|JtT}w113+{g>pi*Kx%6)xOkDPQp$)thU>@BgYp6YqzHt@y&GF0po-NWyE z<8OGEOj`mU++KrHz~4DCIq$V~*Z%if{9TrhcZA7TBut5qH ze$MK;e^eVOkIuC@8!P7q)T^X>%`T}(0lS&$8XvIlWRFLE*qvg>&gs^#M|019=%7G(EZ;HYmYolP9G9KG7 zgj=t4xDi*JJ&cqjs&ReLQ!D9g)5ch*w<%Mnax43wQBI}m_5-{W%&;o_p`C6A8tKv6 za?GI@yn@WTZ&q@nITlzws;8urY3BtH;=j-q3;3?Kh>3mPiFG*AJAM=rTs(I+DWg(u z$I@h~r&8>aj1XEQPyQRMCR=0Pskk6CnC8RTjFiLKy>XAPv-;KbBvFsQ^5QH7AHGqSjnnCH zHNe(?<*j(i8_6V19iuS1dh4}YvJGDNa5kars?}&is+jFgmASrZnuZVJ+bb5O6oN0#=78HrJs?1Pz0gW^kl(~ zyH*iQ3epS6(`Wc_v^1g&F4B1Xy!k@Bw71!)DkjBg&a>!`%21z*wMzSB9P%=e(YdP9 zun}9s6t!dRSL+R9!5a&*@WWjmg*XW8=SkX=dzHmt1N#ESG?0*-;6r8rv#VrOhs_U zyZOZ3CQi!&y_gTZ!Sw{99BaX0$nRPXFwd(^=B5jPNezxnoZ%NaADL0cVO*3Ufxfea zG0jA~jUr$CM=xKo5T8U2%$%lUN|J!pb)qw^wwRw8DM`Fxjju*?qq~f7OD+8J2ho@O zK6$10{NP|+?`!ic31noY^v~Pa$7?F#5z@`pS&86#)O(1tfdz*ZGNXT zGJfY_I-dE_)$xleW9`IWx>Ki9R10NY<)5iAl~w@2*Azoul8 z-#=T#*p|cmYD~x^PXCl^peA=9eRN^t@BC^8>Ob#p!!+gc-ybUoe>nY2(g<7Ky1 z9_4;um;B>sS4KU5CRJ{xlr_n?dLVzURv$CNa2qd$_QkMS(8491busoZf8)nueh=@Q zh=#G7@trRYm8IM~Ubz`R%*XyiKA&*}WN@?7r35V779_5VFD3aW4!1ob(Ykobd|h!A zUa7gKI~ICf6VX=6pMO~Sy$fWB?3E1FACzfmSR*nttYkmZ?GC|?@qYIOjNRiZ!5i!O zJl$ih1vgF0oDRtBf-(o$n3P!l4%gE zyCyf0v#y-3*t+v$egB=2Ox8y#b+17FcMSV?ia5LS?wNCfZTpSohPuVNR&qC+a_?-g zhEbMDZZ&PlNZZ*guDWe0hxZfZHqdoqAYXaRs@^$shx6ix5%yvC&z6A&u6|ftfJ87ee5<(`v_lGYW{X8eujZ<*(kx zu``kv!B3{GQ;%vMed`jf81y=37Ta+kq1MwzrP7OyS=@fSA&OP@J#W5r`p&Z~k}#^* zq$&~^0ld*;6ch12>`MR~`q85;e0<^!0VZ}xESqUbcV^5N)u><5`~Txd9VJQQ8WF)n ziT^tD*=##0*k8nZdXlf6(tS!(G=0n+Frgb9qML=i#}c3H=xnJbd2Z%T*@VSX3_kCD z+DBu01cdZX%In{++gn zc4BQ6wdeQyFIUF-asl68?rGf~mEU7MA^|vAC}NqS?Oi4gB69%B{dEB?>x--UDq$Ro z!lrJ{!&|REL28z_C+(q-J-wb?U5}j9*Uov0&5bzfD>_Sn^_OTRJWFhNZJVH9}7>LiXi za4_rM#qQ@ZzQMILD3h#b{IcOF+NB8Mt~O!H?Im&3Z?XFJptf7jmw%xMefc+HE|OE3 zNLvfBJr!+E?2a~NGUDG04HYr@UcDxt-VuT%?d%C)NTE;!I@oxP(9EwiCX+Fck*A=p zi05t6>&Oh&S>9{NUc*!T07D(ITIByx*HYm8ss~JAU}z$hjtL4iV~fQO z$gD~DPHwYs3;OvfT3H7%ph)<{Ks$)hipOE#=|pP2L`!XikOcChDxQLleRpt)s47DG zI9jdmP*Qe{rFgQa!Gy}joY+mB>zHh3SKGe! zt78q9#yB6}=bDxN%4$4^V>_!1&HVyovsZhKGl4#U-|%_XIM_~LN)7ZCv;9nvNH=4N z=iPU2b(#+!57E_@KpqJT0Nyu8 z-=GULrgic626JzbEEatFlZn@P(>1W-10%lq3%9^pe7uUf3}&=4(%P{!1M_$<0UA8_ zvcZz&-;M%=zB7wvyx15+pwKiiug4d0F>TR!Z~0KLWc_W z6#GfWD5u*(xg$V_tkfe3sva5-c@vVhzG_)moTS9MZ2lB4KfSMYncPUf9g2-FU#Q_y zlELsR@Q9cK=n*$$D9V7p)>noj8CI{y=ddWGJD19J32TW;K-AoV0ZCh^IqS40S%K4d zWqnFF#q*e9ONgL5wgN0(=$nCLitqC8p;ND&O3~Fe6=YW@t;%xiwq?s?gv!SVHoG(x zn`WBTu0Cvtryo;>Bx#-kr-rln$ZJXRcJ1fzp76DKdw$i!DRpH(s9SESt{&Jw`{7hM ze)~wBG$Jun9;n=u%3LHkA(|LTFe~wcgKhYA3bNvA!#XtyW;A(9=j>$VW_>qQjN58w zKa(9$*+KKujxL57AdM4pC+IAztA0cRwOrQqi2j@gU)rb3rH|{gAyw*x4ELh`)!2`u z(m!-~-GSjv6Glet;K)&`3u7dJK|KodEDE_KA3jX<@i2@|@cm1m!DD&(-cr05Z1J!Y z>q}X{lH4I?$PMO4irzwPL{72|BR_^SwiUgF@h9S__V4w?UHYi8Q9s?wYUZK1d`fiY z+1AI>fDug^0MI-8a|P>1CYfryt#V@ecwcJ^hax&-WT0}e(qzq?6>EJ#w!eym3(WC; z&&^3Fr$IZJgT;gtnQ^N?`QUIqXtbytWwuk`_PH^oAVX*(Awfv+4-0RuVBcBeXV`71 zx;7Jb`7{H&T>lz-+ah&$EdT>K)El%@#Lz2msNfdu65Jr^O28SM$hCn}2_;H1S%6P;R_ zy{V;%)vidK6bdvQixqEqpXZy#O|@}I=;6a$1Pf1pa@iu~xNV(L;DP|mmu{JH*@ref zm~^uH5L^O~ROR~5$KzmoG6=h{9F((;UrPwK#FDE){~&T|Iec8oC)F7cNg1>Zm7_Yk;p+9f5*h8T`&K>fcQ%0tx$lEZWzT9okh;kePJ zMkYU+Yh((lEUs&9NKK06suI%$LfaP5u(^&XXuB_cOq#=Kp{x9@@rx-QFw>=hyO%L% zK)$qB3osGjUa)Q_p>0Lt3zqs1+h%M+ihKt2#aFv6@({~m??&=>GZXH4%KLtzUTsok zuj27y%X}q_0Ce6pFi$H{fw;+2){3-1WD~8YGut+`_?62czOxOpt3kNCB|kChQ#?O8$TGF2K!-$t+V2!rkdiPORZ#Yze5 zrY3xC6*ZKEg&gJ!<#J&9Bf-XokBb*L$ZxZRUMOc7Lp#$eE*(VOQvxPuZQtc7jir)dQ?GkGja944wTe!tR0M+br~l`_nZq5N%(CA zN;H#8BK^g)rXgyKEdss-eW+SdGM^P;(XBCDm{0OifU}`&2BC*L56S*3540MwX=~Lh zc<;pYK7PvXw7x@)#X<2I_1bE3@toIFVY-@6E9FMgzv5ZRwFOxHm&6>NgaC<8^x{n| zG$sD6@==6z`9r2vx&gwQ`w6QB5F?otHhCvxjxXUD=gzIF=u=IOyyx`$9v65C@3e|F zU3IgqoQrsRG9s|_Azd;3L$*5EFp=SO{1WRrO-HUb)QYGLQok!dEMBa}*Sj4aJmiib zQ)N1GxtG`3nftZ+B^oO8AY4z~e5@fcOBV^y7ip?{?v(OOX`%828XM*>@I)=o(RjJ% zA!dM$(dcT8*Qk_z)xmzx8TUBL3=~SMQHZU!!=&^}4{40CiS3eP60J%TJLJu~_uVPD zdCchR1q67*iwbb7|NL1NOMVinwBWnri6%vEwxd{ID*A5; zg>QPi=mf5__Ke?XJ6V)!?!k$WjfPL;?PvZ@RYK-c<}KH`DDE+_jTs`2A0{w+iS5YlnxQHwkM1I zy#u^U1XCgTc>N&UAcxTscQ;vfy;a5WcYc74&6~8BM}+k1Q>5}e?o3!i-ZiFG1I`;a zhEa(w6Ee8f$ojb2YOCO=wkH^=n)6{xaRVbfHEl%iQ5Mi}c1(AD<#F2HQvswD^Q(Qk z6Rnt&6oQ0{F)ZWp>H8jKOGmXFp%ZA^h_Ir#1i(|(+d9IOt?P-H;<={ao=X0aXWf5L oe*ZyvxOtcKKW^mxd)Em3??oPHsHk!%szv`t3*=?Ke_Jg7KLuT@w*UYD diff --git a/apps/sim/public/library/best-ai-agent-builders-slack-crm-automation-2026/cover.jpg b/apps/sim/public/library/best-ai-agent-builders-slack-crm-automation-2026/cover.jpg deleted file mode 100644 index b4bf5bdb1c9cbfd8800ee005112d770b65fc1c50..0000000000000000000000000000000000000000 GIT binary patch literal 0 HcmV?d00001 literal 30678 zcmeFXWmp_twlLbbI}O2If(Do1t|7R)ySuwXu*M1Q-bf(0OB#1gAhNJxk%sL&qz`)_a z{22m}0$>2Ja4;}1fPY9}WJj0np)~eX!xMp&wfN{?+Y2E+B}S zlNA2Py`)W~r^*qgYHMcz@Nc&4YF?|FTHS2!h+Feu6g!LjSohHUf1 znNJY%pG`3@bI`u;GEFr z97CP07z$eC+Xw(m8J}Ej z$i&~3&?nAE&cAwOx&?&*0NhX4?`;z&_-FxuB|DV&|JLpwMNIeMi2~qcS=W?fhQe`4 znMjG+`v3|DgF=|=Kn~928BGemV6_7Pz{h>KPW2-noqi`;EvvQUE_(;WcrI;@?s z$y!`wDkLZ3i_?i%ICj)Naq@zcoz<2y7z>+Gy8Mt3`UDSp{CqF?vk&dQF3Ln*jj1y_2Lb##oos~9}?s7qug7{Q<)%9@k?t|S2 z?cP?@aQeO~d7gFIEkvN`Q9buZ?{XNzR)Nk1QU}DQ7))&82SG7hHTLu^l!iD{fv>dQXXlX~#^tP5 z#Vr03+a5vSYl_eE)kc~%9KI*MDeDnANi&4M{}iXDYCUgxk*U6p#HUzPP;N0gr)8m@ zD;mai5hR-k{HVNv)c9Zcfy=x18~>KjCXE$36L4w_Jm`VJ%ac5y=dB20eisv}Lx=R> z)t9`=YF{iUd!E##pt3x<(9s7AZvJsOd>G_`{iNI{c8%2Cxz1XIF%DiSV7I{_ShzPqm-P{*rfsIC-M2v zDKM6fz-&w4qC$ao9ndetUf_PRUViDxI2BqcVX1T+Ug~>(vzW2iDrH7!$#ItM;PfPM zDR8XHUOtubvq)$>qB*NeNkQbYWb(wfWDA=KdIXXZVii612xkpW@;oGZ8pQ;*Y|#b< zRUrfS#Q|*=Bl2Aj0&~~pU6@Jb+#*QJGVUK`fC1#LNP^UV)6SK|^U7`i97Wf|trK+z z|EQvRJ3(c+vr*75L1H+BS>H@YFcPPX2x2x>l`hMaQ<@Y>fF!1&MTt@p*Ht$391J4yzbU~Jk54DAgP-wzf_GOG zCRSd7!M^^-Y_vv&7!D!>*xgg7lKU_vONxWyy*L6U+@Fh#tRaZG^4tJ3$pf?Qbf*yP z?(+?pD*_PeiHL(xV90+#{=bxb{No;!9peTZJN}o-|M3akCSS@p*U9~`@1+WnDl_{1 zFxnr0Ib`cVwuYG{K>XOetiY)7rPBBJ&CjmZm(pKXK)})m_hJCdH%)?a41`OM;9lNo zG~KIf(hWh`lUyW4f#W_F#^{6@0@Za9GWE$*nsdfr5!Du zyzhU6O5|RdVB~?X zQOEwyu3IkB+f3U^BW`iEASk59kv%Qv0eooMo2E|AApdKI&=v261021A{{!ed_)Ptt z@{PYtJ$qfL`j^^fk9>HkV!j}b>24T;?7V087-uxjr27=2Ht;Rneq_(>o;YvXN!;go z(^rBL9ZfZ<4Iga$SZco-(kXxR#=!8)5MsX`5*2b3hxn#eR?Pym?-5+I$dIP<2`od9 zG8B@0B+!*5$1c~}HdH4EUEBLp6FZAD@ zsDx13Cb)a;xk#ULagY=)%#+LStF+Owcq{SsCPrWu0P`bJInxcuBUFs1q@K@`vop?R zXKz(5DD#@RPzAq!^R-Te#LcX$J#TQCnf=)2k2( zqR3a0+}&Nn!00HPqMuJ=u(CL==aJ0k5mV`S95K7;QFiWbIjA-@`5st`Do-LviEJ?c z>VSC)v`y$fJ!4|=e(%?TvOxvkJIZ@q+DbuUQf=)99(}1B*{<1iR;X|+WmAm>z|}WUSaQ-xRh) z8F=NbUAfjcK5b&#?>?A_RohSC>!nBThN*D{1`*ZFtOu8xZPc;CMhy3y434liucnYZ zkeIoD?_Kv%s#!<7+@Bm^$kGX!p9y@8OQp|h-GO=ex!9sa_S5a_q?Bq$)A#E$ZrwdK z{**N}Q>hJ!4ABKM4L1&F;x-}3o^R*;ICHiwPn6k6cY@3=x`QwsnlCLQTJXD^p6byU zurw(Jj_<-(%BMg2VPb0-q98WQa^JQpIWshMQ2xEd|Km@XgF^t|QKS!m^sx{C_x%Z~ zRsD}Se=x@&mBU!98r;l8&j0}M)Xrj{FQA(LVn-MjlCJXkBx$v)A10~~BS9igB4U4y zLWSpZq_llkmgf(B#ktYeKG`p&v9??Oer*ZKJZg(|ICM9!2E9{gn~Ri6O0(gmFRomY zlk0J8$K(T1aBo^PXnqvWZ$aY7e1i???gH=G4!{N5|uL;UUC(L;yBjU6&SoGQwhv z!!#(gd4B6kQ`*hU%zahgB1jY1^c+K&x%t`?Nc&EmfF4|le)$j~XNOlR+~jG!7>t%u zFqf!-f7sg@wMF@Kra{?Sy!K^I-u%Y0R3}8%;=P-&?M<)@F{9H~4V#Zlw4Z}Rk5?|i zC9o_L@AXtd<=O6oVrMLV>t|KNQW0R=I9ou-^+8W;*_UibTbFcpu_djd{JoP(MiSyr zt;u@jwxw)cYeA$Q31<0Cu~Yv;r$%HK(#znmS!r1^%{5dgMv~s^W7q?NmmDTzNDoYw zmryKc`0ij8>J;i1Xcjd4RWg;@;e@UZdg4k@(AQd;CkQqx8(@`nS2K<@ZyAJd8;dU# zOfEm$qV>gF4WATW*MG~lYA^GxD2u6M@KUoDKfTN4#CpHJh%%cY-5{T}d?xG~LwX~o zDvUC%9g7&JaBF#TjE{MOMcdIAK1I1F`J!A!kk_;_feY_Vm5)kOihl!O1+7)L{Atdw zhrcjlj9_;IpsYh_-h&k*AY8+8PQ-{|?;*VXbxgz#sRDyruv%5alU_K>k*BQFzis?q zMGOvA`@07mI3w8P1rH(0hK3~Do{3y8UR;mleA^-#FCJSKDtz2xp;c*8@{s~% zD{Tux#JeBovN#Otz1stkU5w9tY1(Z-Kg~c^h%n3R_<}5sI=k{!Ml_?U0#FOuTsu!b zD}*ofKz@tcb;e&X$;w27_fEpYQr5{vq4URn)vBY!6?3@p3{Z1PLr{1>p(j-K`>sVM zKeq1>X{8@m-qV}#MXxd~k47e&1rWD!A2!!{`OPGq*Wx?YJxO+&y94o=$h*cuE4uLM zTxD1OR-Vq1>%7rlw0v~BxB3%NVu6c>coKw5(u6z=3yAwnDy)X|r~eHPFsM0kdb{@k zA>mKux+ z^{9|Sjf#_Ury>RDK4QhSE#e!j1yC>Utq2JZFc@yZMx?EUQ&2KPG#ddm-KT}HtgrF& zC(o&|nHyn@k60UFF$C~)PG9c4H3#CrzVeob!*UkLigX!unqk<>ox1X^b~VmhJ}6FW zEakre{{QBj}3r4SPGk)eJ|7-Ya7z`17Bg*}z?`$|@={Wj|SKY&o)N%PkO#HPZ| z{sryb!XyouvQ~((~CO3!#fNR3l+O;W8^PFSC;*uqE>o z4chhP(P_fu;lWv+SNtYFxSGnNlevSh#U@qRWLf7j!e`1@>2#(`P+i>u`~;s6@wGIt ztgifveQ}th(>3yvJacVAH5EACzeCVOI_Xnj(B;0*)hn3RYg)@=6510X&!|6up3~1P z8dogX!EU|+?)7b;EP1Or!PB7MltZ`NDB3dZ$TrBZ0d0#l^$NC98jD!(nBqhxmmO|2 ze+IRKK5GIsO-`MhfiuyTDA-3Ct~$CRy4UBOBnopTAAIu!u$8tSAeokN1$1Fp=PC?5 zb#_cXslJb)^RN(_`q7KFYJ#t_=;BY7V@__vM46hHGQ92njge=zJA^QULvqjThG2;H z)zjAdU#~h`@$Lh@ZC_AaJhM##g%sf_HVWR^>n&(p(cbL+19x|VSsMFPF3+`IofFM_ z*m1jG!I;#esf_z(&4v@D3WJ3%o$w8wt|@GPgP$%c!)74$>m$GS^XT*aWb?2JZa>lk zLt4&KD^N&Y=nr6L>otAGmlpDPYqpezNQxoG=a1LZLeIN%vL~L*D!Qb&4&7eUJjSCZ zzLK|QdqXO#H_^de!9AxBlX5J!g#SL$cRr00ffJhCVl|$L8e`+1DH0Fe<uMp+828Y{E7l4S{itR9`jrK?GZ|>f*f>W}xe`2R>94Q$ zSi&hh5GC7`ILZ7pk*$jp?&uLYjePm+P}g5&p+*Gl_5&~Ive&5>oT<*@sx;ar!*luv zz_#dZ2b{&ms4`e1a~YF=z?Cw?62MJYioz4%v`RDGKMwfZaZZu4vS0H9!MW#YaK~!| zH@i^Bt8wDt_eXdBMy9XnDr#F_liY`i&hK^an6Vr&zgSAun(wBh$&xlbKZZYKJQ8l9 zOvaGOu@s(o8l6-A$`8Jl`Ss0raJ$ZS{7Rb1OTP7l%J+3L8V@e1G(h28#h=Em0&Tt-Gd`91!no{vq{vBli~)D9@kRy0%rygh@+lUP0dtYfco}RWo2bt!Kxww zf(A&SC>v;YKOr|0XtN(>p&1YQ1DGB8PPiI-20qpz8ocOg{Qg*+_9Wk859q?h#Nr+x z6GlF~@XgR%!Ze?B>W~@bIh&&Nq-+kaTlA;p=Fi_uNtI>xpS?mlJ^9tFjLsO(A{{SS zEd8619E~*5oz^1pg6k13=(6kE#Zt>bD;cs!InR$O)GHea+5OQ^lbKjj|%{$R0u|DbYGij8^c#Ju4eJ?sG z0EROUp&5*>v{CcE(^{U~4n%=nMr1hg`_YeMcB-kPxz+DeZCyuGbqXo#-e8@wt?Du7 z#934~83-v+hb0ovwM}=m`z%&dQ&UTTHu5Y$gCy^=hokOg(|8Ysv6L}Te2!T49 z--ItxCLV^lXXI9?7I|Y?3_{4-1@s$qml!qtsDw`+Wj$V(tHbXNA$6x|dE3)xJJig&|MRRF)Yi(> z;7Y`{ub!S6-5TJ}^=Hr#?3x8Gy13$uUsyQ!ZwGeb38iA`(xxgsh!|9?v(%7}DBzT= zpy5a!B=V{`7$a8mxUAbV6Lt){JYTT5Jzb-6sL8HuuC57Byc3g$?WQG9_RdbWAgFUU zTI%Sn=et0gm<&HA;|jkk?@!ujL*h|KeN$qoE5GEa*`$-eii3lLXWOP%-QLN5yf?x6 zM$RP3|25B(yzYkAd;=ZJl|15StK%Bg@?i}-rYeguQz<<9OevVl54O(bl_~k74#WmZ zQ4-a6f%4Nj$<2)=H7yybx!?#^&weRxM>7woMwzlR?Ep#t5XYGHsT10k! zq1Oq~N%E(FGNMBQvcvd>VAj9M+Uv7L2Ti< zx4U_W1f1jcdP}W!TTe6sS{`e^UnBrt#U46m8jH&SW+?NnY`$!rq%8OK@Tc@^ zX9c_CR!RNUTB)|2qJ+!2ZN+*{hAt*$xxtCxyfEabC`PJZuDw1+YVXs0c1VyWpILqc zTY_>n>m&&zm`@XG^1<=S_85H}K!V>AJwqIq)bBY*8AMF7KfY0#tgH7goY8e_lcsJ# zF}615F&#o-b?lB*+_)UxL18_W7~Fh~s$svIDxkAU1DU^u1P{lE6l1uO`O+a3KNaPd%fBZ z3~6_x%1pbxd`aE46I{b{Q(N?FZ+*{JBh6ESY5B_Hs^3X zn@GzM(9Z=08${0ttX0(l;|iSE^#&ufwua+7}jp>Q|@&L=?fLp%;Rz&zI-4OED@rh|#sV|Xyx z^~3uk+a=Y{EHeyHOQC^O0wV-a&QvLFPgv|#vo*>wv3%kUULe6E%dPm1V3o3Le%~>+ zpe?v@*WCQ3Ra&c>$!8wQ6a_AVt^IlIv0~;?AFRIpAWO3A{h;;ib%G`8Daa>e3m(ZO zdtske@uvIsmKe`0??6$IItZs1M2uM-bi!rTlV4TwThVXq;tL03AART^qA>DxHq%NPNZzXA#1(PUGG=I!?cmH9Q zXaZcme7(bP`VS#*tk=LSm=VnOPhn`jmTaF>9k-}qMBZJ^lOFIGb}J@k{dPnkQ>*gK zFSLF7b@mGkP&5%rTn{CSickdRM{^G1KavZ`y8NoH{))1|9(87rgEHbbD(slzDb*Mn z!ub{u#U{E;@(_7|71UbsHa`M2hE`jiW-sMhGKtD65B#reladKxeAcSLWzCZDZ8Qp9G2ZD-wp|ewuNcouX+yfN$x? zH_X;LXsAQS2Q@>6(Mp*+zjSPH!bP0#=?3h|rA&Q?t?VTA|0U4U9 z@{>M|H(H6nmO|EyLbDxeYH9}Ir``cZ^A##i#P{Dvgs#%yxrRRl$(rXhdmrU^ECsZP~aDx5>j$BB{^=^#rsDh5;08mj}(?Cr4#h!M(=`z zTI(Zqp-MD2n|NSWc1tK-;%koKaNU)tt*>jZM{?S#wckQ_InjUh4*;rb`OLf3R0?i+ z#OURWF~P*hy~E`iqSSQ&&^LBSQ^O(?NPTc~WJz##iAA8M8l@KNxqN3J@wDixyHNAP z6|M>)oy@5Qv?$gN$E!n|cqKeLMppL$xekBQpKE>N^#FE%mU@T5Lcbe&*rd6hp)NaV zVV#5h=YS#-o+BkKg{2&DJV3;y5DJgsT;=!dX-Fj=X>Um-xJ`guBsqDJdzX z0%ZBv%PE<7MXrolF0pFqGnbh+s$a=P)Q>dbA>7xC_L?of zF>($(f}gqQIbC>&t9QP!VR{kgTV}g~q=U-T^$es3Pg!)7iCxQY>F8krf`_!;Slgr4 z_JHDM3bR`2*Pd%U-u_giuIA_B&17Vest4Od|07w_k)O`M#Iv1 zs5AD|I^mN5N-(C$0dzEYS}`<0<6bL&=9^IzqlnV6T?f@s61P#!@@D_C;K`>iMk6Oa=_Bn;ug;{7nZV-mTRCZ)bqN+vs2xO;{foNP7DS*pJsfDKzK zjXXseKV@vAy%6zehQE@#+I2t5%CfD)*qz%QpFa0J(HlHMX62yz?cdbKgR`2G-@;24 z;_iF=6(}7G$x(ix$badR5}}J|_Fx(}s3#YpKu-Dt$ki3sRR06O-snb=&)_ba8|5~; z_9vnt9nCFQBdFeg8(nhW;Gv9Ljzm)t;)4*|*#EoWp6AS+qHz|YV7?xrL!>K+O&cIC zK(tC%u<2+?QZ#0o9;=HkxXGlpTFaVz*@Sz75a|R)0%IDn-YfA7}lIXdb6ou-d zx(es}-XB};@IV|n$1|>trS@BR9RPJT_WGJOyO}gMS9>oZgK+J;WuKWpfPAZ1psGF% z1(ZOJN$rc*io!Jr{+ zs=|YaQj)HzzeQZUX8~cOzueZi&Nxl(A><5qZ479aDE+x~+GHua`IcEk*{v1gZjbI5 z>0?*S>8RFe{|xl1(nhGOmXc($+mR^|g>$yk+0t!7;Z<}CM~qI?^8x*v^qePfN7*~Y`}*7bUF!)n}blsUCMfR~{a z1VhXka!P5i!FFm#i%=BazBOgzDstnQ@Y(_&>0BQze=SA8kkrXYpOeX0_x@PAkL>I8 z$2@$^5}j$YFE7Fq>S!X>lqs&>xL}Pu534sUc(T;=cx(6*^GtzkiW$Wm|6I$fNTYu@ z)3jUf-2!=F*L2csg{qXKsi$RyOE>57uxl^* z?OU;98CL{~uC)($;=;22Yw@8@?UDjLJUlDj=CzGh-&_KvOg`~HFh9r*d7hunZe7*T zt~Qhg2a0Vv!TSgJty4KlnMzvR3~+NQ^VHufB?ovWud*9`cMI*@K)o*_@oe-)ucow{ zDlt>Aw+-w`q zWtRZ&**UlW@iwsT=V*NZefUa32XWUMlioqET8WJQGz+S; zYdTG-^XGaIeo*hCyKhGTOHwm3WmZ*j7!H3hMbskajS{F)8^miL-7}IW@SQO$oY>8( z6->Rcs!)&hbM?d?Aa5xwg(`3Gr`q!Lt4 z%QI*!&P?rv9>&DPo&S?mw{75pUdb!l!iGvJS=Y@QCnm8 zBpNe0It?5V$W?s%k&tKz-y{#;EX|hWKhk;~pW8Y)!2@;kqQvI#n_@*@c-ZW+cu-Tt z&V}6z@(CnE&6eL&lwa!Lk$u(0!Zns93RZg(Hd;`MqIb5~^RXz*jfuTKHaUK4(T%8w zU=w2La;U8O0YB2gO&^Ar=@Vhkc=|!vLEb}-Q&ijl|L`n#aP%+k?Y%am&(=9e zIt-wn>ZQ%E9LiBx9ASYFd!41$j!&?IudUNSwd(Q3Fy}jjOEn`LbHcTb(gr{nDQ=Nb zo(B7@XJLF?gCiI&m2SLNH5&%9U*p%+r%%4ByPB)w4*1gjLzp>)j6n(fa)7M-h<6UfcAZiPCo?v=^X6ib{55=CSRCN^~H&vXH1s z%P|gZ0oa6Dz4}#SiYkz}%U*NJ{=IlHvJPTAUltL@nsCmtQSErC1BW7WdBzQdWqJl+ zo<50mz13&3^d5iWg+6w0*~7KELN#2ds{O5=xcn7hXKWZ-n0E}7$!Am1ZwvcVF>%!~8GtfR8#8*BKY;SwpKUKbYOxde>rEWBpSAND zRZTGK6Gy7A>pE z&uW|3d@(u)2ysPzhk(^FwAY?I*Q-fH*Inu`d5Y@(QBxPTSQ)jU?W9)U?Sq7A6w3Fq zR#4t!oXO3Q2963ApFpjB-pS@zb6G0tp8K9(P2{rEfC<;+v>yQ1KaYDk`ZYYo5M88yg5Lql5Wpi`L!Hj3N%V9=`HS%D3iNrC>OlW9>_Migsch-E3~sE{gzzzM{H zS++}WANq+{;nDbSd9!939>QIk76{ICWpER=wv8I$AL2?W7XyHSZ#NqJx4MMMHbGJi z*n&0&MNe<`7B1-5MzoKB5<2(>!_+jYIILqPt=3N2@Z(9dxMfJ#`8Z?hD)>3rAAHWhK#J+yXBEccSpH%JwKYKJy5^TYh zX3WmIG{IFz0eose}WG3j9DuVmJ0 zCbjMcqXPA%8!Lg`rQ_x*mY1e<>7q`tL^6?&KIP%A> zF&QBb8nSbeM`nJ|9x~~lUmg)_u=F&u&?UDnTV~ub z7O!8}l$g z4AxTX)A4bmUD0I(BnRZ*L>Y#ad9O>NnbK5atpRNO4P#})AanPAwZuqZEU9vK2vg>r z=TZhnC@wx-P>K~@h(#|tviesK1RWc!i;la=GUJlZ*sSnJq;Zyx?=|1}P@6sphguUW z@+3M`3rwU2xT}uUx#pl?F?pjXR~x8*#8JKZyb#*ve~@HIaFnEBvIZ->tuBw4)Kd}VL1`FL+LALot6k@fN_qCwU@uS`dB>nwHl zIjHqX^Vy&>#D=c>%=JMpIxhym&t0$qE}$@f{`3yO?Z2j0FZ+FtfkuwGG13CL>Ff#j4q0ap~-L7?K??-4gw2?!7RT$E3|+_yK88@=1j6;9Zv z7IAqeXRaw>xr=aU2fYK4Pe^ZXNoMhq4M++f^FfH!-w-ta00@TLrL_*?sM_weFtvf9 z25Z_K_>SHVQL7OAdVS1!1W}tyGH+Y6JvA5Rc)1D4m$Bg2Pi9Y}ff|Qt{#ue7u zyuFVIO+th;k%gN=Us7VNC)PFuUAx>IdDs03gjUOIDsQE#8>=C7_h)NcI>^vQ(U zO@_FG*%SJyRELQXej$!4vVLt8r^=$!jJm`m6n#3T7!=}y6KlSUtWVgySDYqk$$4T zGNCL%hnQW)kD;apf_D7G%;>SQ2y}$vxv}LsTpP9L`I_C@*X2ojjD&aY;@6J5Pjxp| zYG2*UjdW)>pCQbVVb5qfMN^njHg7n;L4S}m_}qH^xjV!ExKl>jk~9ItD^DT(--i7x z{0-j)jh5mKOY@BrX4M5b3Sa#4=hniD5sow4eqx@Gt5iM_sZfvB!s_-teCh_Z;a_e z#hr1sDm{FJRoX*oyvP)L5d~#=5Re!~m?HE>)(D-aD2Z06mPcWo z8OR?Bi~TBH|4%wo4WcAg=T0EL9 zFf^@f3j?oAX{L62o4QSDo!tT~usf`;W7)*WA3M1FHowMR%~AjbMsF|egjG6)72(6$ zu5{A0T5}bq#6iDDyxIsU$ZjEWPE5n%;uopYi_5TaqYc!3LwoZ@X-=6(q{7@8&w7AA zK>byJdCYnyX`c`I6>X}}Z&6>>eG9FX_$JBg37 z=Zn$~*6Q+gLv>?Y#ur>i>TAZu!@mrf+1di()Ub1ZLmWPr=)ICTcC(aJp6=?{Ey%=p zsD>gED?cBKENriJhE)&P%l5K#`P==?<~O;9GB5kNT0T!FNL;<5Ux#%3Os20c`tEWx z43V$JMD?7!mkmrOWdiH^>BO_?4#akM*p21;#(6X}oDC0#y!||nX5sTr7Ng<3z)Efi z{F{HS1X)5M)wh%$>f{z{IF~RCzX@*mU9{3nWDGy6{EDve>j;z3S}$cY1h1|)8ybXr zv4ypNIW6-CK7dHu(nx_s4OX_TwW!6fml?vSO+hHhz`mUK_1}IZ*)qU?ev#_Mq4+!v ze3{tNW~{!p89Y~7+6gSatqJ|kN#@)K%wE{)DSD15!qw?5VKe=2sp+4x-ffdO_0+US z=tM|sXlOwG#X0`~WJ&&V*UJd{M|<8@W=KeNkia;uuik+&Sti))2E5vP>-rM3vBnOX>dlb$(DdAmCwGQz^iNO8 zxvp_$S`=Pd`Zk{cPO-I8Zc4?MX_uuNuao7eg9b~ty?)v)x4(o4`SzqcD41|29d#1- zO-rzvy%-AiT;HRrDsGZa_)}6Qx&3hZn4B?@S+a1w{lvc!_RPU$cc%Z$wh6>> zCM=jq)I$@wx1&f}hQ_!4p{Be475RjwIGz68K1S3UY9J!1p)m?>YIi=H5@7 zW~Z%Yg=WKiCI-MR#lI@vUxw`uHp#Aj7rwpR;d}00`Y=>%|DNTcHQ}AsGK_i>7|+$q za^W}aohQA1>Iz~-1wocZ82YxUmcBR}238iW59cd$@>fOipaB}FaTF$usOeax8cYUW z36S|-MPjTKSeq(CSaL6P8#G#wymCFzc-*0j)43Qu=mekP?qJt$+AKY1<)Gcvy2{ej zkDod&RKPv2_`;zwpelRo6h+`!H0vi>LMdZ!9a!XZPHRh$w% z2Fjb=mjG7-hGS&YFjw?$$4G(7Rr@^!dexf<2zl?sB7dg)R?Sa{g=IU>30CX^ir__v z2+6BNIjeME^Bk&A;WTDa^d}A?J3OIWP@Ugx^_vIblCqwD>=Bxp*^j8LbAdX4gHZ?k zxq0%(=7PfsH3HM!?Gs>3^S0x6y`d}FC?Yu!zuEG!T233)@{5n51)3qo+~N9?C6wE1 zn8y7FFm}>>%dnzB&IJSlRXsNcm}Xz$lwK?*nEmkAH#Lw`=4vvnFW(PT)@j;TdE16XhLee6Gtco{FL_z6<6kQ*GH%K`z^Ai16#5eD+(_oD)_(4~E>9{+$yCX#EO^Jymr3Gie-Q9@a?_J?w z*POZ^nFYaLR$(x*)MGim;^}D=r>Xh`awoZLzuDfHjK4WASfqS2CX^LLUf#y^ff84^ zc_t^PC0)Dwx_dg6q127zJlVfBHX%vWc*DBkoJ()bfz83-HGDFT`4^$3JNE?=MzESg zg&I>+;toQvd+TzpFKG6Yk7<+|UD7KXCVgdn?315X+q#YpVPbgPHGBRzGm@`nB;fKcNf3)JCSOIe6rnlJ5aK&vx@S6@NTcn(x*hwl?oF!S{L7=-x~Gjc(O*y z4h8J%&%wE)%E+%t_f2V1%rEDkpTm_Ir0>@*EKtr3)rIr^3-bTgz5cVkw$%>Wq{4@@ zQ5{rK**2mnRc)g#>C^YOlLL7Lbo6Z)Oy1CVW+gi`M_3sf01E>PgNTfRi~4^0w= zrU(OY;IVP3IK)-))YLUhU7MN_D5*K!LQ_(UXe7+c)4&6>_;0wx-iHLT*Ik^eL;vq!K@Nctb=w1^|GA!7cQJSZHy=98`Co)jv1*9EfujjpF0uK#Dw|t1ZRsqhvCG)*?23&Zsk0u0H>XgT0O%b(+l={7BmZL`J* zlqq9tC?+&_s~cwc^;2?})P{J-JYn%8_l>AN7qP%fl{L)RLO6HctU*En|1~je;cbmZ z4QVnTa%2nQC)icGd{P;_?-Ln6LMQWt%HjpSt3YCLV zH)OXStUxRzhy)RgC)33y_`*uGR*x9QxH3xp=|nyF;Rq|iwa^|XiAyVB{>v|iFE&Uc z{Bxc${Xc394lZ-PTU_-MJ02MVMsMXcT=3o zJIoDaE=cpF$M60C!g@-s{s4N%0?ZTPbP`;1!21Q(CsyWSA{--WwK7-`gVeE5+GzyV z&UJgdw~aFT!gUp+@{dxG4h?u;iM#y%9#3#&a;jYA;=+j*AY2ZIL1KiXN@2FM^lO8R zOTErA&gW4E9Jj#WuML$ygD-X{UhRr4(U4lz{;0@)TFsh)E~xYbC;UeH&PIm`K3}F0 zv7!=WLj+E6aF-KC>>^nG#c5N<&b&&Kh*;DitsmvZsW>xg={E8yaZ7+dKJDZW03lFz z_YWYb?BFB6)Vngs(#U^bf$`Eeem=o9n(WI(l{lxc-DlJRYP`BzFoo{82%}J4Ucau` zV;xts6tdcPf;s2%J}m~Ki;BmkLclo7WZ%L5r3IGA+BxPi$m$!I>yrdlcAyqqkik2y zZ51fq0@2>glZNZ@$PZ4f!I!zaYrDmN`~yfn!T**K_c6`UtV!iwS*PNezSbV|)5mAj z%LKZl!qGP$)%73p6j9Vx`)GKZ0{?2HoR?ITeu2)06&quj!I}rlu>?~q$=!^Lk8ip= zrMb!#f~qfD{W6-$N4#zBwY9|F;Gs_*ZngZpORaIi&3410t32kfJhBS9=rE@KF~yxa zCs&naGr^2t4gOwsSaO=9HSL$Aqdb7bny_Z9J8cqB7I9Gis*JIISZhLG?zmR{S6VR6 z-pBUgl+pn_E6LIdg?U6RKoxO}X#J=>sVNPWc9}g1ZAw<&etVKb3e8(j#%>BWU7~4y z(x{8bfa&j!%sYYFn{}@1odND1?70+$orp3EJ~q!J53y?u{AFsBpK6FN^A?;SxVYY_ zLPqZFK9BWt-%!9($~Q8%73fJt8BxWUAA(iK+8y1-*VA%?Xbkp6wn;!v(>n&>`(<3V z`BYkGIk@+=oKjr**t&*pBBne$| zpNl()guJe=*ehm~=|gFA0gRg&n)N)S-;`~xXu_No^&UTJKG++0oqr^%qb&(buKNR^ z`d~BEM`x5MSh)&3{T*)OOhN*1+@Rh{98HXk#hP>JZe`*p)j}E>lBFo`_qWq<;jhIs z^2r|=AFHsApR&9l{;u6lYq#|Qu~5e<60Nzgxiis^lO!vD_YThWDS=vHf~Bkt_0oYV zJq+`m<71+G)^mW80i&}$SBULm)EysilFvGxiBiZ4wuwXFiH!t&r!xlU#L38gJm{C3 z{8i{5sK|&4I5bQ2N?k?B5u1C;+pvJxE=x{$RFe(*)=H)JJeawe=@c&G<$V{aCQ7A` zH4|L;Ela6AjN`?iFoN|^tCZS4#|KyGFdm;O5)zQM#6zr^jO{^_h(@(JW4nSi+eT4p z8~YZ*f#g$Vyd}bA#4jQ(nrS=|3C&E*t@8}9^%(JaEle=iGQ{~ooCLYyjs`vvQWagf z_tfxNZ@1m!5%T@p+}TmLoiAdr1O}`8-Z#0FKnx{Ll5woDUOl@T(q_7# zCmLdGPc|h5bw`=%gEB)KP4Bb^`52KZl|mm(k=K&EBpaGXSEf)8He=Z)VotS$uzSS z@_@+{n6#oWJ`_>vACYx%VBsZJJD!!o;h>j-*&s8;%`m`z>G^=)R|61-(r@qe^txIECxG~v zMBC6T?G)4$Cn#mB6P@wU(JdI1fHA;~Xnn}Lz^J3H=bMZ~A#-1osw_J&)klteD0TQ2 z;=_!N^#Anr-BC?+eYc_a-U%I~1XNl;ibxHh^cs*Fiu4k?iu6vvP!vNzkQN|xP(VDmtIs5lqTNb^St-I-@5DGeDP|OV^jUqhMOL}( z-qsP5YR$0|ANLTJ*SLf~JGoYgymTxfE*-Pw_5NSyWu6s;1mC8-Y4^BZ=A~81SvXaF`v4fMmrt17P4j&cbuQ7Yw{DI1q79uFg^t%`eRS}bKtPm1e z#4yp_;?WR}d$(b2Qk`a=wFmOj0xwZ^Xcx>??z}sIE2K~(T>Y!k)-l)K^Tx`T-6Sq( zGzi+7qSGmIP1f??0rNbTCb{Ng@smHklC#qfE_clc{mYvKda()*RoNTm7m3TX;>p(N z=pD96-S8ES1lK9G8Q`PmL?Oq7xUGg8cTlony>+kXHbV(aDEn7kdu#cKquLMis0Ul4 z6_EfcfET>!?wE6gY zsEX)tZ<40-qedUntS8dkkX&5ey@L7O)6}{K+(v)J4PC4oBJnrcGtcKW_Fi8np9wpG zeo(B0+ilA`OxCb8L4H;5moSpbj!>J0-fOZkbhp@y?zk=xoB0yc<1Fs&!od||BxzvE z%4WKy(+B(IMH=)b0pH4$D(aejcd$OdvkLlAj}<3{0aIp2?Bgk5j!~V8UQ)L|*7(pV z=FZe!1SrXQxtNxriLq5HRE0NmI%ee;!r_am5G{wU5JRSJCJht#G9%`$8bqNi^)0&v zUug)sn6PS!#z8QAAIN+HoTwqRQ^QLO05X0oD$Mvxymo1WXMYs1?J=$qK49 ziaYbNKQT4WDlqRIgblWTF@?m-Qr}y0-?|#XN?`nDMUowVwazd+tzFY>Fzt4#JVT=q z9+LC41w07+(3L4g2undU!7znk$Y8kX$`8Q3^qt=k<VwQS zZb7jhu2N~33sjd-wUek!TIJtizq;QkZb7$knLeR;yCd&EfO3SPDUhpziVr$oq;${4 zv;(GI|7OgHfJH`IV8;>a_5j%P#Uss)_E2Vet|?L1^kMEF0AT&)>tBZNR@u2siSF|#}lf!VS*n4SXB}8o=~Sty7TGyJIj9n?f<>65<*0& zo#$yUX?@Hx`?!xyqsAEP1*esWEc}@=9aHT1PDytG-zGo>y3Jx&oc*q8VC*-k(f`yH zCIG@U)TB^?Od)C9D>Zzk1mkl`Ub@vgNE?!tS3nu4($SLexq6#<#}nEG6%lozv);M{x}Nu<`L= zwQ{2I{`~RV#Xm*PGv~1VHTBXxLxUB_-kw$3p<9@oPSgpCw+@!R^{xo=iK(*2I(Ix! z(Hk-kzkBCi=1)&7X0F?UT7Eo1?U%k}vgC_Fdx+aM=4P~sM%|C45)YRQxru%_=<=y+ z{Y$h{T{~5VmpIgj#$VuTvV%+!Qd{>;@>C|nq4>mNNSDw|HcFGWKeYfb2WbK)^*-9Z z{2$wDG0I~^!jxDWTHQ8Ww(r$1_gn-3bqQ@A!|f!UMh8}uQ~5Oh^o?==xh^A z;QQuRHbrBQleF)G_t2R3#Iox`Y4|bm+7=(oY1qrK&of)=H z+BWPCs6BFL-;{_%nlfwC*FVfSk8oB61gymcj5Al58|CKV`e#ao& zjATc$)0nlm~cfs&go0dbG-Dfr%|X5*XrZF zs(%c$jhpZVBHl?A1QZHR;oRhTu8kEm5<&Pnaf~!g3)CMRbN5rc(;*xW2hV#l2pP7s zCx6fUTu9IB#v-Ux*WxOaIi(M=VCH`X;dE94V$9;WxK2JhIJTM501=iGU7<=^u;RqL zurjq>P6TXvjG<&FB>U5_)5FwiPGQ>Typ4{(<299;>tbE!cQvDFI#^b0Y~LR?DapKM zWOa3d>#FLCh_yOqd>juKPRKYUyo}!VU?u9dWx7dvnc-2RBF6F&Guo8UrVC9>R6*A+@ z#x4hulGcqW`pbgOSvw2eg}haR^EwmVp__W=pAT7%?L%7>7&6h9O!!dKZ?nxfROAsQIncvdn7H-_w4wmU< zS&i(@bUAX9mSu?WDzKNISSW_EB@gTTG}Cg*X-1?E4>s-ZS6eB;{jcq_48y)rqdykF7gJ+&i?i{wd)L02M7Q{2~YW!h%cG@$o)o1nRB$}OU@Wf41hU(csZvS`n zsd5uqP>(y{_+xNt244_HMfaHVd)6~$)w*AR5{W@Cz?iOV(a{Umel+UkS2*6zGL@vw zK(nR2y-(YOPMHYSkmJo0HNfgp6E&7pnz7Mxa7q|pqBSds!11mHi(uSySo-jDu%K$P zJP9lB``5x>dk|<(m5RrP@9u#9Zf?DNY(W=$2fAJX*5u4q1w4#5vxKm^IQnVpfIC-) z*F0QMmpGkZNX_=#T05xjpKr z{=qrP84>paYQyQivZYNDf#K{~O`NAp-?to01|xi{`z56Nl5Pq=W!6sr-}`^@LIS?&d8)o!8$O_0*92 zFY#I;lTT=MqbV`xdx#iA!>Fx%uz1r00=t z82tfcJ3YuORK43Jb^I-;z6C(2WO!86hzy7)mmXOXTggsfmFJ&rT-3`F`MGSteuGMtk%l62D|VYI=xz zBvDUVB^6y zKhpK2FxqGjy=!he7=@6{iI!7|{YgoiMA1r)GZH&cIk~7`88e&jveZ!$XHve+$*mO% zakcG_icsvLWy%gfpW0vDacZniN3L>AHcNI~a&5D~t5EhseaAg{mK6oSb}DCv`x(WM zeVjV7w>4WU;apA4f|wp3-we3>B;P~6jDcY3n0;a&fm(VQUurW`AqL3?3DxMXViY1@ zTh81Vt`Ggh|0xG?&KyX0$q?s^XsV=2?p3x|>kY-)(WgE0KhY1>0xWPc9f0Zj8Jxto z1O5O$v14F(7H8#=XJ;r(7XbM3i_43Fdw4}Z)RJm9TkCGx+sPQZcnc!zT!~@mlXqss zfyqI)A%6gEt!WJ&*|&>6cK6UPer0>zCLVWgj@2|lY;`1bb!XpRF)3>Unf;1W%rA&1j#Id ziCD1#KQ-95KQFJZ#Z_M9U+m#H*@scUOZaqq+n2cmnI&C2MvjMbKe2JkI(Qg$*R%#2 znY+qaO{@O_NcQ2-@!h}}`<-rzEpN?->4`osPLsAuqwq14;o|;Av)V)j3H4S{Qi;oC z9t^r0Q(5Xu`P za?f@(?%rc^0JI3dGzj{VZiO)D{C_``{U3ln5yN)#)I{pnJwklqMInpfcbwz<+g;l@ z^M70henXR-kjcM)U?X5s`3ccyJ zs4qabHA$K|^r3?YySgL&4uMDfb7o#5A_^BGA`Fj2Bk_ZVbQV;5aa41d{5`8^q#m!< zrCp$9iG$P*Z{T(}q!vct11K;EA+hPlTppk9SQ~OO9Vha0#;iQiKq8soXBAhUS(snl zVS>)0bU@$vMhnY+AL^u>tWWzS_f?(l&wJ2Fe#b-O#L0WV1xHtJKR3aQ4Tcb?Nsb-? zzU^MOV}7|<&P1PUK0&l^ea$vd7Z}zLze|rB#^W?bz+L8N5(KBcri8U01 zwCnuZwWig=w=0%Ok0r}m^N6`UKI1uI6WRL>+^S|USmmJ}aRR60ag8{z5SxnL=V74yoI7N$QGA zgxUn=!I}}XnLBrQj3}kE9YVPBy$dhQap)M$5yd;jX>PN^ZPzc(-YFQIzHDv0kvK1Y zCid5)EjzhbMlEHcKTVFKL3)=IKLYl1A9jCNsC^;RAahKA9qU^|^}r8a^f_`G<=b1y zf!6?4Hl?FeRobk+q%o1;#nmQJ&9>(>V}kxwC28gFk9X{pL|;b=mq7 zl*|`>gS;Bwf#$@UkPAWuYQKr^Q-Xu?)L$hAa*%jZ67y#_t8~^v2Rc>Vd}fnsJ{o#?ystup1vlNf64X_lJ02)Q}t4hnf?< zjPJ_yp{HU|$}!#O1KI>0@x)lOLc8tf{&{Y`(!cHYYTnNi!!HbXjYN1uD=iCdFopkwG7ClRHm zU~{qoUJZc`dqk>Q*GZFFQ;G4#J~p|-l2yX54FUBB(rd~0ts13-%d(LP&au-yKPtw| zt_9a)ru3bDa9!Z0ejcNG{Cc$a@&cg8oF>@fXxeD<>HxOTP*gtra`p518jlZ;(lIjN zi$AbI2=jdTI{JfuB^%v4aqa};Hr%U{6&K`P*~$$*#l2v3ic@07?pRWRr2w3&O>|u3 zQ0jdrB2k|7{bVZZmbc#R7Mc;9`EBMAQFmml?nJABLBNUjfe&Qbik{?y2pQB5 z=d1@01$7GuBLPH4L;@G3JY$95oaLQ0IQ{^R@!T4|o1ki-2W-o}S|$?Q51D(6Z)hdd z*oZl^R^~4{Wr{9(9AkHT;~L(P_7;YI#`l#MaJ_(=>)mt4m8Nbkij(n8;Nb|0zHYnz zPK}%Q3WN*k>q*CKHmJ_bSBL{zCGA(6V)E653}V}#hN&+o$ieJujeoa&kbi}Oi;@_z zDb4RHux^7`d~F$$bG#hgKK@14#Pbc8WFtn662Winjmq9~1H8;X za)rJr+mmSoRAssHdd!t2lMk>prpY9Ml{2|V3->)OSiUS+lpT|Qln05YerjjSjf!(W` zM|P-TQ|UXpN0e;B@i|{O7Lj91R}PWBI$u}q`un)AN#N_$w)a=Awqc)@zJWfEe}O|D zsR2fv1nbs#o-Si@DF9<(pdfd*QC-p6DI>x6|BN%`Nh#ZSg;d*=u^!ho8$EWAD7TDwcgO;`5KQh>$rSD1C+`JZoE3?-OCw*kRv(;P8#Q$_SETo`j$_x*7H z>A%Bi6SO}tIR0*OfO7gwqJ2Cw?@kc>cf|5vJCD(0s^rzGZlTMj{C<7JRP`h@r+GC) ztCdWyd9pf@Z%5!N0Md1Sw_$zoY`|-8_p}&KhY3QT>GMogxdqDgnwG~w5KO1Eg9Tad zN|}9V4K&wQe7lYWjI>gr5f=wHD={%rM0LaSU4vt}p}&h#MHB6s2z zMV(_KeEm)$N#GD+CGf#{^X81GO$Vsu$S?ygDGiFv@(IEPLSmzOy`u|<&bAW5Jj^rX z*(_`!S1(5|RMbp~@ANlwMJ&M4L00fag?c@t$I75n?9xCZlwxMn`s_ko?Q!?93;?cK z8dhk$c3j*sfK{U7Vcsro9)tu{}12xRa#h!xa#Jo;|fFbZC7 z)LuM%&hvrfuARmnksouykq(qq1Wyzd{RcZcFj9t#S(x4(bJRV&wlH#@rlNdf&Yz;r{Q9EKL01q)!H6)cP?vD8{h--w z?3QB{)JLF=n<9J#$Gi+F-DY$%zE`~MI+N0fV%2#na8`+rUsu?m=BZ9k8IcE#{?02k z(q)9hINcbmeozTG{rr`C9r7EG^0gve$Ran-029lX&epTxGJf5n1baich+Efz^|aRr+vz zt@pM!u7`$#fAEMl`^2b4_=Dw{apc1Fs`NKE6)pr%UV})ioQDY-YM_a>)}*P#?O&l> zb+aGp`&+p^nW+036t^ac?~OTI_fV)*h)3-@8CO7}`zs^>eabolmSie_0E%@vaKRT` zQF-P17cTg$WR?;w_L*`Mmb&)S?CD)aJc=>J1KyUt?qqK1p?)L0%7~b3SaCL8w@<>& z#Ou)RJaHB}FMn*&L-=Z#b22_&qg@~NDqD`M75pSECM*JHRz$y}Mov;hc0RF6!L=|N zH-hSMysM!M2eP)1fB4!**$4ExUF46o-Z&7x92BSH*k-&AdzB=AZJ|P9B(Q;7;$xVC zRH$Erp-?2(ZseZ6AybkdC850|j(`{%gGugDua&W7`B7c4hxK8vE@#|dFPs90B0j)q zUmFKS=ooi3FLI@gLK6k-qHQY#`CrI5Bip`Aw$>d;dof0r&Almy1;~^6P-l7#d?O0i z=VX7;_v!sG))lBh^rL>2Ukl1A&GV}n=E*`NVe){!Wg1tQ9YxIGr&ivrxxJjfej8Gd z)T)^R%|ggwpO~WD9Fj@=G#1yIoFvz@r4@*jOma^~T?afVGBu`da99Q`aMm6ie@}rn z!)yz0>>JyuON_mHt{_*=g6N7c*Jt`Z&)smdNc#Nj@M0y0Ao}X2A8CAH)p${Tlby)`B^Nu0B;TZ|}1eydh8X+g!rO z&g;9$dF1CmfHBRa@M#|9llI9BZj)tt1FYX9&DPZ8+D&s2fKiQhDrAzp9Nk ze8Icn2^3%2SRB|fE)teABidjJ{AG}Ap2E_QB4Y6+TeG=dc?p$w9XiYIPcUTY(1$<* zPM^0vD$@e>B+vF_R+fpb_)FZVBf|K4^xo3AL~(Whq|Rx9ZCUU=>~Sr69gB7|36bnS zGl)$M8-551NJwz@2lC$clwJDFW%|8zML>97{*yubYL0X*2^1JjJ9?o! zLjdLrDmn!65d?I7h zkXtfm)Rh<76)p}f@RhRs^dvvZ_-$GtuHdaRJwk;+Jfx z5})S!Dn=fo;A1*HIpaCCIImCGy7?w*Y-q2>L*#m7*v{g^z^jhfZ@;5%&e9s?sm;me zG$NYI#iqzhTN3?{56PgH zJX4wUp)5PTN&=5t>Bff7^PDDHI1e2^5ichGfUh_>CqStSk8D-gZ)ufw=)^~QRcr(i-4_IX^C&ZE8CEHPbCLb%w} zeLe5Lne+uB5-a+Y?%1W7tX%6J(0qd>NN$!Nv>+8;&UwCzp;m9oa0fH{Oi?b4*G z-;X32zH}eY?9UhzHOf7LfKQDRM#H&U4Zu0jed6C0?_zjMjIiCF**v5*I9(E>lI$)A z-6LqZl42zZKA-^D5YVVG5c_-|zX{Jdnby@n>2|F`YH5(S1q}&lK_5@2I2D*YAu@J& zD8oZ3?Mw5?OY4s-+CMWuk9sExjhR813BU-c+Qx@N+UDM1{d*~%4=3GDC411}A~lLu zEM)hACUdt&Kxt7(BEfPC(?(8tDV=`WgWAzUe+J$#+2o0nSknUq06;bMp;f%U0?C-B zv*SqTWbM?eI-WU2e@5GO@-5r7AT4eQFVQe54BKLG=5*G(EThr#A!!EKEacWh49&}GnAOX%(b%h1b47d#P*7ZFaJ zxM-UJlq7wH?!!IAVzsJq!>bjhaNdY1Cxw%dchU>?UC3KT!CFDCzi+*`C~}a zb~x_?&DwPA7t8s&XFOwCBpHNW*Yq_C-p1y@`)FrBbZ5cB`z>XH@~xnHtf=LMI`SeBVgO?8H?Q zL=J&18o!?8kYZQO)XWfdeH6KRVVxl?t;6o;oTpJtxtTQYg(uIvI&Z#SJOjJ(k3rW)}Peh2Xl89NDsFA4= zoAOZ6mv%V>vAs{k2_=30B*#%L_Wg|owQ>XYAqM(c2iFzXeM+g4KCD233UN=|@+Mjq zcW%jFrcz%adS7RlbcvN}`HF8ZKPPGZ8un&k+!W{cCR`q0Y5#F*;Y-B@LZEzT-Cjco ze%$Srf9rv}@C6t_l-zdByr8A_H5828TdIaE^Oc_v>nwi%s1YI_~|q|{)tw4rxp;w#8UeCQe2 zr*fC$*8i}nwN2^ZX@T|@n!E|k_rqTBoBgX#jMGO?iv<29>>wLrRyW&-3Pum@q(Eue za$p=fnY!UE^*%0s3I}j{2@H%Xb~chRdXO)Z=e%}bJo_n~9--$Zh1xHQzIZq2eiH5k znEVt!T~{{U!3#|JJqv(vysv!DdStsOM)64;C8Ja1l$p*J>b&6>LTd1j#?OUUgpM&K zT;wpRMk@c?7N(_yzzhJG0N~z>5s^5x^B@qw#;4`J_l&Du^M1ZkedeqV`Cx!jO-yDn z^T1C})ym2S#)PR_4tF*0sfZ{Wd~segt#5JN!xgX?*&LvEMfivk;%wXb?oKPH?W-tl zD>6!LD$I*D=1~@z-1s1-(*@DcTZ#=t0RR$!8IpiT}D|> z-96wC;nR|C1H91w7d22z#)Pm-mJ#4CXcAn+g5q#`mqWU}_1uUl7}aUnFp8JSr6KyC`R1^;X#c zqIkCS)+@?tB0KD}4jG)XXW*uWT+{@Zvg@AMeB%B3S_HGtYy~tWv&N#WEfmil%fvKG z2XudxGtg4PcNC*Gv-{62iEkIAzP$enVT1P~o7#aZ>K)g&`k%A#iczM2)uZf5QWDCE zYA3H(Or8hq88_2JN9izM6nDFE6NO=ZIN4hX;Rk$Vuli1&T9w`q@4I3;`mH3!J8@ec zId@vtUBX2MuWBAKY|8nq1XOh9c)&)%qML2n3n9gYdKyN@=X1rQZzu#xehYdSVel{5 zi%7JdLMF?HmgxQ$ZH*P(<4U@fq0leN%=mvC=fdGyhSotdL3J$4-%g$IJ<|7(i#j~T z6{>Uz{R8MPka~Lqp1M5E?V%5ZY({5}K1MsMOfW~dQzt6TvV#Pk!Ga0YXTJ{lG(5yN zwwdU@&o;?F*Vx<|T!=;$pKt&NqC(k#4dBPplIBMDm@ov^#o5VcO86kJPXEv#g5M^o z@A!S_tb*&_fdc#Pm1bZzHmo9qjsNCBAK{Q}L4CqlV!K%(-bSI6D#qQcM1__&dGIQb z!R4>i=zpXn{XBXHf@!rfbS~f`3FG1A)#?5vv7{hCII}|c8mmj(J(|JAzgAxjb-5CP z;sxmnt06!UfxK+IrZwpwjihh#fUtlm|4`kl_!V5&7ET_bcCR;<1HPAJk5i4`8 z$bMPr9DA(0wOpWIe+Y{_vsT{YlGWO&7S2A-%&B04F}H>F;NV@ShxWok%j09-Vb^a+ zqzm6p89NpW1G-O;t~8skcE+sVg}jT!MDzWuTQnHY_VGKkYNrsl@u^mvSj<&apqI{X ziv^JMDMl6rhDjtm#0>Eg*m%)3M%dfM$%!2a zzDoKF+f;9B&5esQJbEn@4v|em_)d1}_ZR2&+=lKorz?d}VMyrLm`D7>gmDNBK>_<1 zq=XLf4&M}Kn7xAogL{0Xz)e()UmsC%U2he?9LP}1)8^x`I3e62c7QR?k`5K1jhQ0v z6i}VMnfK@{?qBZ_^G%F=p0e$0O-%2lhOrJ*CB&5t5c$7r?SOIrq>wRsOl2DVQPsH6 z%!lylWeO1~JV=IyUphecTNAwN=9S(_P~^Xom-2rU&f0@am7C2!uy0BAWMx>!7>({O zg#nVg9NItP9ihwN_kVpPvn+>7EK)OUF4^`i*m|1aq77qoglmb7hm5~5HAsUM; zqb-#pXo}wIv4Z7wGS$=aGj;QX5Dwx7Zg1r|(i$`=gmN_5FGkFA(_Iu?oq}OLx|nCG zrR$uhRU=+%SRN8%Rmw_WBw0Jui?{g?VEx|@qm@ZwiC+kOX(wM-GMuJ~7vqd2lpTZ?ZP zPpl#+gFsmzmy%Uy)L2_RB{X^)BrR`BWx-8OrnbUbKt(k>l3>w*ATt*Z^xY@+5XgeT zX3XLutWn?X1b&M=VRp{~d6$ILM$Xk1X>^(se~QzmiFbsio7mOv5_-EsgQ-x;H~3OG zpdMSs4FWMj$ZRA1O3IfWjZrQJOt%PW-(1h&+slvv&)kSs^u!-;aj;13O@Z&oCU;eI zqlNux%d8nPwBm#1gM6iiFJtUzgSXYO;v^l2DnSxF^MTS7N4FA(i;vqdZ`a0?hdWl% z3evHgA5J#L9O;%&rIoPP_uLMLRaXdA(R6c*`4P@GWnYys=^XFD(Ugv=F6DP=ss97Nv=CV;ggzToMv5b zbbpZ$Xsg*p{a)M{2f}egn=2|OiFv)NQuBhx{YH@qff1-zFsM05(J+ZU#O7`bN%R(F zm$#$ozD3*%C-d75W@@uSuAa5F#DP;RDwFWe>LgP4Yn|II;CMBzfFh|C<}I@BBpXOI z=N_9Ca2G!^O>u@z)}3W@Rae)jE&o|*OuP~*F(BzHuXt8}E$?xkrN(s-%+yhxrQH!P zYfXnECHvZOA~TI8C64R;ivdqGgd7+AL2Im31>!XS*Tt?LAH(7#DLt?FIs>Bft;xpC zS*Zg%sP{-hT^%zVqEe>6hC2Xg7<1kTZaEy>h|xl7PWsD Xj36W+xC79A`Tscsze?`kGiLu6piK*b diff --git a/apps/sim/public/library/best-ai-workflow-builders-small-teams-2026/cover.jpg b/apps/sim/public/library/best-ai-workflow-builders-small-teams-2026/cover.jpg deleted file mode 100644 index 2abc7d371db0df2a974ca311703bfa150735c5bd..0000000000000000000000000000000000000000 GIT binary patch literal 0 HcmV?d00001 literal 29204 zcmeFYWmsI#vNt-o1b2tv!QCOagy0PB&R{`LRY85X_FRK7C0PLSHI5=1YMDQCK>CGD?WGoa^@EeO1 z69*G~kP^Nn#0USeQj?QYv+^)AGV_QF3W}>KseSsC{5lB4|K~w^0qF3MdC=8R5aa+z zbOXuUP-w{^t!ZD*z-Ya33@%G;l*h_kVT! zk1t@}8xv;#$Gs#BBqj@CCd~y4y_S? zQ|oGzIA`Ga{%2FnNgmX&y+IvGc$4|1r5gZLt1J)DOzg~Vyycw8Ra%&kvrYtFi8+Ox zD6C=UsLTJo`Tt7b{~r>7(XW_08VQi+)^=s8K8!r`=tA;+4S5J2Z*V3j7uSfB0d?}FaOFTgy`{Y znzBnMf`Hf}TJqZUb1N;O=vS^=?`Uyj)jK5k%QoJ!+);hagg_lVJO|~JG!5^5P-RES zEV0eEb#hB?1E_;zI@MHne#PoCLVi`pn=NU%2v6(+u)Pi1qK+K##VT0T!}gJIOqwKQy;6jWg20XCTu=&KB@FpaR3ajwd2ShTbO$S z2aD%#@0|?dIGBuo;+{Zvace>ii+l4rRXs(MzMwMgI_Pxk7hH(MY*Uy%P#s1{FP_$V z*s_V$D8BYY)?i~sy#hOmxpZ=9+K+Y$} zdef!+ff?9PFH)}P_*K@H$GmMAM>{3F0}<64A4gN?L#iN5#s$&I$om7O)~7|Cz|zcS zg8>;grK}J!YdR|vNC;KCo1g}kt82i}L3B*99PnU5{o+jpL+f-qJNB$FaU@xq7T zG7#$2Oy1p!FTijgnPiB;VDlUCyY$f zISmn0wt;M`COmWM&^NSXadjn3qSPWCVs|SH;MjO1gUvg)_%~|)UpGK@b~D20F~S%( zQ%{i^fQOmxlNBjk3o>n_-UOr2XpQc_hemdG(i$~dXY>7|fdoyoSH*uv-7$~$uNx3| zYhrmfCARmn2S)$y^Y=N&sjb`?eJc`#X9GO5gh^?p0f1i*0LKH7tF5|r+Q;q+dReA=K2-DaW#8BF{^_OIyv#S3hdo@A z{Se9;;IxpET9t3x3f;g!H#>&)V?Zhsw~PN;1f$gY!KvM(MpNleRr>8+=12e&Lr;GB z3yWv)=YsWQ07WpBen~={=OY_~%)!?=m-tN2Lm^8VeYEm~v%3--6>Z?jXG$Ns#TgQ= z5Wcm*dc1E`F90pH8KH-rlPz{U9k^I?D<32kFkACVQ1I zK+r(=hadUBC*zgljRpqi?G)A+)-b~3M7!k9f~|gn${#*T=gkuTLT!5I8IY^bN1>^H zViIUBG#rTO_5$G4FWkxEL(M|Y*pk}#(`!t72%3%mlOSF%A%kYlS>>$qrMUU{APuzy z7OmplVSi`KGQGlFsSPZ*qQmIVhyJ+oYpO>sA98E+(hHO{l^m}Yg&pl1yGrRjT^~`L z@0L!Y0}5?H(B178NfRL_SXS*Y3*u^g|OtI)4$t%11ZC{2$-OyPP5R8POA zJ@2j05ruxzf0!MsKnYkr;WGFa0lYeA{OJjvHDfCSaab$F+K~*X8J$y2E>m2mR41F1 z52=Falh0_PN2}o(Oe3mls;|C;j!Ey+5rM{2YfJ4h74xvBuCbMil*kNsgS8mb+2^Ja zS#D($cg0(L14ji}M;VUhlGRWGOpl5p8x;lIW)}I+HDxVFR(=sl;f`N#VFVe|ZhPkv zk@uuV#|m~XC5}?5eH*X1)7btr)BxB7yQzW-mi<+hnlJk1Ror0zPv2yaP?vE^ciYP- ztk~(DY5IgXTMrgO_NK~8dFXOOlX%(>@z!E8bwO5#1QM`TdUAaZB6a=4rG$u_-lVhq zixd!XrCg1%dJnLA50{f9W)iXXhJHF4Z7(JTOKe4}=%QmK#kPLVsfQ~sd&9ARqHyN0Had^{#?C+sZzK)%^ zQWD`s7~|t~Cu7_U5D`f~9Zhb5b=w6`!V zveXC3a1jtX{A6@oYpXtu8M#`KTF{IkT8MDah<0Bt6x z7>_^2b#0gkZidL_Vkuq4t8Q%g>^GCDbK_2kscPC=7UgDSecTn0$JTOu!4-`E<4qOE zKv=rgJuyuE{$qYZ7R{+LDl&sl>kDAKEnjEe3iIbbPw|iI{xO2Vadl^#4#nieK~5gd zRD5r^){4*!z`-v+3xQ_{7@7}+XonJv7K8~})`^~9YgIc#4}rXwl2{->x806<6mG?BvO=0xlD3E#_97n=-IXF!M?ij+gg0BHqp;pAb-J zTPgG(WCR}*I5lP&SJGk_8M*5HHs|`i>xJ*(5kj_KF}XtGR4*?ELH{n1Ylo9%>|p$0 z(yx0BKu74FL-tHDu}*GrShf(35mivHM1YcT5N~Kbow5X(-#B1z93WerkBoJU>5-NO z+9;Nl?+cg&(3iHZl*r~OWpLBcJCO(_A>fS7t>!mn*gXK!PmCSn86Rz>B4zEq7D^2N zH*Ws%^ZZoTW7}@suhJY zy@Re1D?6&o=5krC*rOc>4wp)lhHqyemy;cJA8fjdZH@y2u4A~J!t!#fv}nO@@3t3< z@kTWB^9_^_05E@2Qpc?nXa6OEB}7FRW~!jXP$dNq=Cj(k;ocO}!f8k_6_fQgsY_f{ zy_oKlF8$=7?l0MO1FSWn;R^YfyWWX^(Es1A;049(RH-_60w$3qU}lq2%bVP)44fEE z5f?DEU25v19b)&yP?GNRSdiplL6%in(_J5i?nc)M z{y?!_B1VP@uuz~MEbl-S{+A_Ultt{Q4tnFISd{TB{Jk5V=qgP?F1TE;P3SW*`s3s( zd-mbUki24wu3(^3y;jA!mj($UMZsFIIr?9!sJOu>1p|zChSD_9K#2DV@Zunz!qJ=dBs;uAYmH%<<%KPFeTH&d# z^5Yqr`Qx)=ANT%w;4Qz6f~^jc%P@`lH;?`|P5+0_sNf7GCmH|(8WI8m76uv;5()qX z26_;XP%r>kG)_2l3`{ICY#dw;E*5qP#kU3&l;py(pM1N)DNJ~9ViN)$@B+A_>-*dg z+f=NIQ|73RF8>1fNxf`s$|LrXX7aA(s^g65+$(Kx%EBujBK4f_wy*(MKz$-yd8Sw{ zQf<`!v3j~NwyB-Yz1gxmR65=QsDwUgMNsKcb)<2Q*e3wZN*RDqUx~bdA=Om&Zg}|p zciR$!Hh)^02OrIJY)(#kWxP7+lbOdm!jaN}Q2UbN=*~2KI{7Ix>3JjlP$i8PjNF0b zhCaOYz}%d+1{1B)(rL9#3^q*vxZC8$bk@pngod_Z3Qfw&TDIcgNyf9?I!FfRzLUEU z_NmZr+i?9?~G(M`TDmCG@KP6I@fqG09IUA zjyx!jwhfbBq$|d{NwVlnK$)YPVshB4f~A|@#2}4*cTpk9S@E$Y8|7-wJq-jnd{!vj z*+ulZD+{-`4bq^=S%Aq`AMJVp!0QffS;M5}_Fb)Sc;0=^5?tT;JeIF_gIbS6mnR3= zPtW7I_5PZsrV9JFCW51UX=qkXgZ_jpdIY*#{bfnzb30Z2O2-4{es68{$P}MH;>`J>Ke!C`cVjw3>U(e7D+y}G@Q!$fJ z9W$Tc)@l;Z$=Q)RxiaCOaMra7TjUUhKywR<=Wj zUHWIJSFvF!%Q4tXUCy6*Hh2W8Xd})$EvuW!$;2f90{ac7F>z1My=+oZ&&3}2fg$}& z;cDA-4NA&iuj?w1y3E0gc2V@3SNHS>w9t_+E4u>o{0w5m;Jto^4cbvg7VbKguhrnK zeu4_#Q5v4NoN9m7@xX`NC^f z1GZf|<)i|YCQ7kO#vQSWaLd~{ABq}?aA+lpn_58+2eh8i22K_)b^OkSac5<-+eqh# zZX<1Ta42jC*=I}km$9i%q~vH`EZ;3JdC&-;r!Lb53|DP5eH2m9XpromcwN zj3hJ!wPkNJ=xx$3M<*)U4eQ%B?1H-W{R^c!&2Jr{NsXc_p}oOlV>Kl(S0;5Gb&c*N z0&3q^=36>51B`LbI6-z^Hkp%CK4x?4T}!U{7#*oq3-yRfl3vwJ6*UYJg@RTLeaz@e zO4vLxNUI0V%9SPziTgkX_P=n;Iq-D5==fVzqGv2|4-O)7Vo*dmv4?O?$jsr!O7#Q5SU<`vwN46Z72WR!)(sz}dHeG&y`gnTE!?RTu92;-TbwW?m5=&QF<-W-kvd}?7-BvI1DOBK>D{kM9Xnf8( zLF{RaDqO0EN0pI@k6lvk7?r#e#yST?Dp4G{y7@uJmD=oGm9M4*UM!H-Nb?4AOekB^ zkZ^sxLako}Gl_klLT`SwA(?mBN^7toD$`;xSaYwjE_3N9JUQ)4S#EnM%q;C@6o#Xd zyW5Tf;wN!+T$X$g*JNdVIE~qo)p$EAPHCsC${-4B{R8@rWb*`FY%-mgqqON;1gwYT z8UeOGGt8c!b zRr8#gKcp6>99x8dHgFF?4DcVPp)QgwBRFPakm_XZEje4}uo=DFxJ)H!yAS7CB)_Za z#mEYP0#<-YirRzJjlJ%5Z-P*DO!)q`2I zZl~_Nu6oC+#j7A7OOozLpmRbf}U zbCPKZIT{9CZf-XFLw3Mh7g`d*y*y$a&-vhP69~Jc1Pa9(+hgQ2X=)Nk436}uKh22O zV^l~^VQSO_Y}Xi8-uDy(stis4?cCf_rU=8M@?G>Nuq%8|;IosovHtWPFZ)Ys0q(KI zcGTzErxVv9xlan(&iWW5gr?Eb8>%Q+@o+xo19$SSR$aJz>WFk^$)p8&K__ikc{3eR z6G*TU`mAMxXHW>m$|@Xe$)J)Dm-+kW6apJ@HK`F+qF0-luL`mH4qAU>%0)d)6hkb6 zNaGkJQG}frGkA3WOUHtkD3ePs$b88BySZVCDJ_r-nwtX+pVc(&H*>dXluRs>i??!p zLceY^v!84I5J1g*VVOH~a0Vn{fn2ZUYea1N*m`0fJk9b0=ZwC{WAr!Xmsc zfUM-%Wo*&<61>MBxiYL(puW!330He8>3TehonM0PrOi~&6{xK*&8A*IMzK^2$|RPZ zlXW@kdYyqmR`UfH`W{vq71t|gW5}z-m{%Sr*26yTq~7O-$3J4tHu-ZJwxTKDD>Bp_ z!zANKl~OyfSngtl!iS0)vnY+k@QeQ@je37EdT!fGT3JvX#9nv#1oM@A75Aikl7|01 z=AhCblda!__K03ui+A)jT;-*HmP0J1Ql}+Fn16SVB$10tz%B3zWD!=Sc4uFm%SfEd z1conzHu^Ek_RE(v`8E0`-#fS&9U7D-7HnsCq17*1y@k{X;G}4x{x<7o+ofqo>&w~X zo-E&u)ey+h68Y{FOwy^*!lgpj@6%IR0eZ}ph1pN79bBQfwbGDPfh-}l@p*k|xFL&rZAu$CU|hZ!C$e#ES+SJ=9rdjSwt-f4v$GRWZq z__Qts#Wf2Njx09a_kWW{3P(x`&UD0B4zyn(Fh0a)R$dD|Z47E!?xeCyG1krI6P;1j zgUGc9ec7s^q;+2{3BS-IKPp&?TmT3z331xq;P?>=e#h<2+-iMbIipB5 zQaMi&$84BOBx#&{U6sY#yAKd0AKcY6L$$`}50W+Ym+y6FzB+bQy?OyPIlKkn)fjl& zXMuM=^$*d((k6pdnr5RY6bsI&XpC}F6C@YmM9RPz+Jtd|m7Aw}la^89A2Q=sTXe_y zT=WOaE;PE6gA(#K?Nq{o&Ow1d08egZRa9&G09RKl7wrzlv?B~-=tV>|JS_d_5<{vD zNJQymyL~L}Fg1M#$ds#Nl@x6iD_9!atZ{?0(0rti)_&edqA2D&K3n!oEGmEYfrM@F ztECKy2>e$!XKmDwwaA120#aw+p)PFXx=6LM7z zoi@KZT)_nI>t!(9zhATXO-@rDz>z*ht{)VaUbM4kpY7z}y6ldfMV#f85lly(oBJIo zlsjsF*?x5LF|g+N7F5RbrGzudbzGBrrO`~=RUDO5tDv+rifUq)SxIpglVDMB!b%uB z2b!#6HQ**u;?BM@=dZHipybU! z5J2JB8(orG;0!EQP*t7o|HvlBx&PII*|WQpjxP5nZD2aP0iouiw(OgNbZ#xpa=5^r z=4rmT&WYr{ZFQK@X=u0oaC-8n*a0G?NAA)WfSf6Q7_Jgg67QVNLXeS_92;RHqH(ByZipM8R=$p0`GXIU{xgMKq3JAcDgB0XhP7n%gBB&B(F2CAP6EoWAm z8aYy87&J_#Dkidj*Bie)P)Vmfe{LmnGwQK;%zM9^TIdVIN1@;k=Go%6O?nF z4H8+5O6Fs#8fohobaW{2qBNg=_xd62yP3-6Pmgx|h>*#N@>nB-08BB+e=I7iK3C44 zIn3<%5h9Zl=CMXA;y=Zp_c5nPaP(5>QMr0Wc~VICD0fJNY}vQOP6-IS6Ux z4mOlqX>ELf7t6aGO_#|@0&3&Li4oj2vp_R=2CCr9+c>B-Vvvna{=b43t6z4zQQQ4YnUk1v2LRQHk$rMeyg%Kj09XX4`O+HJB?xp8(o)S+9meg@^^cV>lxHlq+!7x4lpb-AH6`@*JT ztGu#|gjr{@q%U^Z{$8=LTUrGYDd~3=I8Ja~C`%&vE^g*d!6sc0=8-?j-)s57Y8h!J z4oOAgf$#+&MHoeqBh?Q!w3fs!@4e@L!6Yl+gl9cx1kcK9vr(zBUJ90*xM%iJC`~|e z_rlxDJnW|eASYGtxFy!&YbRo7bYlBGO_W#R=RJSvnwcsg@Fc1muZQgqy#Pv_7Z~dl zo}$~LXWoZwr{sNm0ch6VtO^vIU<-_3IkIXp?Np)a?h%n#I<~sw5u8ucvdTqK&~IXW zVaW@@5a*KnGpf67m~j8x#8oxSQ2PoOfD;muRYzqj;h~+|U(4}A3N1dd$-3o#t^pE+ zmxhOjmOOZ6;py}5`#T2V_2#{o?C9(Zmh9O$GoD4sK^@7sk29F>%C z*SGB=`hK*Y;4ByYmePGq8C=ND9Q%~N+Wl8anB9N0|JDb&p9GbY6ExWtFI{q)AG4#5 zpG#nvg?%Gnqvj&^ofHdu;dZJTCT3zh9;V1T;0d(D`R;%^($ajS%iVgG@6_{;@Iue_ zO7Psj*ngC9gAJcDj+c~JXwLv@6+az;1^<&J8hEx$yW5&lY$@N96DwT0DSs^u-`!8G zx7&TY2FXO8;6eHQlLtR%*|ce3-BrUaFMsGbFLumQQM)@OWVtSmV|fnV7#OXiD=jUJ zw9hG}o#!7>Y>CrE`UulQ^z)N;QZCkh>-PF^g$s(NDDM>EjyfJiQZM3xmNuLs+a$1X z$J*oT5@oWC?sojULca}KP?Idn*9B~SBdRaslsc>CUqNi90btIhYCVPg9b(>tT6xcO z7Uj5PS9TLJr!{uDI-Wo-0mP~d$onh{Fb&XcRq3G~Q^J`H0&RtO9CLJ8rcS{RTASFEnp?Hag!2du5g0hE} zQq06cnZg6ha*hlJm~)-P3$l#ipo6LoIgHXI}n`oIBZa{)kW&m7=VNKv-C=a-K)> zlxRFP--kspb~6oJnvhw3-+T8`aA$LRufm<8u44U zV0^d)v_6F8`w@b7MHc9;M4Jc%_B8PbHzy_UVobg)a{jPNe}$VQRA-W#SX>KoQW0{c z>rfyFcE`rK&V{Bp&woUlO-y|C%{bzIM+#;nHg} zKg(VhJ4*7-ur=GCb}M4%`pe@eKHjZB zuoOOKtsquG@Yb}W7E-cQB#{At(PA7=r!DMCRdaBBzLB7mOBV>D{2?5uT!D>r$e@#@2o-5pr59p>)H+cxW^2^4jfOIy%Pc z36$}RCBUd*E^=%i+JzL7W*6dcxy0KR#XT=M?Zj#^!O-5(nIoEW4PO+xpMx0WHpl(v zSGOJExsxptxEw*ioQR0Z8k;nqUFWseU=pi?r08J+9OUk?k*@*Cf3X}Sph%O|HfT~$ zm4wfj3AJ&^a!)dB%KFSOx}Ki@ivEl`eCkkFqJGhFEoooyTZkF+PlR`APH&hxl7^{d zYYsfH6}KHQ3Ipe$T)TkI>!npqDh^GC6{>0px03EN>Ti zwK_KORxMcr8!|q0MV;reM=@Vqxc-IMqCvcspb5xo+#F;8XtCnSWOZ8-BuGgGZ zeRtk>jRxl02Z>-#G3tktXcv>CS|m72Xm#)@egqRdLu9 zS-N9rO=eGP(2s%q%Wee%8f_o$l&x8@?-aLI>dQk$^^a7}Z3K~HZ_uq^-Jn}M1X_oAkQwIIMMQz`%0aKuiKK_g*T;^u7_MqfUUT{+XOuh z>NTzBy;I#|yys71F=bplQHZihop6~R_MCLGp z_RLt}|C8N3%O#A};|TEuS~BC6Uv%|t4d*1gX<|u+us{zpT!YrzXd z+Z{w|t~?E}XD-uLvnfFU)Zb9d1@`3Zh@aRU7~};1G_fVsvN1V9a2>PKSc!_(EX?@L z>j!8f$K$DYMpps!r2ehR8C5_x1+OS~a5|_m+-+0u-XzU3iI~B@Q@xSrM$XWxEt;GA zZ;h^U-eTHTATFSa5)Ys4NF*q6nKGFJ$9G~xEX!cm1WpT#p}2rOMk29oS>|83a(`X? z(NL3T@^vx$(hM?wr#XJ4KchdMs86~G-7f=b9se?k>}hZRcaX=fT*Q+amcAR#@QXY` zNIJKF)|@o=aj>6YZ!z zRt{@?jIPHfL?j$M0)010wHhK9|A$=1uX~jXpyLry$<39Y7120n8HZ`s(>7EBG@&GE5Y z{ug;`=ZAC&^#o?~9=aF60IK;Wr;=G^etFpofZW~yaBI}A>@A%;+L}WL=eq6jb)lou zYWy2*B+x}>`Db|7#Soa}uFHTjWkB)1J;%pPd_$iREKPM0T?NWogr2k{-v05Ay~kKR zaR{vTR+2DNamT-i1D-JoQ4lk8gW zZeeOt(aZ6m9Qrj3FkfmV8(MR@5cwvHV*V*kBA^E^uUo+9$-i|OCNCNO;K5HjOOf93 z7{=C1xMHdpyX6M5Nhn)wk7<~3w{8gT`ZK?9Vvr;l7%I{`yu+}_i+itlLS|)j^^v|F zLZj5b$U@>X78#zefi2`C>w6W0j__Co5gyhX)bjH22C9DSOzkHjCVEDs5~}<+N+7}Rk90%=TPNaQYME@uTi=>w z5xipp3AR9AA=6U_Xm-iw=q%fxTiZB(ul>QU*(oAEEXomlpOc0#^f}lBC~+<)X6lwP zKllQ07X~C(*uRGrV}04D6u}fJ0dc=2+L5$Go6#oU+S{%XHMXYP){3Bc3l-hWbiWn5 zA5|xrrjH>`rsC@+lV}{fdy^BIgf@y5K_X)?dwCpsacE#TZyX;6mpe?syUOs!HwY0y z^B25l(cCi6zh3||gRdU(QUnbFDPhX0WAcuVZ=?#0Rn+^uoFT!*gdV9R!xiMo@3u{9 z>^)kAp=0nMU9@MDfIpg>K&AD>lZR3?>2j{bYi}s=?FDk@vjpDrSi6Zdehkn;w=-?g zPP?N-u(yuB*wZ^QBV3x@?TI#d)|qo`BkD$xsiFW5|I{N-ZvsMX+DDrGE7p zkLtEuEaJ&k1aCQ5t?v~KwCPlSm{=K?YAA%ZxP=mz-WUKnlvYL&lDdMCnz7y@>u6|r zF&OLte>oWWsk$*vP{O9z9du=UgAMysTSxasr3`|~Wr)6!kmR9wnX|xJ;lL^??i5GP zgsz0+L#jJyNt;L#bW>vfiNkJ076(Q(`Q2vvZA>M~B#x*KR5ybq7U|)iSpulMgx*r} z&y_QtkB-DdVllibZ1B+2NXZ=bl7}>61nn{up2|YJ9O*^T%4!qo1v}mNzA;IBK}q=x zZ_Q4h25ZzXg1)}nBMPU-onJu|1t&km8+MUD9Nw?eTi=v{il2_IIy~ao8lxwdf#MV& z+DT}n!^7uQwcc>gYa#QQZCO056qxYJ7U5d(Bc|z1X>Pp$!dLtUQ%8=4@oL9{$@(pk za9|TAUjPfR+oa1OuL;T)RE2gnjs%9AbWm}!<&}7?C(S5SN07xrC2fdEqGW=^P1?Mq7?lwGVjd>C}Z@8Tmn>LP_od zKjkNBGT(w-XZc6_103yQG5dv0;bctOO+r4iTvs5}N+z&z&eGzJ+*oIymuS>$v>UFL z9h$*0Br>jRn)`P&_a9FJs`nsc3(PDwQCQ@K^!bRcawbkSC0CqCk*atu3Q%F)hgr_Ge}`EOrl_Oe+NN12(;kFN z7qIw9Yl;~_^6W@p=SBgH`vvgZpnGgRmww{@@(m%s4PK-N=?-?)eH#TBe;lfn%=;oe zjSC&Cpdg9NAaAfS@d7AE2A4db{P8p+57@8in9e0bNELN4V#J&)(4(T`<9}P>BisiI zutcDKIP<7|T9m*_7$H(pRAozReS76|xbvF&^-6yMeC~tV5jnqSB^lx^xBB{{% z8`+g?{wE5aOXhjy_{B6i#gpLtGqV82$TP3bXz4hiPwaBdpKy>-6t0n90C`b=WI^lk zv7stszE`~;l2U8dG=_p2rXKT-1CCeH8@94{O@Mu^I=^q{_{XlpQ^Q-AAm&f= z)ff5VHPiiPGSq;rcn2=QLJ#LpJ7yK73}yJYymq)5KvPmI;E^SWtkM{t5O2`Fb%BMD zD-IK^B~=!4?tUTZ&rp9!&-MGZx>^XGDKa*45IAh!)AbPoE?)h4`f0y~gSQ!A(-=QE zYoTaZ#1a}pgbb01(|pmuhu)UNU9*W-x-a7A^(FQ1n9lBDz0yk8r7m;Y`?4#1GF`t% zB-GklQ^Mj~-Ca4A1#Df1L`{iMsbuwF%ebS?3PpygGP=Tn&k2-0>S>jpmjEfaJ*BW5b0 z*+75x+Sw&OzXgZN^T@XJ0eZY&dJ9sOmJ@{Crdf!g#l=P8n^6qrmJV$E7c*V9VG9TSC-#+AD~oj=Wg4>KgRBRM znp>F!+MAUIprK?OkpdICwPBLTxo7g_Dy?6vz>p0}ssUzC~NMHId>%!u){m8xk&-of9L z8UC75o)9gtH$F_ZP>e54SDtl<>vW+ThBxaag!=B?c0Ptk-u|3&NuA@~h=&j>7}z7} zW)YQUI_qVq=e)l-Kt?USZ0487+Cg|X|Df2gAFa*PzL4FXj9~iUNrH@sB7jD%ah$k7 zchKrRWhMb6=6_#%{sJI7d@H)^77zo7V?%#Vaj(8J9sHQ35HQHyJ=nz1V;V~d`vL$D zPfVn0uYjeaVJW&qAYun6jfDyOu5+rW=U{I0>gcE2^}PUi5-@MjGl6I|GInt4Fs6vT znqFDG2lN3CKIZ-|K5ulQSqOff7)pz8PkqE0-bV8Im##utaQ-O%@U#=Xfp|;$5Y&f| z!v_LX(+q9kIHV)XCsFyt^5j^f9Po^bFqmZIB1)=;HMMNYD#rE>esS?RqDBcly)#%8EbNK~{<(F+Vvdvnv3axLqRg%B zzsfQZ-h<0B&*{X&VT{n7pm2zOxC_TyV#Ok-$%0+dw6b5O4OBjJY5!c`t<%rv3>nnr z@C*|f>{BNIH{?@R^^jyGS9TkB3&;&{z}irTO_Xgj!#9*~gKw0pL0pwb)gzx{U?;EF zpW?unKV(^=vhmKp7?&~~Lc1imOV#(=2sv?NB%a5h#v!>K_E)I0is`;K;&3AU^`~cs zPz*U1<;9MQF^x5a+XW^!dpRStq*@f0>apRGJ{$ZC+?}Xm|1Wg!+T%ONGUv35BcCSF zmCm8(IC#Ej$3A*xPuDo_RIwwAt5K_}ur{lc5nK>WX6em~Rvx{fukAUydK$ts#fLM}a+P9jO)VnNV6-4Mk@ z=UZcgMS#{XSw5afn&VMl;p+h94~IG(yJYEE!q|IJu2CDjWEjJK`?ktMmm`poA8-r= zoL4Nxq(X!rXbg?(C}HZ~NV);Lzey0{?W&U`asT#udl3!(H+&ZF1hubo? z=xHp5uJ5FTcBF!9#4y#V5tKl)6355wsay7HfW?2)AP?ajiY=l}kU6fOjPU8jx z4EB}OzGp@+hye9Wc7`5jzp(2YX^DTEQ@5`FGt4 zjj!f|Sx;6508@ss_M^jd6I7JNj-LGzGn+>-Sx>Ti+=g;EA=f`{^2rSPqBBIc5~XZ+ zl_9F!%11vAGX*9Huh|PepRQF}Ccn=U?&=>O#nlzxlsfmzA)u1Bt)8OfYnd4vOZF+A zQTS0YS031;OKlFRo7CZMj~>y2-E`750Q$obLS<$Ts$t$lg9TPnVi- z^L+*kMi?>W3=S^ABZ^DA5<4yb8wbRX+}%EW8se-H4ocE|;!CKj)C4`DKT!{4;l&7> z(51PFI53S`)NWO8%J4g4a5w;n>-d_a*od_W^*Dg=H$_;cDbhO1hMiEqrX%ro5jkKQ zV|GayEM@ttflBr9)-bjv^DPN#2S(VE3)7US79))$d=7k91l%%;e2+fUP9{G%cdG18 zwJv%;Ctg#3&y?&Z4A9nNEjo#k>jp1w05kt@$xT!Q0&1bknz%MmEzD(*H3*j&k?C zd!2}lWu*`Z+a!jxd&y9$@W!dSRO)S#opxTZgBuQz#&0H}elfbEtD5Uma?GNQ82t-i zj_xpqFsY-D;vnzT7B1@v-|hP)H$XRs-1K)w7en;Crm8z9(L~?)fR~~*8OM zwok`W%`8xqvegsc+e@kQi}I%!zoffeI5w2o|E_l39Z(%$|715ta1{G_2*NK}Q!Hs; zVpwBzim^4UdiaPR6iwzj337;Y5W}a--0<9E7a4n#9^s`jqTtLI$`5Q&-R0Qx2+bwb z_$^ukCl*h+oqcN_ZWLgxFerA)!QC$Urd|AL@vFL|{9O0N_Nf5%w>DY`J@6S`gK&1D>njP0a#fNZ-bmugGyTlU z{oG?1SEHf9T)q-AzUQGJP;#mbbIM^o+~aqGy*WnEe9h7+xwUUYstVFidQ_C6!y5hQ5iVaKV`_Dzv>frDE-#wtYWE(KUi~YoCB z2Pv6(=@z2PZ&>{l8a(c0HidO@Rm`n&1nWN#0oAjcP~m_BOiaAdj-*}wn7-;xxxH^o zVm5OlS?~*yOH(GStIB+e!T^GDVcFFDE)DmifSJOvfYQ_ZF#<&`d}U#jUd4@S4ZE{) z9Ky#EmY>h2q6&AI%JuPF3&sv0t{8ZDs^~l#gFFYvtk{wO}0cg zTFOirRRbeMc|JP6=hXwBeecE#Z*=klS=N$*OJL|rbe`NpqzHQRF7!QP+;;EkD2hS4 zx6=x>pMw>=sg;!*Q9OvduZw~LLolZpRuWWl#W)3MwMp@ z**A+U5E+c^F)YGOc?F^U2I5bf36lBpLACK-2bte_pV zx|>;QHEcsTo>7*dC;eU-yi4l#vi{l|(3<8Fi7upH^30Us1&tbIF`9qpxbq#P9n8nP z9qR3l_b^@RMi6W+|HyzJZn|%lYu=-cUW~k#iKN&$x;wZ&0kEYbCgb7?u+bX%Ty@&)*_s7S24e^c-au z&PRVtTCOXN#w+@+sdy`bJZ1UV0K|#WtWD&Rh5huEZ}T9z6dABc6x~kHBbYHm6vvaB z4{8_Vc->4-M_vF&*+y}GHK8ltf_7aHOS$4flF2F!XouP50wk*}YFPM?q>R0K!40VygccY?!ybN6KLMnE55UO@4&- zU1c~hXMgkCafZ>bCXa5}R~0JWAu2NDZ*j9BB*~`70*9sZ;XbvX*gE)88E5+HHJdym zdh2S6C(&!JZQ|#6;;IZ&82vKCD(|8nV-`y0$PK2yn&{JampT4d-DysEs58bIWNDbXche!<|h_r(=14v5?N_Pzn z(%mIPgMdg5EeHb)2uintfQo?M!Fzw-y1(_^_5CriX5Mwqd1LSW?C06%Jyjo|$UWbI zl9|Zi)`%4bt+yZAHk0S0(n9v$((2BOb*jxJbV-|X7g>|Q8Ja1@o_5LFAKImNXGahL z!853ZvI-#B75Xm`^;vlSZfwwF5gyNCelV6X-fK=U3+3=ELDSI0mX9NngYz5QMN7hgHxrY~qM zgcN>=etWy&ivQ-B>O4wY7`c}s!gisva6a9hV#l|)()X8%PE6S;jmhwrbiU+XH2Yru z8`LpaQbR^qb(RxvZ+Hw?g=w(_Y+du8#RoQ z5^v=(mIu~>H7m=Dk)hkZ?Y`;q74i0t{e3R`8kIVY%{glq;C0!Lf>{o6+hkZC@sb?;2D#MG@T+AV?B917@{=I<4&`1t+o_oI zEcVzLK|2kbuh4|ucD~N>(0zm4T=MUom)DAXJ+r)ozn^|>DTru49 zBXOH7y{IMJ@LljTQ{VF}y({OCWo_Zm@sl{}iyfg%AK^{3JQ-W;1wv!WiS#+f=tIHe z^dMvL_Q9b(`P;3&=#MHM6tj+-{&cs7m7Db=6I6*yJ;Gl7w<*xbj<_kT%LMtT8BeF< zjWkKDtXa#ugs~1LkdQg$M!cdRf3~O1Qz-L046 zu|2U0jAT?1aE7K1J9%eUdl$Cb^2#PbWrN3;7`1lq}_^1j@#{k?vT60oab6tqkzDy0wl8$`ZOl^8c`hrhny z%XdRebs}*}9r4#&-K(6fJAhlaCR$=F2qU~<8wf;cFBdVag zsyl`U(;VYdlJXFW9RBA^#l`bQjkZ2#I=hS=Zk1{SPWCJ) z8L8}Lo77mvS{skf`egiPR=uubQ{~siHm&}e4h1@n=AceZ2KE>DMNeZE8D`37T^NV` zf|L(_Zf^eu;hI0cg!j=`gj=8<8WX)~H)L{l=6e?_uDGM?9I^C5u&2>I66OGX+_mT{ zI%IHWVO7?gENxUKbPGo50tiZWwA>i4s>~E|h_N{t8Qd^E8>q#EMlbk6K3Buy7+J#mVjAj)whp_?kuYl)z ziZ$$tJ`uIc=+_>Ib5`4Y0w#CGN0U%oXc-wt!CpH3mz%RwgBWzgY8xsfS3dYrFJAJ3 z)X_Mm0c4u0yHJc6O^5I3Lz+4}pVTCz{VzGloVp7#$?8HsY8A z5zVpnV<;+)_dSuXI5W`8NJmyi%>XI}oQWU0QSiYm?LIzF)>=k)S-}`yi3D6s9nDvjVEg%aEmk^3nQ_dN5$PriBkj7;XU;rsy&|4JqP?FbOU05^dp6J0GCw%(As!~ z70}s+T}s3ooJ~2B@yM+P-C-7N2f-!A&&KRv%348jqb+5qO3dg^N>_35%};d==r&RM z5NoBu*w@n3J^#2E8I@YscYfOL;Ao$k;ND*(R!`XcbCORaZ*v8faBbjNmc;)+z43=^ zhjqMfdP5C!DMgV=<<`g8(jVq;51dH95uqQcCDuh^UxF0WDnM;LBJh*|D7YQm1WwF6Ay0 z*X1T8u_Y$bTe?uG5^ZIUkJajdlXYca?!#THpwqkj?)4~`6%_JnAL&K2z-{&36Z>+c zD9U|YC!voI7j|)+$9hyX1D)bMXFgrZj%UqR2aYRL=rRGtiPxwYrZh+Mn(%YK3S5Bz6Gt}j29ZUXL(+jG3HQ~XZlimP-(e=%MU@k;~n;+>7q_{il! z_Tg`D58QTnac2nE)_w_@q(}Y|(J9qk&bv)bBMqQQbBHz(rc8qQ-drmA15goAkQNrjV|Ted*%CILwO{{ya#- z(bErn-Ahn*m)+B2J%PNaP(0yJ4KrIkJy;;^cl}l`&6n?FDfrI=zKX&2CWM%gl770T z+l1c?STH2FDHKo7MVf47^hoxtw=8=-(`aK3`F%8g9cn1x4NB0V`FS4G@fiBTL@0m9 z8U}w}!}egb_&BVA{-c%&pEWx+U2Z6mtz5FgF1}cabva6|xXpy_{ENO^y}9NagthbX z-!w%k2*?8^K!h^*;&jM2lGP;E;IwaVs1GY_ts1Hj70z0Wosbiu!VggiN)gZ(#ho@x z$Jwxlr*zGm<&p;F*|He?mj>`&rfPpj8{>hgU7i)%8 z8+*Zb;2E>PD;+s}3cEDW3C56<&7iT{3ArKyo9C~Y5~7H{dJ_JRt~fet_;mnF~o{{n`5You`+QMB3ii;uvQxa?qzwi7-ap``0zanH1@}F})ao z+g;zzo^2K`0$u#)sLkYP*4z={*ynKFb?|#PM>F z75&DXs33kVUY`o&QZxDuo9kmMe_4xGKKj1gc1^N}J1*ma4~@WH>_@sV&vb0Vvi?*n z-h{6kV>gyaaiLOLs@&3@Hqb<=Av406s~FiG9&XxJ%|{g{Fi8hGnq9gi`n+uXfs@l0 zUxNVMztie`-wK8jdeJXe2dDur9$M*m(;1~qMD~{dXj_0iw9a{;!PrEf`ZAJ*IMFHZ zC8%N`b*fx}cOXmPdbX7~fS8wyGIE*+R#w)Cs)BE^hBK%HT~FGf?Z*P0;*9n!6zS@^ zJjqP>jlLC8=begusU>wcK>$koZ_rWcnj#rpXOb?f4@Y zNOREYbjaj~?L>Z`INp0STxh1nCrF56O@UloASDt77j{ofhC(>5lBUC z7eHQhdZr7%i|A|Mzn<3TWD}!tE3Cnoa-M^Q{swiYTAlgrGv7{JYfP&gRgX7JJH$3x z$fcAX=01;=Gy(y8zL*4hb`S2~v9J}fv(^AFr#K8qc}s{=csvq~9kv$hIB2b)ROUG4 zz+CB*CSjnL^huc7XPxE}{I|uENC+l}Yx3LEsLj-95hmfYER#ExXe_?Uy!89|`nb+3 z6F{3Oc~L@l%#X+ZKqVW0=^V8=`d6FgSn?ZZG{$8Z1|ZM^XasARE!|eGaJnPbNb_5; z_YLJaq!tqe*CA~bi>Xp^?}}v_X)xq6uA2dhY$6Ov`9XDa{Q-q_ANtZQl`+=_=-rTw z!bXgROpw+`L}ZS~O%kHrug7s4144`Od8m^5CzrA z^PtXIMLa92D2o{{%-yg4ap$(vZ^@BK?vxSts*e*iFUFP0$wXOO9kFWgOZV~YAcsU^ z;EOl+AU8moql@+>Wmm$sr999HC`JP*#T7HTPiI`si!rFn72Ztvu7B&SbG$?i5k+n} z6nDXN*<|mVz8QnPtO2}m9+*ZCoyTN%e&Ee_M=GsFWJoUy)dU#0^o+*@S`N*Iz*GTN z2_#B=S4r7BZOoxhDPV_>gt1z}A#Q0}$)0TXWum8f7k(1A+~xL?Nf9Q+sb=WPXFahL z0>gfL^5T`;4$jnnoijL5SbSIyFZhS5N&$YYn0D8AaBSR5`w$1VJ4D=QH4!@TaccW3 z2%7Kzok+tYis&{Sq;sD%Gi3sv+|yP18wOrbT}5>18sbdr?zV;E;MonqY|UL~6^ z16)fgv+E5_{r=MtVX~eWjUCJqM}^A2IdtQK8quUq#tJJrT1=6MmJo1J782kz6Ak?N zTJ~CEZ0ef0lacA&<=ww{157j~jld3;xd|!291!lz_}FeKP}{8?C)(fVD+MmhEA2qq z_vCwmxvbB8a!M?(N(3Q#Qi{)ef)&?CS*i<@Bw3TTDh zXx)dR>!$hxQc&BUdq$i*zpmP}1othwFwCF`vE30F;@D583!2^BL}Zy>J4p`kJ34&6 zm$d(_s$`%zR2@|Es}Wq37^pk0Na%Y<^_S?*YP2na!|TW3ZxE@Psa1(<@r|gW3p&#- z_?UEc#1VZ8!tp(R9%&}>E3Qw*qv6VwW&at{DAYQPGFBWc2A5!v(4} zOmyAHag|2zN61c?*!CmCW=0{8SHNaR=0(lo*2T%{=vor3RBM3&ILSdEf`O+wUrlg0 zH}%D?2+$!8SyFSk?)6s*A6qtTMuH4uyW;i(ShWnN!}v%F;g+*iq}(7T+69sA2NG1G zW3g72BXzKSt^z@TjI$}-f$y2mNoyh2(r}1}#<K#Zl(GG5B(OM*Jw=sNeFo zp1O6yA&-d^m*rdmByc+GLteysxQqPLOoL^wcIb7mev6A<^!8@NM9$(=T6Hx>1DDAi zE2Vibx(mDiOW?d^5E&Not(_UeF>#k@r|!jmtzJ@Zk|G+vI7l@UE>yW9{vd_j*;bND zMvWFd?(&${Hv}58x$USlqyElNjYrItEgsWva^^p2q&X2zQW5@uDeeYX*E5D}Sm3+2 zr+%6@OAVGaexcz|-IeO>MX!+r9p0Rq6AiJY$D0r>3x6D&9Ls50rHxi%$;!^bqoIGd zlZCHCj~gRSnVzhKy-1RC6)e@q(GdiopZ@MZVg)!1COiuo3E`8Cr1Sc}h}u`(y59oM zM^du)ZN`3>eu1oOlTOMwM?Fa#vTZM2auLeCcCr<3^1oo^O~kjW$GY9Jv1sb|c9RtDnGHwZHYFF}=e)K9U@LAdmogITTW zWZ*%~IBf`ItC&A9mFYoAZZ^sXd0D^w^YrN0+2=*U;B=v=LUX0!8%67j{5qE$->;1PHs)H)C zDOt*w4n(pHypWW||hZCO-1^3lEqQ zhi~X*9OKtr-V+P^Q>J2s=isdKC!?M5{wy6KmwjTaWEi`9?k@)pQFTQp`N$$i4X?Fi zGuO*%*SjD9rK3nxzKWc1_Ak5gd2L|Ti|9R4(6y{Nsifo{8q{@O&pG4)lj>B^9r62; zgDV5$Dy=ld@UkV_zl-nRG#7UC+uCbat(8&99hJCvB!ma!(cckgTy@Z7dHN8mbbjvT zSK$)>lbFFEUm!xtB-R$a5rF<&_T(n`iJ~z4X?k3+;z zkkk9si%c|LXFsq>ZqA;v(h$tmW^(;bLsyfxuc@rs>@RBp?ALn8dqDxIY(qI0_wx%= z`WdWs!;V4T5#r!)-qeH@pMh=Vs^PVnoL75U>*~Nk zIUK;yJxohAV*2ID6MBJ?&*-akpu>;f$BL`=G}%%!{|F0xEXr!BHy1hp5Z9Ing(WCI zP?bY;>#F6yewn2#f@4vhUlT-SEf!N^X0*hxr=1xA&6@E^LxSb4e}h^EN*#SH98NK6 zQJ5NdaCYp2ZdYfcvk5DXuzem%j0H!pjJu;1Au=jq>2E-_j79R; zuqigH)~TJ*FE`(54nM~*$q*RTdo*q^B%OdPqX%D*_Fo1e+nycB#h(8LVedGsI(NnT z1ir`1*NiUtHKkLPt-;_R9~v=XwQ0D`dw*1ux%92-um}~RMf9YGq;RwL04{A-ZN_=Q zcdFybJ;I=tk8VsE-sSmf>PhjWgGpMqqtOWrLqB3+vdfKrC%b(ASSMhcU#?8OyN^;d zIYmX%NZ*>bpNc9xIDi$|&q-KK00!7iSx$0>JIX@Fdt(G*VR_^YKMjN!JY~`qpM#KvKP9XDR?d$3=ZVxDNi#zZKk1#Pw5=GW zaMLA4zkf7(_m02{$l!Gfo_|$0{;0#kI=Y9Q|Df+Jx`l|hQg?^B8TfNtH^KIS$p~MB zF3M5duMW_PMdVHUxx+mMh}5I`>$}RA*V(){6CVb6jeNnz9;W>w&SPhd8Pc4p3vO=| z9Pj%qlM$hkv=$7eqBfSvPK(r2#M8yrM-wYjyw$xBNOfmZWKVeRTx2XFhz06fV+p)0 zfgZECb7Br%PE7JV=oB%+OA52*seR27b1`>jbM6qCyHde&zDfSvjb6&cI>Dz9gxO0N z0ggX3qAAv2LYIE`nO&u+wjVdnN-D`ibX_qR0|^?wrxRIZPTX-`k9}&D}1u@Zn$tZ+r54E zrQ)ZY6jvuls}0-JJ=4Lt5Nyvd$l9+zv7&M@g?jodp(y2}Fw*qJw%%8qRLamtdkc#U z%}G_yppIfbqGTR~G>->WYGvXudKb)0&`5&1PA6yh?5Lr<>MYN*+d*|n-dtNA0|WU@ z5riE%0K-h;+uYy66?yOh>%r&CX)<;(Lld>D-63*=%IiHK_Mp%_0A z4)-07kLnfk@G^*IJ8@=esL4nycZ%g@^&;4bs?v5=7h8)+X!w)!SmdQ|Sv45ZHe$h8 zWw_&G!pF!j{@Dw{MGPZ5BsJ|Az54NSoX=fslJp@Yd($|E1pJMh^*j=KzmIWO_ z(-;g@-uEL&cmdYE5`iX$;ysB*BE4@9@C7IsR&| zIC*=Q$(>d2T3cQW0eszT(HcOPU50(Q)xPL8d?fhrX~z4iv-8VFo7D<12ePv`$A98M z0eO}F+HIq6L&{4-i5c)bTG4!Kl0oMTarYgo7sf_mnZZgna@22N^hxB4jKNa_QI(r9rd=A z#h0X_^<6ejn@KtHJ&|a2B~#Y^WxFP1>l?R1CosgU0wGIS?io)Ll}QKMifK@2?T|^~ z8TxrMQ@>tNrxc?BpA+J+0%2o|al3Yg`ne^b&T`h;y2?#VZXZ$oJuX66Wu&`f99&tV^b|AJK7?=iR^WQ~NT|d)nq;BAnK<5pqLu4Oqu{M(Tby81mBf zq@zwEt@qD4qs<`N8$kvoy|8}H@JD*fuzMEnb|FLN)}n#@Mt)1;p}sl_NHk z3+2Zg!Kux;m$5#NA*oG5B0s+Eiu%lF(EX%794k@%cv}fcnyc+cYaqZFmIUvH<=RH4 zaBf5R7WHfQCwN!4_#L!p!w|F|E#~CzMCP3XLe=QsqTpYE^Yh3a%h%j*7-|r4u9wly zir=%9CI@c7zE}~$8K_ZZ@xwE>IGkcQf4!P9m&LqJ{i~h%!Q-HCjPIM*8#9tqYDbfX z2|!ZqI1`6YI#tGHVoO!BbX(xGr^(y4rJ*rocUsy)G5gD2Wk(rdHOLUN7b6=jr}wJu zZX#g}aVfbPE7kReBOu|vJ;Bq`UjI{m;jE>ac*ZjZia%t^-IZuHgS3g?|QQ{Z<33d;JL+l!?TSWR`+dPf1ho)}2pTVOa+0%P(K zOO0k6yJ>FRY*t)TA8@P4E9lSo?6H{m!m}YnucWD*_1%4@r>u4fR3!UME@>IZmTS|hWy8j%^5h>&*q^w>B=1?p2Mj0`g$J-a6@M~j(R)GGDQaXBeS+~iM2R9 z7Qf6?&->V7g;z9wlt9VVAi1%N^={uF2ZU-K)B^zq1As0IquN3-C#eO z|F^YBRU#;~g1DHIwn-l2AKtlz46KuBY-qQY#I{k)+9nQ5!@)hwKz#^;iL_xARRlC1 z(=^^x=>Wp>Q$Whd{N>C4I(p&c5K#2*YEK@xjt_05LNhJneLY6s-DkR=>BZ_jf*v=0vtjc-!-la#rhcq~y zqwam*SS&l7C)wMZ4 z`x^orRNec$_lJ5LSi4x_^(EsisR0#>NBXF0}GmHX{R7-+1gyKUalVkZv>; zF40{!4r!7fE3W;5yT+VDVgo)Iyk>7jc|Z#s3XQN?Cb(D_`)g+wDU>R=Ku}V=e|S-# z#%&59!bHxmr*3$9_nY*b;_vi)Gv*Zp)tc3+OYPZw!o3Yc@|n0K|1@hnp6`?j0Fspd`xB~zn6F_Yg_*Lbr$r>*pzGugjCQqarm}avgM1yUt z8Ul1L!|p<;bvb6S8YsS_b{>9>%_=3~NjU(aLBHDZ-)p&Q4p!bm@p)NUP3}EzJWC8w zlfeyK(#zE<$#{f)r)*N(%gWA(2HS|^>&6wP7^)4z@FEi7@#yUKGbbo?h_Ona)|>|y z;i*LMRh<+EmaRweN=;eqS@VZ<${{4wg{JO(&S$DFr9y6ykEhH_}fSU#SI<|-QC^Yp}4zKpm+B<@_){IKb_x| zPw(u^O!g#MNmf>tW&K|My#^2gK>y={fq{mD2mO!`5fBiOFpyC|KMZ1YEObyJ#wW$c z1>H=PWMq^~oQ(90oT7YuqUuWO4i3qGfPno!2GR>ag9XortcCz11AwD}L7;*C?gtP6 zz`(!&5Pu&3x?rFop#zXSHW_R&?4K}F>zM2wo`1H)Yu>8E_qWa$r z3`Q<}b8nnMAmIm=`#=T&vb1a=#mdy6nARuiAWx)>*=hv7y{CiB$K(MncUJ31VqHnV zi^#;M!vAg}%*MRcF>PLR==l5^ssIR}v@tDWN9G;>t6V%Q^ zGHJ>~X+HD^3A5e*%~YTbpCyvf%KtQB@$sE@%}gJiUEaYC?Q5n$f0KTso3Rz_so4pEhFAp=pV91T>bBl z0DkkCfC(%L`*vz%l>f=v2v}0S{{~2V?cW2q+i?}YxC;KO9VB+)Me`8uQP06vc0nPf z@9Dv*RvF-(cmEP$wApH&dJ*H0P)wuzcm+I0U_>#vT9v!?+NH@@Eh zfKT%{Ofx*lqFqkj-j|UO{ zP)HJ|#7Dlc@+1cm))2ZRWpn-ply-HGsc!#bxeck3Lr$+fJ%-`%$(Z>jPl-34s=MLn zAqeZ!>n;5TbJ&JE8MkO`_DH{^2zTh*kd3AHDWhsgp!861a*btKaU^7}Iz21;m$X&? zcb4@%cr1jvMI0J_gi1cTqhH-{^R2gVwCR3q$17Xn3dDKU5ZU!h<={ae|Y1%N_Ba=p1!5l%J0bc zD6O@3&S>UZchN?g`J8*7!ZJI~{}X^5fh3fj=v$h;fMzfBe3GFImKwTFx{sI3So|*p z0Z{d0+jlapXUQj~Bn;&-B_>>*Ya>-Z?#pO_VINo0`MzNv;y+_Q&wlw>BzAC_d&KV| zpZ68GhT$^oDKW7L1c2kGFYm@m{|0zo^n-=3$&Ex?$(Zxk@R-ULsMP!d11>qDaI5a8 z&k2#`$&X%yfHTgv7w^5WOz1~ZXc4-p-eL1HG5s=?tvbK@^$pryl5E9K@->!&fzvss z&f_;gSGD~&z%mm5WRTpB1<%j_5I(BSNxA*`NPr)sRd+U-%xyR|`_4S$4aayFL}<;G zyANa^ZEsfw?&#DH1hQwCHXe+X0%y$-ifF(YXBe6-D~WvWP$3TQhCbM>@s!Yar7mZ` zweW!*8(i2+ni)O9dWaXxoctR(kl4{9+gTwlyUn|<&Eyab1IjTj+0%;P3mYSJ5`E93 zKumdE=|N}rnF;^NNmYSQnLGkjsCS5N1BA>(z+3a2$)1~+Z-#rFzyN$JjdbdLv?WLR z9fZ|Ag{s^lH3(YtM$yrW0|YA`|Dd$W?(?s2e#mtkyvGXCKk#p>9j_tB>W{u2OyVCC^Cfy>fbTfBZhGhROcyoY z*54lHy`CsjSEp-@g(>ft&&^~x-x)>5Eaw+y8^kSz5#AV^RJ~i4r!U=dyAfummxMU; zDFCxV-*>1&5pG{2u-rs?Ye>2VD!7bTWQyd_pN}PC+bO$GR)H@5)`Gm2Ela{TJb{yz`Cs<`P;{ zM*u<%D-s{Zf?G&ADU2UQzu%rd_L zNpf1*tB@wNjz80Welhk9pUkiw(pdqxT%S%|-6|hXe<)OFr&O<0xckCpGVvP_E3_AG z-&>T&wj_d4=~i-h}?=29sUg)*2r1HMiAhl_0HN}UeH^c06)1j z$sfq``QJ7EPX?Mi;o}9uxBv6upE++SeeBB|ZX&ZrH&2aUnKJ;~RG_oRrJxo%-L``_ z|8{o@=3zD$)Hz>H_N=-ZTFh$!@7n{Tja$FRtwNvT3X!?^_{lOjGTD%+lCx2A$C;;f3IoL_ay0aKV03Ka_vt z(^tyR>zznfJucsgw-0^<@_Vv8J`JKza=>qX@jN-Yb=AP{b=9I8GYe* zk?nvvj8sbP@f4T}9aCFeM^0!1d}4kHYA?6Q$-4`P6*qrs&---ud@P?JXZVW{gXTGw zDb16EMwczT z;@gnlfbdvGeZRXXoMn|<4r2xX^OxRC(QqBuca6tl+L{+t*!T1Idnb*cJ70oKeclay z)=uv`;%t5eQkBRxWTJz0t=vFt61c4+J=RUM|IH!DJ_RNGnb)^yVMS9bI7mQ^#fVS&_p8oamkYjZ6|O_#^hzi!3_DS zfj?Xz`7fatQl*jE&ph6D;TLda2>fQm<@wB_l6`z>t+?O&N_c+*gtC>QFSGez4tt{& z)0De;FNgLVVhunU#gzxZ(SRQ!V2 zAv9AUkN5J(KXTQ0_&kfrobBD`4nuKU?{#5}53A2k@GmlHw5n~*f!0t-KpK&P*Gu@F$g0@%1!89Q%{!;i#(Qq#d-IUa46(hWOex9V&;r4gkG_%3 z3xGL<5X70lHsRpXjup6wuKGITA|V2Nw}n4E6G+)eN%XN2;md{QG5+k;?Mm33b?T+K zeG7RRu0V+G32?z3MLp&D47TWI>{W6hi0Iob!&Pt8J#jV&xN?9jP#` zE#tKD?Sc}`-2a^c3lP^Z^t^C9d6bF;k!hoDgY^VS*NN@lI$jt-^ zay)_hg9HPIfC50npt5~H$G{|EN5jHqA!ij7QzRv$F#PQBjkz1-b%F&|1%n0r2CS&# zca#Tr2S%!IejygS+jx@V@klY}+!csaqws1ekeFc9c}8SxU7|({{pR zhOr}RgeT)+%wjzTD`+5bqQQNTgD)lHnlMqDa($yET+uZD7D5qc#GcgCRx9~%|Ah+| z=GXZG9kzoQ6gO6D>DAOMXdU2i;d&&0$>_+XEQvv`ZtS6z$7^Wes*0FnJvzB`M-n=` zylszP?tF%@-jf`gXO~)ZKSlZt@L6%wEZNZ{`3<1BaXQqqrT-j1QQ*4#<}DO{WbvUg z+75LZ74q4(OhY$uHmszSC_@dy=X>SKLD~7sruMn2np$8@OS^_JVIqad548a!(`WOs z)+1Y5sQ{q^hdm({zb$sXWE%HmSb-7VjKzlHk{zkT9T-^z)|fWp%GF{We*z|$Vy!%(;vOcqQc$u z+sxCdnod1L9*p<<6;^IT{o4JyY})8lujx?z;yt)3td&&y6Q+Ixu&$+8E5GF2>2SG3 zdsDrWN-V#!Oz}V60be$x=N2U((;4;ZCN$RW+jD<0IqiHpd+zJn6xed}+AqLjjeJcQ znXuN5$fC&ddq!HIn)Dn#I7*CmDqWgeR4aN)aZ6SAaQ;s3&ZHg+hx3P@5;q0Pe0Zf8<>|GQ zwq8M;0V1l1{RZ$qF0RDA9r<`s4q?*qt$)D~s^fQZ;x6M#i;Zv?!}NvxLe%?(u4EZ@ z<_sz{r(uh1qs4~NG~ZQrslr;Q#mi!>f17|u*u47*h*dK&~pmd>r7ICvLU#q(;a zs;XiNG8v0S1H;9|5gCu2jpL!0A}7`CG(W=_RAe)5if9~!Df^3q@xE(Tm6#B;P&keb zG0S~dhz3gAnprk603N~_3kepS8u;X$wX?B37l!B6zLt40O{4UNabpv4a?Q*~(tzSn&lBf$F1$?9ih zzW3Y;yqvY%zSS(OGJj`>z|-_fge5Jf8>ftt1-Q1eH!j%{Y# zQqTb4JJtoAdK(9`BvxiE%Ap3XR^3OtHjMqudOd=B4J$KjhF%2IvA24wB!}HhKYPPN zx|7n<(!AxY4|&mIZC|Sl7mGnOlY}16r;4}=XP)hIL!$rUbeNrdcW6P&Q z7@9oyo!QLa;MyY@W#jbqW9^5T&OmTZ z64?E`(OPA$$W|}!N~Ix+S!T}DJKX)~D$XooiDfm6EBK0!yBMS@L9!G*nl5gFh^D4M z)WC7k#l;|vHH3t$qI^X%1VxzLg`~1~_fLkaN*y6wfQVo5fp*84La8fO+sf|{; zGPJs`B^jHeoX!--w$DYBh4;s%@%4=oNk3oGYboe4NsHcMver#WRpPttyVy+A0E^@G zELk8RgM!oK`X=DE6^^%TcxDH|QpI7C)u5R#*xtF*wD8lQLI*-Lx{|lg$B1X%jqXZ?p+9eE&Lg=AZj+(O%E>uck zGEWQ zs?rv|TsdUVMv>`7O{4oci-Y$xgYr5_Cdei6E|bK`RXE#(ojBF%g0#+>3^A3Z&;nD( zETClX&Z6V~Er_%igNY7=c^bPnH(5~)8Zvb-OhfW;EPG>Ss}Qs?VW$e#?~ZzGiv1%lGn8Rj)(mih*K(q zt~Oj}&RZNVA7i|Vn&G>ENg5ja% z42U|>>fI33u3(huC>W>^8_GQmNjl`__QTCu{1* zMszu}!fD4AbRu71x^Q*eH<}7~602W>54hJl6@Nr52L#W4v3xzP1%V*)`fDLwoP3=65YNj=w_qcmy|@(%aZERi|FBs6K~s6Wu>4=V>Wzy2EHn_Gu6 zG{`_tew@J_6SrcI#^loFGO)GSpI?q$wpQ}ti6}i&Cn^GINN_LExnlbRbEY@fWnYnyNe#RqFrB#m(?Wgv-7SJ5wOMMinKWD43j zypIa@Nn?UXdoT31eMPwDpY+$#Ep+>^kGRN1WF;Fy%BsrH*q7BkA!b_ocC_<|7 z2|HXuWx>tP!3VtPQuSCxvIyorelgCY@%Uo*C)pZU*F{%WnWQ~|EF4{LQd9D9VR_QjKYRoDn|y)db4HKeM~Ro2O3p)E1&ytAyt?Lc&&1T~)+ zN-+$TgQY|fATJ6&SUc5d6V>PpiOxo|P+1h(RI5&3A%)7Mq@;$evZAG4iWWBkht{Ym zMP<07CAIQ^!pXo&lwjIrNe+kghgE@;zbTi8XWe?_7=?VJFX{BKw(&F#O3DfDnyWqM zVht5c#oK#Sn%b4gzX8rzR1`#YtlUeZri&B#wDU#-uQ>$%1Gmkw= zOY_9Pum5F(ofJu3EL%mj$ImbxNL44Y#zzz6ESNkQ)&v=fSxcC;lpFGFk1|KWejBYa zTL(HcFvEx#QT1hQe+`5}hQ*^Ye2%OM6~i{(qrUH%xfPXFbWMC0z)P>Yh{g0Tq#vG0 zZhK&|V0o{E4rX+dnkZ@lfFwcPq9SOxCf%Ys|3NjrrE;dfLTjFI;sC7VDCzp;ySlsG z&bmWfu`ya?`twdQbfuBx+Lfx-u6%miT#YbzI1k|C)YzUgcQ;&PaYPO!D@4w6wA(v4 zL&FC(X~~T*gC#%A9@XfsV@*&Pi0{dL*zb09qpH#DMU3Rg3SMV0)zH}9xScX_)n!e{ zO*F?h_&PNxs;IlE&%{8gyK~0=QyZI}nG9S$Yp0HS3He_9U<+uevABK&s~N{@ebk=^ z4pCH@I-azSE^rKDr+4Wkkp7y!{R4#_B0u4%xe)ot_r%M`Z8URIs{_MxWr>+>7}aDc z8aDr6qLTRpBkcoBhE_J>S-fQCYK*c}_KRFvV+q}CzfUNAXNaIo+z}U3Y|^=@`8P33 zBAjpkl0(4h`~%6W?~wUZeih$kXtPpQz8HJsC%KC2#4SUWmJ{Vr1^G`{uO0}PE)TI#iC5v#t& ztPMHuTTA9`_Ij)kP&F%*cVt257{SoG5ySEZiiHWsSydzH(d;OkcIiJ%~%Y}x9hMpPIresli^Ktvisw)@;rQ0fRLQh6VHG9XYX3^># zPFcz>Npy*AWV9RXY{0$X6)2YEDy)ip-G4_zmY0`aI+Ia)DAV^d$qz1r9~1i;eBPp0 z)r}0KyKrS|7?|}NB^+T+scDjLUxi)vS1ufvsSy)%)Wp|}$88Yhd@tL!M-h|o^U?3n zPp)18Z;J56g$w_x79PRZQC2E4_F z&H#z9ngb&Hc&geuy~^UuxiI(7T5wcBpan`&RlA!^?e)v~OLY6sjDS4v?75QciDA&s zR*$URGT@c&k-q_R%JvI3z%zwn%ocqYcA;)vfgZgrB=MOIfqD4#7k`~>MOx<*C_5j) zim77EW>O?NrEv3`etWgPDPH9Ab+CGFK;FBEMfJ^zan`0bRHzl%>DEh|mzUi zP63QfViKPS{!#MtrLX1+j5>FmUh=J}OvZFUYjcvC0rZeNOPg%rCOPR@zyu~HnhLvI z{1HOFI1YW|i;8HBPRDMW*HCHN!@*aYVfG<+THyGSTXV&FAbPTw+SP+Psu=4m8tdFB zZqPSb)9Blfp|qYt&AAvU-KRg3v-I7gZ}5wk40}$~Og`A8IXzFbR5Z*BsP+p|6l-nMnUq zf6rF#zQt+dA?S5>B0?#<&LCBSsEU4w6ia2wZWVXHu4Np+F-G6`@V>TFmTuUVIeIwo z)e%0gw#kzphAq_J@MFq3iELe((5cO*NKh*_+ps*tCbGpTFJ_IFq?xnFsVs&?t$)(X z$fc=`^Oj?4HmRNT@WVu!B`$%>)v72B_jCZs2i#9tOI{x@U0f6N*C$J>B5n>a#lE;b znn0w@mC*|~26m!9$sP3dE>x?VMH_5eZK_nAzD^eO%(2QK@AMjGkG`+eFVx1xkI~(9 zYFVEm)u*?7;3v)~e;TnZT~!^d6NocNK!+nEB}oadv>k3?+mnX!a8s#wB;7u{Fs6I( z6bobGH@cN*bwuN@^c_Uw+En<(vvOi4{HbRRenajAGjn2BCo{lwA zd5SX`gQuUuzo4QjZF*TGa1SAB>D0WWEJ*748sTi=X2I?xppXdM_eokm^XVCuRkqYn!++95Yk(fqxa1 zcSnu_Jw`N70S)z=4hpPtN@=v1+KRn_lQl39Q7C~KY+(D@iHSgJ741s=Fyf#k#+kT4 zuk{0OrB`=*Z^3l}r?+{}0~mw-z+-3_nK{b%IkLNT`i#$8pP`_b(au$V0LN}T`veI)ej+Bn>kiQJ?h5O_MCIniJHi=Pd+EffDISNN{ zVw8PvTgPCB7dIwdQhyA{i*FU2JW@9K6bjqyrzpg(+^j+cpK}MJ+{y7^4P)6Oegha< zsH0I+A*h^8h}=hsz{w|`2I9Mo(5bBQiz@>lK~^3O!-V>1x@B%8_-jgPXGZn6&$bhW zV5KKHPmj}wKFGGl!h|!_P`Gx`o1HEbed}&q1a`@iyI;PzU)`tEq z{eaXBa3`~jd$;xv44Y2DlEN&kaWEe`sZzGQUig zp4MLCo3M7(Gor-_!5y=^aua zG;%MkRC79&@Yga9PQTrt^nR2|SKCmVH*U9?Dd&|l>G!pRFRRh~;;i%ls9Apsninq9 z8)h5{Pye-G3lTdXQCwV-9@>27R9B*L`^WiWNeIuc@t?xSy0zj3nK3MeOdP zq;!PF6e&CCHwehv<6dmx1v%J%RK zzmofL>RS@J@q2CQnCJT)?K;iAbuu_Ye{@j0bt=rw;V%Z)*w^#pS6jOM$Lj``3=bzK z)COU2COJHQ?ctv3^rl>lzn|1;nw7hLhD%fDDcUTDE{0=_!m*ku1t$=3Vh1xmd}=>W zeua83GV|;lFKtr89K~2IZPMmneu=!P5=~>W(!7uAaS7@Zqy<7$bz_jL`*X*!HZdk^ zBro<6FIs_7&wPLB81#zEE9 zBkDd|Me=LsO(hrC?U&N6TalLYV@E|)sH*{|X0GXx|OjgZkdIjhU*3T&^42S<> zKU(u;&zC-E8TB#2#I%EDcD6v*c#{tgty(_3AhzTEhb)i(WwA`zOZ@0FTCGT?q7zJ3 z1zCbW>>fpukE{nL_>^2%-^Je@+Gn}Xyg}y}Q{uk?GlTnuzr1yqtrsVAYL<(OOCyWU zm^4~`;()9}XEh}>HMfoQj$~3=>#trG1jEC|E#t;5XCS+T^I8|K5dLhs@3P_(wX-ow z_XaV%Sq^qoJFW(uh~>{j@*g2W2g$gu){S#n9d}L=EVvX}tp>dYX7#f3)_Qsh*dG6M z&&7&mD_2iQMfvvW#!jrpc12FWOpx6+PSKG5r8HDuD5UCt)NiSe{eoeFJ?2K@ak11w zgpJ0F%JD?s*w?KLT$upx`Wd1rFw?B@bYhvYK20aUrF##uPV{0GGo&bJ+7clqRm(#V z(zGn&C>8vDWo1>x#KczFQmrXHjDJ45?bMJ5wedi7!YY%sD4{DYZs}+7j{o1bz`FG& z;rQ%L)NZa7Gd;4FStZl#?GE-z*&pl~$p7USA|%N28lX$aryJ^iz>g!73}G~E;fgUr zMdG5l3uHS*52>jlBH*|UJ^(J69D;myo*eHn9Q2H9J2z_$;ka~-fQAF<}>%)%$+uSFX;0>vWTNJ7ileVej_$>kh8M@bA$Z6Hf*9_YU%b+UIG&`Jqx~+yWxagD~)*RZHC! zJB#STOf7bVnvIZtQ;M^u!Wu0G%e1@7)Z}14D6cB!`@Zrng}~9#r_!EpoE^wkT$R?D zeOF1NC0a#7>Mqd2t_`kM-ZI~PUARg7+vw@rQQL}DD_>A8mBOag+a2xQOfLaV-S0vTVRV?+r9a<9(A+YZ0zod zx47bqBohpA-eyB`_b5R&Xz0@JC6I9z&nw5S=6ti=7sMD59YA*<1l)|E^ zYN_P=-{`SO#xb%~EQICgYAg5)__!(<_%*b19>qzOefUu_H+VIS#aaKri&#_*M0LEW zo2RkSJ=S4FiIVU|)a7T>ds^@ema|T$ZP`@n@qlEK0SyLGvlrFKWGpSrE>P$TXd3GN zn1-KWuzB>1OIP$L60V*KzW$Wx+g%vhTs&42Y^jwG3kwSmjtLG^oCMWYJe{6?syt2Q zCJ930Jyk10QuT(eOKMPj1Z&oM&xT2C?W2bCo*u(c89dtO5ij|uwmZ_d3-p6$ASNiy zJU2^MhDfLA8oYEx_7t;3RJOP=!%Ciix&m3-EJrkE|+G#d8KzVlih z!8zc+lt_*Bz84MAv{y%uLl$B@6H+R`5m})R>l36v>w$SAD~=HpyAs;fUSz}DdSwFz zQOKYIz`?-5px}_8prN4tL{WfZC;(`Xs2@m~l%UW_$b^&~YigN=RaA{kocw|D3Ar=m z0d+mS7%YM!ipGY{dHJ&xpva2A?wx;zR=^2>LMzrwK2TwXiH##_+8SGe^E*lb7Bbod zuBZj4*o`K2j%rx1Ub)?U38hf_Fu8~0X_pUZ(n}AW{FaBrE7@||IBKjs0IMce0}@nt zxAMSQwy2ciWEf*-(`M>DEa3~60IFZ~95#!ICTDFFte zSd7F3P#)BM+Z^3Mg=zwC@A zoR*HfGb)J2GQ>`iDC9;m*zWKWTRX$_qDPxpv#+mFq8xLq(PEs{z30Z?=*BG$jSDki z$dwUioyFX7IZeW*F=0k4l*^rX*LJX`@qQwhCc8M|DAdt_K@{M{YY6XOl0pV}0;&+L zQ#SH}fMih=7N=wZ*)R-a#X7JlJ+nkscNdg2S1j&Y68aQvmeES+0IKC7vgTtIzw6`w z**DA}=d5d3&!u|1`CZx-vqKM_T?|{kEhcLJ(D2OZeMw1;rpdU?KGAXUqWiO!padp~ z`<_n8Kl)BcX#vHH9aS;C{00o|@iPtjmy2^vXx%EX+|@DV5aaAhEt11@=*AEGQ%ph8 z*RD5&XD*Y|XD!b4Nj>&Yw5UQl2A}eHd7qji5K)umYp7Icf+LZLW+LDW7Sf8JBm{K1 zYlN_k+s%xO>t^}gX{5aQ?e2h}3NDfp9F#v7rM_sTOoQfAaEcbVS@l*=hY2p9uL+?t zFT;wsGRV$VKv-(+v?lPYU`ASBqgDbBNyIp!@%7B~`z&wLcC<5oy^}Ncr)*Gko`AvA zJtOaPFPSTYT_amI>#cwc3m({Y3~MGR5^Lc%pe<-~{nR4Hd`yj>I;3~^1~W+Xlg_aj ze_gz{qK6JwK&}6+wz+8-YT+<_!Is>&a(R|5g{Q?FzzE$~`{BW*De}eY1@dUJ*;XA( zQ!k36k2-{%?q79&17d>ib0l0@CY+zL zM0XOwQ}nxd4FcKCn)Y2_fuGEUaa2Piu=t<7*FV+U#gnyW<^~Uelaj`=gvnK9^jJI1 z?PM1kVbLEe9(kW0`eANIgG;dktSPG#yh8rm`KCNLXuwFpe)}4y&o6V(QmD+CImg7^ z?zOIIhgAm1Q;|Ts$l?5q1r{lCYqOD-<|4f^bhI1H;571DQQ9S7n zWOV+pcmC55d{|*>ED46R5wLTVaiTFJJ(GtN8!1JQ4|4Hf2mB^3_}hbNYCh0ILTQx> zvVV|&MnL&lppZQ^-P=!yxnT?Oxy!M>3Sz>SZ0|s9(J$g^u-yrHc?_>L)2bWQl7$R- zOT<*7qe)KM5+MMS~DH0>qq-#58iSq)$VT0<1JJ3oILJGUk6G-TvIX$Y}#u^*cj_p6g@h z3>UN)W8Djr9N!ZzXUL(hu_>)FZYckz$+UbQ>bDAUU~j0KAeR|QYnXFw0#kmaiB|6i z+l^lAung|%w`x6D(J^#Xh>uG&&`32aN1w$((`Te&B4=?8;c1t!e#N`A!ys>}XZ{4* zu|p*1F<`hildwdhi%x#*8jdTkwNhz;f)FxH%Xt~wMtiXChlQo$Zrm&3l!Y#~#RCeX z;r%_M3YXY9H}ko>j;=<)Gd7x-<2Az)c|0P~u$2y5`SCT)_4&$Bs~VVn<@Y&knz6zT zIav|Fs!h;J}&G0CL)QZXo?dr-blZjn|_+XO`?)7Gkimcl1#1=tECGsZ3 zSJTlq2;n`0hW-y2XVLDo7UuY8hE~{%^^zp(XyWG-B0RHr_=hUI@x z6B`EN6YHqdu`hD4@d6v?jf?gT>`{Us&aqFGX$MbTjf$gEUS*{^NGGJUB&yXn3JXF116R&+>Fo!XG5SlB3*K$H;o;yCrczp}dT#I=x<=5d zkhs})QhVXQ0Tx1;_?YY@HxU~)=0l#zvf`9drf6j?X4bZ!6fvXSzXM4>0v^<9%#k_| zp`C>4ZGr0E8dR6)v*j+SMeQOfSv?83tm=E}B%T9iCEVDktSFt#(T^W`Pf-Qfc>`)< z@d%aroiUVJ%5`frPt-}gXj!pH=ynBv?S?D@zk9{bdIu_Qnw(ncP<~|Td^xhB_)HSO z1zX`Hj^Ke~Vf6iPTP=Z8gxAUJq|*e#YQ5#U|@6Qi>SD3NKI<^qrY zrkv!jbN*!y!f@46IFCwAX&CCWni^gI2|$Sq`>jCw^sV4@y+6A+SQzRz-fg8w{Sy4CZd}$eul1g~M}d-iK5(`sJGT(9F#GK8<@B zek1ll;QbW4%f3*-G1X>E>!%+L=d|Y%&+RZBSupwr*3VO^Gy!jeyS;b)kMmv-zOGY| z^~|K^#^A-+6lnut;lyo184^+zGls7LVvt^9qDn*6(K&MmBlyrN+66MenveiK@wA!2 z;N@p1LQgahR)Tm->d+^B7FI=1` z5Q{X)QpzA4gleE`$Xp2$5HGU=buz>k!>ewr2FvZdR>aN6kMdx-!$yAt?xvX#YSu02 zrL|JNl|qeB*2v)sjD~?r$lwjh$2Q6?5tGy>b1#G>Io9KOaT6;M zH@a~TsWLI}bk=cJ7lt{_b{+Nws{C3jw^UTUT_bY{i#6aS0}f(eBg@@3dzgl|pWF=s zr{JuoZNcoOEZ3Gsj)hHhq32yKF(F15bSjApue-Hnx(%o7^Ed6 z9=nZ5xqgK?r2ukD_;Ou{1FoutTE#-2w~$U*$WuTKle9i_?ZIbm6(v@iQ6U&3%u-F@ z+qUPTqcTUyq(i~f=1A*FYjsVz`qg1aqS@=c>Im7nI3mAVs1~F`*7vnLRscW~EMXv_ ze2LHVNg|oQdpp{_zNvMEjX>!v;&5OU7vn>8hu9v4UX0o=8;(`#_su6$fi_p5C$Z2; zzDd}3g}gK;yBJ)vMEeB2feQ~lwbo|)Fj`lKEmYNO$=d$HQT(XuyoLdq=3hVVT+sBP zx$&fO=}=1U$3 z>Jl*TdS4V516U`b7C1k{6AEF~uZnR|4U#m#&8E=6J zmBte;Uak$*Th4Krs^07N@0DBDU*RSD^ArP2wkFvSI7$@=j! zjeG3B$E@m>%o%Aq6vVJ(GHg)BVo}vi72I{}eQTM9I(50qB@@h#;cNK!Bvs0UI%}dz zS`RWw;JH6ehId~5Yye`^5&DZwSSJfFE^|TA?}lAYO{W5Ax&u(C*!=3^f_YSrRnGEF zHR@iF{$uf{ryb+m21K)tiW4?4L&BF#NXYUwS4p%bL!Cz1l9t8ATLLDCb5;ke7^xq? z`mQx!icwaM?|ON`bY^FOqkHg|cFQ^;ctbibSyPFj_el5$!wlLk#5_XCV2@v#z-uA0cbeECH`KM3khs zI39wNm|HS9{A+uQ%S~sMdUEIxr?_kc_->2s^1eMI2Lk%~rg}hqT<#n9QCEV@yBqXG zRCxxwl}hJqWAxrVG+(@{b|+0cNJrP5&TKP zpQ~e;E_&)JiubHXP?o!-t)5M94Z>QufUem)?Mzh!fsLGm-N9onU`3Sq!Ct?{48}4_ ze(h4iAWxFR?hO_ke!|#n5wm%9bhCRMs4{7FFzUF8~hoomk_f<2iXvy>`{vEIM3<=QPoWH1udCx$s44T9!HVE~tn&LZo&YhOf= zne-Vg0W5@Pk49mIp+o%dLsXQns?}b9<>=K00O$Z!5QNT4RgC-$(<8gF%lZ({oda0}fWiAPQJgjT7b&LXTgF1b2GFHccT)A_hk zuY=!!AAc=Pd6+^S52Gi!fie8m$Wp?fO7Y&+?3ew_>>bD70B#bR@!Xj1b!B&)0-Ug6 zDCEd^#(X&n7HMs2&^peg5n#wPhp;eg64 zS*WEz#pcwyV%b8p1PMQ1FtaET_(~{_`6h z@EU6e;31#)%mtzeF?#f<6-8^>dJi#Ur6HQlac(hLV9w%Xj~MzdR@)2%^`|W%pe@~@ zkDq?7@J&TQnC+Kb6Inkqx7N;%^VmHrRUq;YVH~_nGNCQ5>b$YgfL|-~S}~0!T^3se zWCct1Ss=m9eORTh@2!`cFhpG44r883L~Es-5zL$TM;R7MK>~Nh7VuW*#=e2UzK=Gp zVIo~phcMD_P9co`@{yVU39q z;i2YVGHWB&TtnKKeL=3q?mWDoa!&a#9?r{}s(vCEVNXgpGV(db1x#IoT?=Lp%^%3pNdg!P12YBLs2M{(3A{GGm?o0vAYfGFYKi9z zhLN0*)xf8*Qa0K7%U`nxN3w-c1u8cG(8`0?+E|K;;LZ1cZ61NJwt`r+*hq<&pSAit zV>G9P4Yeg$PQm~4?{{In(_6U@${nlCwPa-njlK9W4Q-VO!cy9+*vNQf`ELN`b}SsC zrrnD=-q6E&!9)yOHF4^2pkLE!zd-6fk+yRAkA5^`*b-di1q#*5Ju5n?94X3fB-2@K zYoA1k_AXivOO||~jl=GfG$yuUd(E_($&G2XOqc4t=^BU7I1N1nx=ikUhVS!PmWNlm$E>@kyF zJJ77Zh?_IT(s@{NB?+!ic4fkTzhx06ZeMtOa5P6ZK@d)umO0SN9AWiuv|@rn5Scb= zsVz&*DG$4FZFeU*j1u z-|}j1iwZE`lUmW6mrK!4HPyCw$sMy-ogyQ8FXg?CazvOvnZ`9l^ANNNn{1V%Xs-C+f8FW-o}cm7{}Tz5EIZ~G5w6tQDP zTdN`{Z3q=X5h7Z9kG9k%_}UaTT1to!)F?`5Z4#|&(P{6{G*p)@Mu*u}OW)}C`~KeF z_5ShxbIx_0^PK0oujhX5^||kJuQa|Qd|bq(EeF>O2WpSI>31b2GaLNs*2Ftk!0;Ir zPxCWNf8jtnt;sf5lw3AS_C{^8`)yP}>q-imO&q$NgQC6Ng2mV{Ff}yRgqy!gITSzA z&$2m_j(d^%?%TM;SoXMc1!B*nal3IM|7H#|O?0^HmiwJeiaY>7Vo6NKoT^!MqZk$U znxDvNm}@h#1A~}p;Mi;z+nRq_n(bFf<3*Ye;@* zPJ*~Z0-$E;drESGFiTQfCQ81m-;c4*Q3IHeMog0Y8Qg_;YvAKhAcOx<7cl^^-0O1L9zX7kmMkTrD zWar#&G_B`CbQaOPMo1%8AE{0gCK@$+m+7#Fd0Ojn>;~~-?NP0UJ;QbNstJG)0H|HG zNG4#HzIh+|02?m&a(Re)_8NrUHQs)sD8{Js%WT_@@Z9r)cvHFn2Ea6E3Xn}}@Oi_s zG(cCs{>wYl{4IPr9{9-R$JG?mu`ZLvg^o^>VeDJeKr`N%lmB&L{56zpYv~H00uGwz zVY~R~k6l5lYy;w|X-HggCT=`}!X-xZu~7rGyNY$WiW{BOXHt76!4&=xrM(}x^s_wozve0v7D${H}) z4_o#z28~RJp1_0;(;oJn!I@AUJgdKtD(TVBq&9QG>TF}17~wXPh)v_SGrQU=!b9G4vpBu-#>m`%OcL?ZfEZ$m6fVOX#3zLjz`(4ZmcIeLcgOOJ zUih<8yZt11VxJg5bS{n03>7J~0BFF*1gd{n{B+g<95)8V`OpR{ZsV%tIa!Z%p$E#f z_J-_q(CBPMD08<5cvl|Jf4w5M!C%zPy4G#Zd9ta^S|qMxi>5YM=`^hRVrT8{hv0KA z6$tRSW~BaL%OjSnSr$6@z=ZuGsmK_jax==}gY-hkA%N6g*kaWD(62{rWiU7R>D5)w zH1`wOrm$XLlx5ZuiS=#yG3GXFq4+%#s2%ob(b4QX?-OI`9;Vn)G)sQSAh|bby_d)j zNe(VqF82eq7nUNE^ed+n!_R1QFZPj}Eo-gfuO)R~J#Qu^%G>~a!A91a5?C@;1vQtb z9X}0k*ZqY44G@oAxehbNvSmkb6PCr%1eU5=E$_d;zlyVjA1X6;T0`0VS62sH>6 z@}<1~=eq(OUu$Fur5$WSx6o8DM4daY3I?#ERMA^i;`^XRYB#UmAn%pwHfvC2_F#7m z(CAU>pS;8AFa*zQEsp*-E$Ebd%G!Me!oS%=F7LU?&({MAC@mdfCOWt;L^72e=(m}e zt%t~qA-yh*H*0+l$VsY8a!exuY~Iq2sWDqRJS)fmx~FULfly}%n4hf~AHCyF>en?~ zGm{sAD^&R?4^JnfWXv!w7sv8|`fb*N)j1vt+Ud@uD_4HM zL9J8Ib@&c?$a>18+~MW%&TK=L*1U+0co6JO|3yAg@DYR0IYUXLc4!UfUNU>J!kw0Kbz zMxt?41ni#f#s;A5WH(xJ5#lYpyA3ygy~QvB5f@pw3+tAf$8>f9dT}ntTG_v0Pz>5-PPI z=v#4GRnADh9eJbs9n}D`)QY=cc=Ip&829m}yn3i)l9Ha#Z-D2iVT~`Bz4a<0eA3J@ zQC(6rVbAr421L>H6PV58vou?J9J|!`&19QB^Imw9qb-67L3gwPusv5W-*a&4ovsRQ zpvdOBvYHjTPUmHO+rjnqDKfol7yS)TuVExEy{sE7<Ehkz4x&X%n+p>-_a~`@7-4);ew(>aD3MfDfMhzKj}wI1p|>aSLK!Qfw!u65*0y^;U9Jd4AQ*lQ zSE;*e9%WPeg>&S?0<#5Mx${LZ6D(!h85a`W+~p+V?OItikuE>4q1dJ2O1aZk&D z*~n}c1NhdQ+Xwkny!9;Qoian)CeN!-6%V10M_n&j&IdZLI;FYe4U=j}{h5&+OQ4V{ zZ#z0--O)y4*ePKojd{DwnvpY>{&WS<;9@l#9gz!4j;m0ML-t*<0AK3cDMR**Y3Lc9nZ1cCaW8i67Mcq3g;D^PEgC1d z_ZZ~B4%sLU<`>m@C;{|>9Pi3~UTy2=A|Di;#jzrWoSbcp8opRj@Har=u`%$|>(D{r zx}##z|0+$aM1ltJxdPqL`Mzg87q_up-TiZZ#c)tWrB6cxei-QZV=XUR)eOF{)l=%8 z0w@4~q~Le9AvON+tf=1pr(;VB=!og=+=9^q8IP}HlP=i<1q_(7IMn0rUXXPn}Z z{Sz?zy~<#!+=4>CaxH&CekzGW4eNZg>?u0fRl8a`u1Z#G+q`Nr99-XQavK*QF#9~? zH=wjUd_e+>y8Co>!!YuD)#Y4?>m6lA+9kx9SkE49|SX-s}A*a1F|_2HTY(MSk)^F%we72<;-fN=*lf#nn*h)*GDn6_p#ppuVc;bipl8>I~ z<5CQK=fpI0yEzi<Y*O)gOfJr|*BdjJ%GbR31V$PUF{kvi}{iE=s0Kaem14 z9hkAMYejOG&0pTJcoG;QlmntJtKSoO>s(y673Uvn{UN9^&@X0Wr*XtQ`sw5U#fgEj zf(KE^f85@X#~eaP+9Je?^0>ihlc#ir^&>5uvkTdW%{1?KuoD)2w#FCCrxNuB`DISs zoO?ff#$#x1fEIQ9Y-!5jME+K0LtoMl98WVE2Q)>G1}Lk+|3gOW^W zK(3>ZXmcXW=~+H9Fh;k`Y*opC8JWQ}H21mkZm-WkjjZn{*~zrADNSVG+8?%L=XH;q zm}*X1WUb{K0E0Y{H!S6jD&1Gf4KlvR-N0l8=L?yck68Yu@<|w;Q6@u9_zO`xTY_aj z^8|0H79w?Dk&FL={&I3&fDr<{t?DXYG8rOBxyYrzIu^Mta^7yqz_|z*=mCG6P1Tbu z*5-t>nX;uF0VgT8ht2Q#SJVc{i4s0IyuF0-?z*&U-=4XDrOiaD()yw~Y5e07f+b%{ zxa59^uoq%L6UiBF_2GhQP-< zrl|Q{5tZ9*My}28@5#x<8#)N0_zYi&Nt{jFT6C+|wP1mi{xTYvz0=gmT@>qx7-Q)Z zB)!qx+{JSZ+`e<7s`%UvgiAMt1T6+{B#_W3iYHoAmE0{{jmLjfZ;xBB{8bDR!{p&4F;0&w^}1TUx?NkFtL?$L-Qv) zq;Nl$SFGTHuK`y-Hlj`d2AMWGbylJVu2C8Md#+#Y*_4=F_Z2!T#7H4Wn?q z+Ry=(a`D~~dGY+h{9TJfhu|cc=OLcw5%r+*&|RC%U?gv1JrSAaKsm#oNc6PS=Bit+ z7($*}FT-Tjk&{O2F5CV~^S^nh;6n5oWmogV7Jjo4&u1K1)x<_hUIne*gT3lhop4C% z=o{j$t%&SGk8$=;dMeU&!`KV3z|fDE*e*O(D)mmBPn*IXo(fsRoYAn2rdDJV%>5od zx*{{Eg}l`sQX5))_1plzMY}6Lg%G_gKQ||RWDiq*s1X}JhDTLOejXjGc9S*P$Fb~i3Ru3AUDjDAzBX-VrFRcf#UER%sS7- zWE9fxu_C6*7)?};D;v`!v2{#G+D--D_&_I`^z!635CSnX=}Ah4?n{7D7BxE^tMT0O z8uAebTrXOTCUl!jx4cIhy#kYuc3FB8ZS3^?<@)>YVhwTr4!oB6pC%m~=GZU!C75G~ zJR6*qS5dbrA|p8~ecj!h^7S(lvza-_Y(=D|=%Ao`tKuZGIFCF zm4>yC6$KM2F}K4xImGIO$yT|0b!VS+7hAOheIxaPt<<;&CgA`CL#IFXvA!0${P)+FBx8va09 z6ECB$d9BEmejV*yIEjk^2Ii`7!Hh*>n19>=`wwIliiFTcBwEu6MjS)M&zY)~m5uu| zPUf-H9g^%>RLLbzSY@Je{Q9%t{}Hli)HN9Xz$tCBG4O_xdVBurlfWCC>cso|P#$T^ z+PhSQNTQ;Vd+eeopj5isj&AxH_xTFm;@Cs+O)W_QcDswNxc5h&jHB{1jzJt{-<2IP zf@VH;rG7uAN}U*NHvNa~70brkh79f&@_;1bD*S>vq8yBAHiizbJNw z^}fo^jQbKBMy#JxYyN7l$*1Gtc{PA#s7sr;fg;|gEAP+BeS@oO#TzZ@aHFOSCEEna zSC|y{{!Wiw@%J2}0iK?FG#ys3jcS~{x}4<@B!o97zWUS_ATuy?r$uInGhL=R;<|U9 z-I>It!m=Hf0-0|uTPbCJUPXEbC>Jtp=uCEmxko0)qb@L3E#Pq9+wfcF_xJ?|ZU zng&m~B{#KdFDqd3j%wAnh>l9Xw95h0n5aA|&ZUs`OO^?A file.length > 0 && !/\.test\.tsx?$/.test(file)) + +const offenders: string[] = [] +for (const file of files) { + const source = Bun.file(path.join(ROOT, file)) + if (!(await source.exists())) continue + const lines = (await source.text()).split('\n') + lines.forEach((line, index) => { + if (APEX_LINK.test(line)) offenders.push(`${file}:${index + 1}`) + }) +} + +if (offenders.length > 0) { + console.error( + `Site-URL audit failed: ${offenders.length} link(s) target the apex https://sim.ai, which 301s to https://www.sim.ai.\n\n` + + offenders.map((offender) => ` ${offender}`).join('\n') + + '\n\n In code, build the URL from SITE_URL (apps/sim) or SIM_SITE_URL (apps/docs).' + + '\n In MDX, link with a relative /path or https://www.sim.ai.' + ) + process.exit(1) +} + +console.log(`Site-URL audit passed (${files.length} files, no apex sim.ai links).`) From d6f38a01e4dd592f87a20ab0c786aa4d953e8c0b Mon Sep 17 00:00:00 2001 From: Waleed Date: Thu, 1 Oct 2026 01:07:48 -0700 Subject: [PATCH 2/9] fix(chat): keep mention searches open and align resource menus (#8507) * fix(chat): keep mention searches open and align resource menus * fix(chat): preserve mention selection and nested menu sizing * fix(chat): reset replaced queries and bound workspace previews --- .../add-resource-dropdown.tsx | 22 ++- .../plus-menu-dropdown/plus-menu-dropdown.tsx | 162 ++++++++++++------ .../prompt-editor/use-prompt-editor.test.tsx | 148 ++++++++++++++++ .../prompt-editor/use-prompt-editor.ts | 56 ++++-- 4 files changed, 312 insertions(+), 76 deletions(-) diff --git a/apps/sim/app/workspace/[workspaceId]/home/components/mothership-view/components/add-resource-dropdown/add-resource-dropdown.tsx b/apps/sim/app/workspace/[workspaceId]/home/components/mothership-view/components/add-resource-dropdown/add-resource-dropdown.tsx index 73865f1b3ec..1773f242283 100644 --- a/apps/sim/app/workspace/[workspaceId]/home/components/mothership-view/components/add-resource-dropdown/add-resource-dropdown.tsx +++ b/apps/sim/app/workspace/[workspaceId]/home/components/mothership-view/components/add-resource-dropdown/add-resource-dropdown.tsx @@ -82,6 +82,7 @@ interface ResourceFolderTreeItemsProps { */ folderType?: MothershipResourceType onSelect: (resource: MothershipResource) => void + subContentClassName?: string } /** Renders a {@link buildResourceFolderTree} result as nested dropdown submenus. */ @@ -90,6 +91,7 @@ export function ResourceFolderTreeItems({ type, folderType, onSelect, + subContentClassName, }: ResourceFolderTreeItemsProps) { const config = getResourceConfig(type) return ( @@ -108,7 +110,7 @@ export function ResourceFolderTreeItems({ - + {folderType && ( onSelect({ type: folderType, id: node.id, title: node.name })} @@ -122,6 +124,7 @@ export function ResourceFolderTreeItems({ type={type} folderType={folderType} onSelect={onSelect} + subContentClassName={subContentClassName} /> @@ -275,6 +278,7 @@ export function ResourceMenuSections({ type={section.type} folderType={section.folderType} onSelect={onSelect} + subContentClassName={subContentClassName} /> ) : ( items.map((item) => ( @@ -401,6 +405,7 @@ interface WorkspaceResourceMenuContentProps { /** Offers every folder as an attachable entry, as chat does. */ selectFolders?: boolean onSelect: (resource: MothershipResource) => void + subContentClassName?: string } function WorkspaceResourceMenuContent({ @@ -410,6 +415,7 @@ function WorkspaceResourceMenuContent({ searchable = true, selectFolders, onSelect, + subContentClassName, }: WorkspaceResourceMenuContentProps) { const { groups, structureFolders, isHydrating } = useAvailableResources(workspaceId, { enabled, @@ -425,7 +431,12 @@ function WorkspaceResourceMenuContent({ /** Lists fill in as they load, so a trailing row keeps a loading workspace from reading as empty. */ const menu = ( <> - + {isHydrating && Loading resources} ) @@ -449,6 +460,7 @@ interface WorkspaceResourceSubmenuProps { * offers its folder, for pickers that can attach a whole workspace. */ onSelectWorkspace?: (workspace: Pick) => void + subContentClassName?: string } /** @@ -461,6 +473,7 @@ export function WorkspaceResourceSubmenu({ selectFolders, onSelect, onSelectWorkspace, + subContentClassName, }: WorkspaceResourceSubmenuProps) { const [open, setOpen] = useState(false) const icon = ( @@ -472,7 +485,9 @@ export function WorkspaceResourceSubmenu({ {icon} - + {onSelectWorkspace && ( onSelectWorkspace(workspace)}> {icon} @@ -486,6 +501,7 @@ export function WorkspaceResourceSubmenu({ searchable={false} selectFolders={selectFolders} onSelect={onSelect} + subContentClassName={subContentClassName} /> diff --git a/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/plus-menu-dropdown/plus-menu-dropdown.tsx b/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/plus-menu-dropdown/plus-menu-dropdown.tsx index 6eb493cc479..3759f6e3d81 100644 --- a/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/plus-menu-dropdown/plus-menu-dropdown.tsx +++ b/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/plus-menu-dropdown/plus-menu-dropdown.tsx @@ -5,11 +5,15 @@ import { cn, DropdownMenu, DropdownMenuContent, + DropdownMenuItemLabel, DropdownMenuLabel, DropdownMenuSearchInput, DropdownMenuTrigger, dropdownMenuRowClass, + OverflowText, } from '@sim/emcn' +import { IdentityTile } from '@/components/identity-tile/identity-tile' +import { getWorkspaceInitial } from '@/lib/workspaces/initials' import { ResourceMenuSections, resourceFromItem, @@ -30,6 +34,7 @@ import { import type { PlusMenuHandle } from '@/app/workspace/[workspaceId]/home/components/user-input/components/constants' import { buildMentionPreview, + type ResourceMentionCandidate, resourceMentionMatches, withBrowserTabMentions, withFolderMentions, @@ -39,7 +44,7 @@ import type { MothershipResource, MothershipResourceType, } from '@/app/workspace/[workspaceId]/home/types' -import { useOrderedWorkspacesQuery } from '@/hooks/queries/workspace' +import { useOrderedWorkspacesQuery, type Workspace } from '@/hooks/queries/workspace' import { useSettledTerminalCommands } from '@/hooks/use-settled-terminal-commands' import { useBrowserSessionStore } from '@/stores/browser-session/store' import { useCopilotTerminalStore } from '@/stores/copilot-terminal/store' @@ -51,6 +56,15 @@ import { useCopilotTerminalStore } from '@/stores/copilot-terminal/store' * autocomplete. ~10 rows is enough to show several families at once. */ const MENTION_MAX_HEIGHT_CLASS = 'max-h-[min(280px,var(--radix-popper-available-height,280px))]' +const RESOURCE_MENU_WIDTH_CLASS = 'w-[360px] min-w-0 max-w-[min(360px,calc(100vw-16px))]' + +type MentionCandidate = + | ResourceMentionCandidate + | { type: 'workspace'; item: Pick } + +function candidateKey({ type, item }: MentionCandidate): string { + return `${type}:${'workspaceId' in item ? item.workspaceId : ''}:${item.id}` +} /** * Resource types that are only offered via `@`-mention autocomplete and hidden @@ -122,7 +136,7 @@ export const PlusMenuDropdown = React.memo( const [isMention, setIsMention] = useState(false) const [search, setSearch] = useState('') const [anchorPos, setAnchorPos] = useState<{ left: number; top: number } | null>(null) - const [activeIndex, setActiveIndex] = useState(0) + const [activeItem, setActiveItem] = useState<{ query: string; key: string } | null>(null) const searchRef = useRef(null) const contentRef = useRef(null) const browserTabs = useBrowserSessionStore((state) => { @@ -141,11 +155,15 @@ export const PlusMenuDropdown = React.memo( enabled: inventoryEnabled, includeFolderMentions: true, }) - const { data: allWorkspaces = [], isPending: workspacesPending } = useOrderedWorkspacesQuery( + const { data: allWorkspaces, isPending: workspacesPending } = useOrderedWorkspacesQuery( Boolean(organizationId) && inventoryEnabled ) - const workspaces = allWorkspaces.filter( - (workspace) => workspace.organizationId === organizationId + const workspaces = useMemo( + () => + organizationId + ? (allWorkspaces ?? []).filter((workspace) => workspace.organizationId === organizationId) + : [], + [allWorkspaces, organizationId] ) const [inventories, setInventories] = useState>({}) const receiveInventory = useCallback((workspaceId: string, inventory: AvailableResources) => { @@ -170,7 +188,7 @@ export const PlusMenuDropdown = React.memo( setIsMention(!!options?.mention) setOpen(true) setSearch('') - setActiveIndex(0) + setActiveItem(null) }, [] ) @@ -211,20 +229,45 @@ export const PlusMenuDropdown = React.memo( selectFolders: true, }) - const filteredItems = useMemo(() => { - const rawQuery = isMention ? (mentionQuery ?? '') : search - const q = rawQuery.toLowerCase().trim() + const query = isMention ? (mentionQuery ?? '') : search + const filteredItems = useMemo((): MentionCandidate[] | null => { + const q = query.toLowerCase().trim() if (!isMention && !q) return null - if (isMention && !q) { - return buildMentionPreview( - visibleResources, - (type) => getResourceConfig(type).mentionPreviewLimit ?? MENTION_PREVIEW_DEFAULT_LIMIT - ) - } - return visibleResources.flatMap(({ type, items }) => - items.filter((item) => resourceMentionMatches(item, q)).map((item) => ({ type, item })) + const workspaceItems: MentionCandidate[] = ( + q ? workspaces : workspaces.slice(0, MENTION_PREVIEW_DEFAULT_LIMIT) ) - }, [isMention, mentionQuery, search, visibleResources]) + .filter((workspace) => workspace.name.toLowerCase().includes(q)) + .map((item) => ({ type: 'workspace', item })) + const resourceItems = q + ? visibleResources.flatMap(({ type, items }) => + items.filter((item) => resourceMentionMatches(item, q)).map((item) => ({ type, item })) + ) + : buildMentionPreview( + visibleResources, + (type) => getResourceConfig(type).mentionPreviewLimit ?? MENTION_PREVIEW_DEFAULT_LIMIT + ) + return [...workspaceItems, ...resourceItems] + }, [isMention, query, visibleResources, workspaces]) + + const activeIndex = Math.max( + 0, + filteredItems?.findIndex( + (candidate) => activeItem?.query === query && candidateKey(candidate) === activeItem.key + ) ?? -1 + ) + const activeKey = filteredItems?.[activeIndex] ? candidateKey(filteredItems[activeIndex]) : null + if (activeKey !== null && (activeItem?.query !== query || activeItem.key !== activeKey)) { + setActiveItem({ query, key: activeKey }) + } else if (activeItem !== null && activeItem.query !== query) { + setActiveItem(null) + } + + const highlightIndex = (index: number) => { + const candidate = filteredItems?.[index] + if (candidate) setActiveItem({ query, key: candidateKey(candidate) }) + } + const highlightIndexRef = useRef(highlightIndex) + highlightIndexRef.current = highlightIndex const filteredItemsRef = useRef(filteredItems) filteredItemsRef.current = filteredItems @@ -235,16 +278,10 @@ export const PlusMenuDropdown = React.memo( const isHydratingRef = useRef(isHydrating) isHydratingRef.current = isHydrating - // Reset highlight to the top whenever the mention query changes so the user always - // sees the best match selected as they type. - useEffect(() => { - if (isMention) setActiveIndex(0) - }, [isMention, mentionQuery]) - const closeAfterSelect = () => { setOpen(false) setSearch('') - setActiveIndex(0) + setActiveItem(null) } const handleSelect = (resource: MothershipResource) => { @@ -257,8 +294,12 @@ export const PlusMenuDropdown = React.memo( closeAfterSelect() } - const handleSelectRef = useRef(handleSelect) - handleSelectRef.current = handleSelect + const handleCandidateSelect = (candidate: MentionCandidate) => { + if (candidate.type === 'workspace') handleWorkspaceSelect(candidate.item) + else handleSelect(resourceFromItem(candidate.type, candidate.item)) + } + const handleSelectRef = useRef(handleCandidateSelect) + handleSelectRef.current = handleCandidateSelect React.useImperativeHandle( ref, @@ -268,18 +309,14 @@ export const PlusMenuDropdown = React.memo( moveActive: (delta: number) => { const items = filteredItemsRef.current if (!items || items.length === 0) return - setActiveIndex((i) => { - const next = i + delta - if (next < 0) return items.length - 1 - if (next >= items.length) return 0 - return next - }) + const next = activeIndexRef.current + delta + highlightIndexRef.current(next < 0 ? items.length - 1 : next >= items.length ? 0 : next) }, selectActive: () => { const items = filteredItemsRef.current const target = items?.length ? (items[activeIndexRef.current] ?? items[0]) : undefined if (!target) return isHydratingRef.current ? 'hydrating' : 'empty' - handleSelectRef.current(resourceFromItem(target.type, target.item)) + handleSelectRef.current(target) return 'selected' }, }), @@ -311,14 +348,14 @@ export const PlusMenuDropdown = React.memo( if (filteredItems.length === 0) return if (e.key === 'ArrowDown') { e.preventDefault() - setActiveIndex((i) => Math.min(i + 1, filteredItems.length - 1)) + highlightIndex(Math.min(activeIndex + 1, filteredItems.length - 1)) } else if (e.key === 'ArrowUp') { e.preventDefault() - setActiveIndex((i) => Math.max(i - 1, 0)) + highlightIndex(Math.max(activeIndex - 1, 0)) } else if (e.key === 'Enter' || (e.key === 'Tab' && !e.shiftKey)) { e.preventDefault() const target = filteredItems[activeIndex] ?? filteredItems[0] - if (target) handleSelect(resourceFromItem(target.type, target.item)) + if (target) handleCandidateSelect(target) } } @@ -343,7 +380,7 @@ export const PlusMenuDropdown = React.memo( if (!isOpen) { setSearch('') setAnchorPos(null) - setActiveIndex(0) + setActiveItem(null) onClose() } } @@ -392,10 +429,8 @@ export const PlusMenuDropdown = React.memo( collisionPadding={8} className={cn( 'flex flex-col overflow-hidden', - // Plus-click shows short fixed labels (Workflows, Tables, …) — let it size - // to its content via the emcn DropdownMenuContent default max-w. - // Mention mode renders resource names directly, so widen for breathing room. - isMention && `max-w-[min(300px,calc(100vw-32px))] ${MENTION_MAX_HEIGHT_CLASS}` + RESOURCE_MENU_WIDTH_CLASS, + isMention && MENTION_MAX_HEIGHT_CLASS )} onCloseAutoFocus={handleCloseAutoFocus} onOpenAutoFocus={handleOpenAutoFocus} @@ -408,7 +443,6 @@ export const PlusMenuDropdown = React.memo( value={search} onChange={(e) => { setSearch(e.target.value) - setActiveIndex(0) }} onKeyDown={handleSearchKeyDown} /> @@ -426,6 +460,7 @@ export const PlusMenuDropdown = React.memo( selectFolders onSelect={handleSelect} onSelectWorkspace={handleWorkspaceSelect} + subContentClassName={RESOURCE_MENU_WIDTH_CLASS} /> ))}

{/* Plain buttons, not DropdownMenuItem: mount/unmount must not mutate Radix's menu Collection, or FocusScope restores focus to the content root. */} {filteredItems !== null && (filteredItems.length > 0 ? ( - filteredItems.map(({ type, item }, index) => { - const config = getResourceConfig(type) + filteredItems.map((candidate, index) => { + const { type, item } = candidate + const config = type === 'workspace' ? null : getResourceConfig(type) + const workspaceName = 'workspaceName' in item ? item.workspaceName : undefined const isActive = index === activeIndex /* Items arrive grouped by family (one group per type, ordered by RESOURCE_MENU_ORDER), so a type change marks a section boundary. @@ -450,15 +487,17 @@ export const PlusMenuDropdown = React.memo( therefore every keyboard path — indexing exactly what it did. */ const startsSection = index === 0 || filteredItems[index - 1]?.type !== type return ( - - {startsSection && {config.label}} + + {startsSection && ( + {config?.label ?? 'Workspaces'} + )} diff --git a/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/prompt-editor/use-prompt-editor.test.tsx b/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/prompt-editor/use-prompt-editor.test.tsx index b3f1316c96a..97b5d72f905 100644 --- a/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/prompt-editor/use-prompt-editor.test.tsx +++ b/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/prompt-editor/use-prompt-editor.test.tsx @@ -15,6 +15,7 @@ import { type UsePromptEditorProps, usePromptEditor, } from '@/app/workspace/[workspaceId]/home/components/user-input/components/prompt-editor/use-prompt-editor' +import { getIntegrationMatcher } from '@/blocks/integration-matcher' import type { ChatContext } from '@/stores/panel' function selectionPayload(context: ChatContext, sourceWorkspaceId = 'ws-1'): string { @@ -74,6 +75,153 @@ function typeInto(textarea: HTMLTextAreaElement, value: string, caret = value.le textarea.dispatchEvent(new Event('input', { bubbles: true })) } +describe.each([{ workspaceId: 'ws-1' }, { workspaceId: '', organizationId: 'org-1' }])( + 'mention search in $workspaceId $organizationId', + (scope) => { + it.each([false, true])( + 'does not auto-attach an integration after searching (dismissed: %s)', + (dismissed) => { + const matcher = vi.mocked(getIntegrationMatcher) + matcher.mockReturnValue({ + regex: /Slack/gi, + byName: new Map([ + ['slack', { name: 'Slack', blockType: 'slack', icon: () => null, bgColor: '#fff' }], + ]), + }) + const { result, textarea, unmount } = renderPromptEditor(scope) + try { + for (const character of '@Slack') { + act(() => { + typeInto(textarea, textarea.value + character) + result().handleInputChange({ + target: textarea, + } as React.ChangeEvent) + }) + } + if (dismissed) act(() => result().handlePlusMenuClose()) + for (const character of ' roadmap') { + act(() => { + typeInto(textarea, textarea.value + character) + result().handleInputChange({ + target: textarea, + } as React.ChangeEvent) + }) + } + expect(result().getActiveContexts()).toEqual([]) + expect(result().mentionQuery).toBe(dismissed ? null : 'Slack roadmap') + } finally { + unmount() + matcher.mockReset() + } + } + ) + + it('keeps multiword and punctuated names searchable and replaces the whole query', () => { + const { result, textarea, unmount } = renderPromptEditor(scope) + try { + for (const value of [ + 'Find @Quarterly', + 'Find @Quarterly ', + 'Find @Quarterly plan (v2).md', + ]) { + act(() => { + typeInto(textarea, value) + result().handleInputChange({ + target: textarea, + } as React.ChangeEvent) + }) + expect(result().mentionQuery).toBe(value.slice('Find @'.length)) + } + act(() => + result().insertResource({ type: 'file', id: 'plan', title: 'Quarterly plan (v2).md' }) + ) + expect(result().getPlainValue()).toBe('Find @Quarterly plan (v2).md ') + expect(result().getActiveContexts()).toEqual([ + { kind: 'file', fileId: 'plan', label: 'Quarterly plan (v2).md' }, + ]) + } finally { + unmount() + } + }) + + it('keeps a dismissed search closed while typing, but allows a new trigger', () => { + const { result, textarea, unmount } = renderPromptEditor(scope) + const input = (value: string) => { + act(() => { + typeInto(textarea, value) + result().handleInputChange({ target: textarea } as React.ChangeEvent) + }) + } + try { + input('@Quarterly') + act(() => result().handlePlusMenuClose()) + input('@Quarterlyplan') + expect(result().mentionQuery).toBeNull() + input('@Quarterly plan') + expect(result().mentionQuery).toBeNull() + input('@Quarterly plan @Roadmap') + expect(result().mentionQuery).toBe('Roadmap') + } finally { + unmount() + } + }) + + it.each(['selection', 'programmatic'])( + 'opens a fresh search when replacing a dismissed trigger via %s', + (replacement) => { + const { result, textarea, unmount } = renderPromptEditor(scope) + const input = (value: string) => { + act(() => { + typeInto(textarea, value) + result().handleInputChange({ + target: textarea, + } as React.ChangeEvent) + }) + } + try { + input('@Quarterly') + act(() => result().handlePlusMenuClose()) + if (replacement === 'selection') { + act(() => { + textarea.setSelectionRange(0, textarea.value.length) + result().handleSelectAdjust() + }) + expect(result().mentionQuery).toBeNull() + input('@Roadmap') + } else { + act(() => result().setValue('@Roadmap', { chipify: false })) + act(() => { + typeInto(textarea, result().getValue()) + result().handleSelectAdjust() + }) + } + expect(result().mentionQuery).toBe('Roadmap') + } finally { + unmount() + } + } + ) + + it.each(['@ ', '@ name', '@Quarterly\n', 'person@example.com'])( + 'does not search across a dismissed boundary in %j', + (value) => { + const { result, textarea, unmount } = renderPromptEditor(scope) + try { + act(() => { + typeInto(textarea, value) + result().handleInputChange({ + target: textarea, + } as React.ChangeEvent) + }) + expect(result().mentionQuery).toBeNull() + } finally { + unmount() + } + } + ) + } +) + describe('usePromptEditor context insertion', () => { it('leaves a cross-workspace selection to the ordinary plain-text paste path', () => { const context = { diff --git a/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/prompt-editor/use-prompt-editor.ts b/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/prompt-editor/use-prompt-editor.ts index a3e58fff4c7..b4ed623c6d6 100644 --- a/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/prompt-editor/use-prompt-editor.ts +++ b/apps/sim/app/workspace/[workspaceId]/home/components/user-input/components/prompt-editor/use-prompt-editor.ts @@ -217,11 +217,10 @@ export function usePromptEditor({ /** * Start offset of a mention/slash token most recently dismissed by the user - * (outside click or Escape) without a following keystroke — suppresses a - * single reopen of the menu for that exact token when the caret's own - * selection-change handler runs immediately after. + * (outside click or Escape). Mention dismissal lasts until the caret leaves + * that query; slash dismissal lasts until the next edit. */ - const dismissedMentionStartRef = useRef(null) + const dismissedMentionRef = useRef<{ start: number; triggerSelected: boolean } | null>(null) const dismissedSlashStartRef = useRef(null) const contextManagement = useContextManagement({ @@ -338,6 +337,16 @@ export function usePromptEditor({ */ const setValue = useCallback((text: string, options?: { chipify?: boolean }) => { const next = options?.chipify === false ? text : applyAutoMentionsRef.current(text) + atInsertPosRef.current = null + pendingCursorRef.current = null + mentionRangeRef.current = null + dismissedMentionRef.current = null + setMentionQuery(null) + plusMenuRef.current?.close() + slashRangeRef.current = null + dismissedSlashStartRef.current = null + setSlashQuery(null) + skillsMenuRef.current?.close() valueRef.current = next setValueState(next) }, []) @@ -411,7 +420,7 @@ export function usePromptEditor({ atInsertPosRef.current = null mentionRangeRef.current = null setMentionQuery(null) - dismissedMentionStartRef.current = null + dismissedMentionRef.current = null plusMenuRef.current?.close() slashRangeRef.current = null setSlashQuery(null) @@ -439,7 +448,7 @@ export function usePromptEditor({ plusMenuRef.current?.close() mentionRangeRef.current = null setMentionQuery(null) - dismissedMentionStartRef.current = null + dismissedMentionRef.current = null skillsMenuRef.current?.close() slashRangeRef.current = null setSlashQuery(null) @@ -493,7 +502,7 @@ export function usePromptEditor({ atInsertPosRef.current = newPos mentionRangeRef.current = null setMentionQuery(null) - dismissedMentionStartRef.current = null + dismissedMentionRef.current = null setValueState(newValue) } @@ -693,7 +702,9 @@ export function usePromptEditor({ * `onOpenChange` and never call this. */ const handlePlusMenuClose = useCallback(() => { - dismissedMentionStartRef.current = mentionRangeRef.current?.start ?? null + dismissedMentionRef.current = mentionRangeRef.current + ? { start: mentionRangeRef.current.start, triggerSelected: false } + : null atInsertPosRef.current = null mentionRangeRef.current = null setMentionQuery(null) @@ -708,30 +719,32 @@ export function usePromptEditor({ const syncMentionState = useCallback( (textarea: HTMLTextAreaElement, text: string, caret: number) => { if (!contextsEnabledRef.current) return + const dismissed = dismissedMentionRef.current + if (dismissed) { + dismissed.triggerSelected = + textarea.selectionStart <= dismissed.start && textarea.selectionEnd > dismissed.start + if (dismissed.triggerSelected) return + } const active = getActiveMentionAtRef.current(caret, text) - // Any word-boundary character inside the query — whitespace, sentence - // punctuation, or brackets — dismisses the menu. The mention token - // is "complete" the moment the user types a non-word character, so - // there's nothing more to query. Mirrors the boundary set the - // integration auto-detector uses for symmetry. - const isOpenable = active && !/[\s.,;:!?(){}[\]"'`/\\<>]/.test(active.query) + const isOpenable = active && !/[\r\n]/.test(active.query) if (!isOpenable) { if (mentionRangeRef.current !== null) { mentionRangeRef.current = null setMentionQuery(null) plusMenuRef.current?.close() } - dismissedMentionStartRef.current = null + dismissedMentionRef.current = null return } - if (active.start === dismissedMentionStartRef.current) { + if (active.start === dismissedMentionRef.current?.start) { if (mentionRangeRef.current !== null) { mentionRangeRef.current = null setMentionQuery(null) } return } + dismissedMentionRef.current = null const wasActive = mentionRangeRef.current !== null mentionRangeRef.current = { start: active.start, end: active.end } @@ -825,9 +838,16 @@ export function usePromptEditor({ pendingCursorRef.current = null const previousValue = valueRef.current const nextValue = e.target.value + const hasMentionQuery = + mentionRangeRef.current !== null || dismissedMentionRef.current !== null + if (dismissedMentionRef.current?.triggerSelected) dismissedMentionRef.current = null let finalValue = nextValue - if (contextsEnabledRef.current && nextValue.length === previousValue.length + 1) { + if ( + contextsEnabledRef.current && + !hasMentionQuery && + nextValue.length === previousValue.length + 1 + ) { // Single-char keystroke — synchronous, boundary-triggered. finalValue = integrationAutoMention.processChange({ textarea: e.target, @@ -843,6 +863,7 @@ export function usePromptEditor({ nextValue: finalValue, }) } else if ( + !hasMentionQuery && nextValue.length > previousValue.length + 1 && nextValue.length <= PASTE_RENDER_THRESHOLDS.ENHANCED_TEXT_CHARACTERS ) { @@ -867,7 +888,6 @@ export function usePromptEditor({ const caret = e.target.selectionStart ?? finalValue.length valueRef.current = finalValue setValueState(finalValue) - dismissedMentionStartRef.current = null dismissedSlashStartRef.current = null syncMentionState(e.target, finalValue, caret) syncSlashState(e.target, finalValue, caret) From 76a82b0e970a8d942f19a390737305a73419cb79 Mon Sep 17 00:00:00 2001 From: Waleed Date: Thu, 1 Oct 2026 01:13:16 -0700 Subject: [PATCH 3/9] chore(deps): update Next.js to 16.3.6 (#8510) --- apps/docs/package.json | 2 +- apps/sim/package.json | 4 +-- bun.lock | 44 ++++++++++++++++----------------- package.json | 12 ++++----- packages/emcn/package.json | 2 +- packages/sim-setup/package.json | 2 +- packages/testing/package.json | 2 +- 7 files changed, 34 insertions(+), 34 deletions(-) diff --git a/apps/docs/package.json b/apps/docs/package.json index 5396bf2cdf9..ac0e88dcf59 100644 --- a/apps/docs/package.json +++ b/apps/docs/package.json @@ -29,7 +29,7 @@ "fumadocs-mdx": "14.3.2", "fumadocs-openapi": "10.8.1", "fumadocs-ui": "16.8.5", - "next": "16.3.4", + "next": "16.3.6", "next-themes": "^0.4.6", "react": "19.2.4", "react-dom": "19.2.4", diff --git a/apps/sim/package.json b/apps/sim/package.json index c26813ebfbd..5ddb0010557 100644 --- a/apps/sim/package.json +++ b/apps/sim/package.json @@ -214,7 +214,7 @@ "mssql": "12.7.0", "mysql2": "3.24.2", "neo4j-driver": "6.0.1", - "next": "16.3.4", + "next": "16.3.6", "next-mdx-remote": "^6.0.0", "next-themes": "^0.4.6", "nodemailer": "10.0.10", @@ -263,7 +263,7 @@ "zustand": "^5.0.13" }, "devDependencies": { - "@next/env": "16.3.4", + "@next/env": "16.3.6", "@opentelemetry/context-async-hooks": "2.10.0", "@sim/testing": "workspace:*", "@sim/tsconfig": "workspace:*", diff --git a/bun.lock b/bun.lock index 9257e2e9096..eb4bb7cd7c0 100644 --- a/bun.lock +++ b/bun.lock @@ -36,10 +36,10 @@ "zod": "4.3.6", }, "optionalDependencies": { - "@next/swc-darwin-arm64": "16.3.4", - "@next/swc-darwin-x64": "16.3.4", - "@next/swc-linux-arm64-gnu": "16.3.4", - "@next/swc-linux-x64-gnu": "16.3.4", + "@next/swc-darwin-arm64": "16.3.6", + "@next/swc-darwin-x64": "16.3.6", + "@next/swc-linux-arm64-gnu": "16.3.6", + "@next/swc-linux-x64-gnu": "16.3.6", }, }, "apps/desktop": { @@ -98,7 +98,7 @@ "fumadocs-mdx": "14.3.2", "fumadocs-openapi": "10.8.1", "fumadocs-ui": "16.8.5", - "next": "16.3.4", + "next": "16.3.6", "next-themes": "^0.4.6", "react": "19.2.4", "react-dom": "19.2.4", @@ -340,7 +340,7 @@ "mssql": "12.7.0", "mysql2": "3.24.2", "neo4j-driver": "6.0.1", - "next": "16.3.4", + "next": "16.3.6", "next-mdx-remote": "^6.0.0", "next-themes": "^0.4.6", "nodemailer": "10.0.10", @@ -389,7 +389,7 @@ "zustand": "^5.0.13", }, "devDependencies": { - "@next/env": "16.3.4", + "@next/env": "16.3.6", "@opentelemetry/context-async-hooks": "2.10.0", "@sim/testing": "workspace:*", "@sim/tsconfig": "workspace:*", @@ -555,7 +555,7 @@ "class-variance-authority": "^0.7.1", "framer-motion": "^12.5.0", "input-otp": "^1.4.2", - "next": "16.3.4", + "next": "16.3.6", "prismjs": "^1.30.0", "react": "19.2.4", "react-dom": "19.2.4", @@ -686,7 +686,7 @@ }, "devDependencies": { "@clack/prompts": "1.7.0", - "@next/env": "16.3.4", + "@next/env": "16.3.6", "@sim/deployment-config": "workspace:*", "@sim/security": "workspace:*", "@sim/tsconfig": "workspace:*", @@ -720,7 +720,7 @@ }, "devDependencies": { "@sim/tsconfig": "workspace:*", - "next": "16.3.4", + "next": "16.3.6", "stripe": "18.5.0", "typescript": "^7.0.2", "vitest": "^5.0.1", @@ -833,7 +833,7 @@ }, "overrides": { "@hono/node-server": "1.19.15", - "@next/env": "16.3.4", + "@next/env": "16.3.6", "brace-expansion": "5.0.9", "deepmerge-ts": "8.0.0", "dompurify": "3.4.13", @@ -846,7 +846,7 @@ "marked": "18.0.11", "mermaid": "11.16.1", "minimatch": "^10.2.5", - "next": "16.3.4", + "next": "16.3.6", "postcss": "8.5.26", "postgres": "^3.4.5", "protobufjs": "7.6.5", @@ -1562,23 +1562,23 @@ "@napi-rs/wasm-runtime": ["@napi-rs/wasm-runtime@1.2.4", "", { "dependencies": { "@tybys/wasm-util": "^0.10.3" }, "peerDependencies": { "@emnapi/core": "^1.7.1 || ^2.0.0-alpha.4", "@emnapi/runtime": "^1.7.1 || ^2.0.0-alpha.4" } }, "sha512-AJxoUD2/15ESHbvpcyjU274nsAPLuOtPHCk0vKJM5pj//Fg/B1FXNWjPnXTT9PymCYYiHo4zPj0ZomXBKhoy7g=="], - "@next/env": ["@next/env@16.3.4", "", {}, "sha512-cjWZnUUa6jZq2kFaNe/ZyJdZonOZ/QoN0Zka2nz/FLOrfx14pQuM9c5RaSVkWMqgdt4ksgPAMWPyHSs/CyV48Q=="], + "@next/env": ["@next/env@16.3.6", "", {}, "sha512-x9Vblze1EbtltQYnNH38xCPWU3TVfBd1eXqA3+w9+BTpedkkdNpAaltXlGQ/nsc1+E0mVTNrtcbX3GoO09zeLQ=="], - "@next/swc-darwin-arm64": ["@next/swc-darwin-arm64@16.3.4", "", { "os": "darwin", "cpu": "arm64" }, "sha512-iBr3I5LZNk5/bgl5//iTgD2tcym14MX0Xo7fD//u9dYAEgGzza1y9oywluPtf74YnOswVdH1908aK9xVz7zQTw=="], + "@next/swc-darwin-arm64": ["@next/swc-darwin-arm64@16.3.6", "", { "os": "darwin", "cpu": "arm64" }, "sha512-E/7GEqaUkt8mk/T8v9lAnrhzR06kdq1ZBkC12F8tAMkdIadwNp3H1KqHynDHrpcTlGCUdq/qu6vUL2aYVyYBdw=="], - "@next/swc-darwin-x64": ["@next/swc-darwin-x64@16.3.4", "", { "os": "darwin", "cpu": "x64" }, "sha512-2dpiSyl2Jw/NrBPaU2MAKGSa+2MR82pJIn4Sm5Rjr+gxAeuh0z158Su3Z2O8zn7UNNq+ej4bToed6RcRN/Lydg=="], + "@next/swc-darwin-x64": ["@next/swc-darwin-x64@16.3.6", "", { "os": "darwin", "cpu": "x64" }, "sha512-yBE893/nDWTlaiBD1p+qgt7NUen4U5R6FXyH0s67Npq1S3E0cVSef1WIXC2xBRgQvwAvJq6DnS6Y6PrY0cy4Ew=="], - "@next/swc-linux-arm64-gnu": ["@next/swc-linux-arm64-gnu@16.3.4", "", { "os": "linux", "cpu": "arm64" }, "sha512-+t+U8HZT+fApePCS5h89CSH3datz29MkzyfCn+6fpsZBG/oiEOhINcb9rtkv6sdpToLGFn2e6146NzaKCXkqrA=="], + "@next/swc-linux-arm64-gnu": ["@next/swc-linux-arm64-gnu@16.3.6", "", { "os": "linux", "cpu": "arm64" }, "sha512-KJDpjBqBPYlvkivmyrp+Qys6k/7ksbqGQvRVc6ZEGfR+cjQxx+nUkJaWmNZJsmoOrqYNbaXByF8wa0lBwDhB3Q=="], - "@next/swc-linux-arm64-musl": ["@next/swc-linux-arm64-musl@16.3.4", "", { "os": "linux", "cpu": "arm64" }, "sha512-mx03GNs1ocQA5JQ4FxDMmIsNkdrZh8cuezKCrId28e5/gIPU/l7Kcy2+vmCCzdjnnmXJy+iOAu+7K0QppO6Urg=="], + "@next/swc-linux-arm64-musl": ["@next/swc-linux-arm64-musl@16.3.6", "", { "os": "linux", "cpu": "arm64" }, "sha512-mqNg2K+hvWskSRb/QM+Ix412DvBsuSF0XV+frTSw5vmoucNnIlynFwKYew8D01bfATErMOM7Bujrf0BA5DRKFA=="], - "@next/swc-linux-x64-gnu": ["@next/swc-linux-x64-gnu@16.3.4", "", { "os": "linux", "cpu": "x64" }, "sha512-YIhGY6fSMfha52bnVxnzc9zaVBzJg+cqQTOD8tXIBSx4fuv0pVMxQTE0PaS59YhnMOiYiG09IMwxJAf/CFm/Dw=="], + "@next/swc-linux-x64-gnu": ["@next/swc-linux-x64-gnu@16.3.6", "", { "os": "linux", "cpu": "x64" }, "sha512-nFncBNGAYouRHjRVaITs9beZRfhX4ssVwpnvPIAbkZVH6LtGoAVlH4bJ8Cnf9SOo9bsXgPFer/GdHtEE3JNOkw=="], - "@next/swc-linux-x64-musl": ["@next/swc-linux-x64-musl@16.3.4", "", { "os": "linux", "cpu": "x64" }, "sha512-+eaaX6axpDb0yF1GCpiERe6njplvdC+nks/fKfcHu3XPGRrald8P3/X7yv7QLdjA51knnxwl9pxdIJsg+w1L+Q=="], + "@next/swc-linux-x64-musl": ["@next/swc-linux-x64-musl@16.3.6", "", { "os": "linux", "cpu": "x64" }, "sha512-5Mf3cHDGR/Iz0ng2Bj3zUR3p5QS9YK3Hn2QiAfavFmyF48zwThAjpFoiTKNIcOHLYS4zEk+gzyJ/9deQ2ZB8yQ=="], - "@next/swc-win32-arm64-msvc": ["@next/swc-win32-arm64-msvc@16.3.4", "", { "os": "win32", "cpu": "arm64" }, "sha512-0jcXW7Xs/uzICrmgV3MhDYDeRy++1CqnpDIerlPIqYO4bhzB4WNbX/aRnQclustsAyTkFKB0z6rbcjmNg5tR8A=="], + "@next/swc-win32-arm64-msvc": ["@next/swc-win32-arm64-msvc@16.3.6", "", { "os": "win32", "cpu": "arm64" }, "sha512-0jkJy0C2kbrJWTk4YLa3xk80pVBpx8FCHJym7CnUfDAXe/FWv5qT7SQJbR0KuemyxaEDlEx5WT4VQJoTW+/9Qw=="], - "@next/swc-win32-x64-msvc": ["@next/swc-win32-x64-msvc@16.3.4", "", { "os": "win32", "cpu": "x64" }, "sha512-vvBzwu1pYQCp92maZCFCIw/XgOTMR5tur9GjakwIo2cmwRTMKajRZZDS9+e4KsUZWKu1E007WUeAFXRRjZeuzw=="], + "@next/swc-win32-x64-msvc": ["@next/swc-win32-x64-msvc@16.3.6", "", { "os": "win32", "cpu": "x64" }, "sha512-/YXjI1e5OXcZ7YpxRwgP/1jAV/SBKTzeVKqN2mk7mLpcICsyn3Gl5+dIfDTJp70M0ccMhyMMRso4v6mPDCGepg=="], "@noble/ciphers": ["@noble/ciphers@2.2.0", "", {}, "sha512-Z6pjIZ/8IJcCGzb2S/0Px5J81yij85xASuk1teLNeg75bfT07MV3a/O2Mtn1I2se43k3lkVEcFaR10N4cgQcZA=="], @@ -3920,7 +3920,7 @@ "neo4j-driver-core": ["neo4j-driver-core@6.0.1", "", {}, "sha512-5I2KxICAvcHxnWdJyDqwu8PBAQvWVTlQH2ve3VQmtVdJScPqWhpXN1PiX5IIl+cRF3pFpz9GQF53B5n6s0QQUQ=="], - "next": ["next@16.3.4", "", { "dependencies": { "@next/env": "16.3.4", "@swc/helpers": "0.5.23", "baseline-browser-mapping": "^2.9.19", "caniuse-lite": "^1.0.30001579", "postcss": "8.5.23", "styled-jsx": "5.1.6" }, "optionalDependencies": { "@next/swc-darwin-arm64": "16.3.4", "@next/swc-darwin-x64": "16.3.4", "@next/swc-linux-arm64-gnu": "16.3.4", "@next/swc-linux-arm64-musl": "16.3.4", "@next/swc-linux-x64-gnu": "16.3.4", "@next/swc-linux-x64-musl": "16.3.4", "@next/swc-win32-arm64-msvc": "16.3.4", "@next/swc-win32-x64-msvc": "16.3.4", "sharp": "^0.35.4" }, "peerDependencies": { "@opentelemetry/api": "^1.1.0", "@playwright/test": "^1.51.1", "babel-plugin-react-compiler": "*", "react": "^18.2.0 || 19.0.0-rc-de68d2f4-20241204 || ^19.0.0", "react-dom": "^18.2.0 || 19.0.0-rc-de68d2f4-20241204 || ^19.0.0", "sass": "^1.3.0" }, "optionalPeers": ["@opentelemetry/api", "@playwright/test", "babel-plugin-react-compiler", "sass"], "bin": { "next": "dist/bin/next" } }, "sha512-/Ztf6CeRH+ejEXUrYtqI4gkS66eFIHuSwqi60RgcpWKodxFZx2/dqVCMKBwILfAHXQ+F1b1vAudgj3mnxqtoIA=="], + "next": ["next@16.3.6", "", { "dependencies": { "@next/env": "16.3.6", "@swc/helpers": "0.5.23", "baseline-browser-mapping": "^2.9.19", "caniuse-lite": "^1.0.30001579", "postcss": "8.5.23", "styled-jsx": "5.1.6" }, "optionalDependencies": { "@next/swc-darwin-arm64": "16.3.6", "@next/swc-darwin-x64": "16.3.6", "@next/swc-linux-arm64-gnu": "16.3.6", "@next/swc-linux-arm64-musl": "16.3.6", "@next/swc-linux-x64-gnu": "16.3.6", "@next/swc-linux-x64-musl": "16.3.6", "@next/swc-win32-arm64-msvc": "16.3.6", "@next/swc-win32-x64-msvc": "16.3.6", "sharp": "^0.35.4" }, "peerDependencies": { "@opentelemetry/api": "^1.1.0", "@playwright/test": "^1.51.1", "babel-plugin-react-compiler": "*", "react": "^18.2.0 || 19.0.0-rc-de68d2f4-20241204 || ^19.0.0", "react-dom": "^18.2.0 || 19.0.0-rc-de68d2f4-20241204 || ^19.0.0", "sass": "^1.3.0" }, "optionalPeers": ["@opentelemetry/api", "@playwright/test", "babel-plugin-react-compiler", "sass"], "bin": { "next": "dist/bin/next" } }, "sha512-L+otWM/aQbYTx98aZhgEoMb4bZAXx1YVW4UMA/vuCyCoWG5HJyZUili8QAkqzrcC+5///tsz3s0M+SlyB5bLMw=="], "next-mdx-remote": ["next-mdx-remote@6.0.0", "", { "dependencies": { "@babel/code-frame": "^7.23.5", "@mdx-js/mdx": "^3.0.1", "@mdx-js/react": "^3.0.1", "unist-util-remove": "^4.0.0", "unist-util-visit": "^5.1.0", "vfile": "^6.0.1", "vfile-matter": "^5.0.0" }, "peerDependencies": { "react": ">=16" } }, "sha512-cJEpEZlgD6xGjB4jL8BnI8FaYdN9BzZM4NwadPe1YQr7pqoWjg9EBCMv3nXBkuHqMRfv2y33SzUsuyNh9LFAQQ=="], diff --git a/package.json b/package.json index e55b140888a..18fc1fc8916 100644 --- a/package.json +++ b/package.json @@ -130,8 +130,8 @@ "overrides": { "react": "19.2.4", "react-dom": "19.2.4", - "next": "16.3.4", - "@next/env": "16.3.4", + "next": "16.3.6", + "@next/env": "16.3.6", "drizzle-orm": "^0.45.2", "postgres": "^3.4.5", "minimatch": "^10.2.5", @@ -154,10 +154,10 @@ "ws": "8.21.3" }, "optionalDependencies": { - "@next/swc-darwin-arm64": "16.3.4", - "@next/swc-darwin-x64": "16.3.4", - "@next/swc-linux-arm64-gnu": "16.3.4", - "@next/swc-linux-x64-gnu": "16.3.4" + "@next/swc-darwin-arm64": "16.3.6", + "@next/swc-darwin-x64": "16.3.6", + "@next/swc-linux-arm64-gnu": "16.3.6", + "@next/swc-linux-x64-gnu": "16.3.6" }, "devDependencies": { "@babel/parser": "7.29.2", diff --git a/packages/emcn/package.json b/packages/emcn/package.json index dfd6a36b6a5..1650443a219 100644 --- a/packages/emcn/package.json +++ b/packages/emcn/package.json @@ -85,7 +85,7 @@ "class-variance-authority": "^0.7.1", "framer-motion": "^12.5.0", "input-otp": "^1.4.2", - "next": "16.3.4", + "next": "16.3.6", "prismjs": "^1.30.0", "react": "19.2.4", "react-dom": "19.2.4", diff --git a/packages/sim-setup/package.json b/packages/sim-setup/package.json index 1acb6ac7b1d..e0938b6fdbe 100644 --- a/packages/sim-setup/package.json +++ b/packages/sim-setup/package.json @@ -49,7 +49,7 @@ }, "devDependencies": { "@clack/prompts": "1.7.0", - "@next/env": "16.3.4", + "@next/env": "16.3.6", "@sim/deployment-config": "workspace:*", "@sim/security": "workspace:*", "@sim/tsconfig": "workspace:*", diff --git a/packages/testing/package.json b/packages/testing/package.json index 215c13eba67..427070039eb 100644 --- a/packages/testing/package.json +++ b/packages/testing/package.json @@ -72,7 +72,7 @@ }, "devDependencies": { "@sim/tsconfig": "workspace:*", - "next": "16.3.4", + "next": "16.3.6", "stripe": "18.5.0", "typescript": "^7.0.2", "vitest": "^5.0.1" From 96349b91e28d119a2e6907c0dd2718f45b74e574 Mon Sep 17 00:00:00 2001 From: Vikhyath Mondreti Date: Thu, 1 Oct 2026 09:44:52 -0700 Subject: [PATCH 4/9] feat(library): AI Agent Marketplace vs Building From Scratch: Where Sim Fits (#8519) * feat(library): AI Agent Marketplace vs Building From Scratch: Where Sim Fits * Pi Babysit: address PR #8519 feedback --------- Co-authored-by: Sim Pi Agent --- .../index.mdx | 330 +++++++++--------- .../cover.jpg | Bin 28873 -> 30662 bytes 2 files changed, 171 insertions(+), 159 deletions(-) diff --git a/apps/sim/content/library/ai-agent-marketplace-vs-building-from-scratch/index.mdx b/apps/sim/content/library/ai-agent-marketplace-vs-building-from-scratch/index.mdx index 1789a6c8fef..29cfc4084e0 100644 --- a/apps/sim/content/library/ai-agent-marketplace-vs-building-from-scratch/index.mdx +++ b/apps/sim/content/library/ai-agent-marketplace-vs-building-from-scratch/index.mdx @@ -1,243 +1,255 @@ --- slug: ai-agent-marketplace-vs-building-from-scratch -title: 'AI Agent Marketplace vs Building From Scratch: Which Should You Choose?' -description: 'Compare an AI agent marketplace with building from scratch in Sim across templates, customization, integrations, deployment, licensing, and ownership.' +title: 'AI Agent Marketplace vs Building From Scratch: Where Sim Fits' +description: 'Compare AI agent marketplaces, reusable templates, flexible builders, and custom development to decide where Sim fits your workflow and governance needs.' date: 2026-09-17 -updated: 2026-09-17 +updated: 2026-10-01 authors: - andrew readingTime: 10 -tags: [AI Agents, AI Agent Marketplaces, Automation, Sim] +tags: [AI Agents, Agent Builders, Workflow Automation, Sim] ogImage: /library/ai-agent-marketplace-vs-building-from-scratch/cover.jpg draft: false faq: - q: "What is an AI agent marketplace?" - a: "An AI agent marketplace is a catalog where buyers discover and access pre-built agents offered by a platform, vendor, or independent builder." - - q: "Where can I buy pre-built AI agents?" - a: "Dedicated AI agent marketplaces let buyers obtain pre-built agents, while Sim provides editable templates for teams that want to build and control the resulting workflow." - - q: "Should I buy a pre-built AI agent or build one from scratch?" - a: "Sim is the better choice when customization, integrations, self-hosting, or deployment control matters, while a pre-built marketplace agent is better for quickly testing a narrow use case." - - q: "Does Sim have pre-built AI agent templates?" - a: "Sim has a native template library that provides editable starting points for building AI agents and automated workflows." - - q: "Is Sim an AI agent marketplace?" - a: "Sim is an AI agent-building platform with a native template library, not a pure discovery marketplace for third-party agents." - - q: "Can I customize a pre-built AI agent in Sim?" - a: "Sim lets users customize templates through Chat, its visual builder, code, APIs, integrations, prompts, models, and workflow logic." - - q: "What does Chat do in Sim?" - a: "Sim's Chat provides a prompt-driven way to create and modify an AI agent workflow from a requested outcome." - - q: "Can Sim connect an AI agent to business applications?" - a: "Sim reported more than 1,000 integrations and 266 first-party blocks as of September 2026, allowing workflows to connect agents with business applications and services." - - q: "Can Sim deploy an AI agent as an API?" - a: "Sim can deploy an AI agent through an API so it can be called by a product, internal tool, or backend service." - - q: "Can Sim deploy an AI agent as a hosted chat?" - a: "Sim can deploy an AI agent as hosted chat so users can interact with it through a conversational interface." - - q: "Does Sim support MCP?" - a: "Sim supports deployment through Model Context Protocol so agent capabilities can be exposed to compatible clients and tools." - - q: "Can Sim be self-hosted?" - a: "Sim can be self-hosted under the Apache 2.0 license, with model, infrastructure, and external service costs determined separately." + a: "An AI agent marketplace is a catalog where buyers discover ready-made agents, editable templates, workflow components, or agent-building services." + - q: "What is the difference between an AI agent marketplace and an AI agent builder?" + a: "An AI agent marketplace distributes existing agents or templates, while an AI agent builder such as Sim gives teams tools to create and modify their own agent workflows." + - q: "What is the difference between an AI agent template and a ready-made agent?" + a: "An AI agent template is intended to be edited and configured, while a ready-made agent is intended to deliver a defined outcome with less modification." + - q: "Should I use an AI agent marketplace or build from scratch?" + a: "An AI agent marketplace is best for common low-risk needs, while Sim or another flexible builder is better for workflows requiring proprietary logic, integrations, governance, or deployment control." + - q: "When are AI agent templates enough?" + a: "AI agent templates are enough when the workflow is repeatable, the required integrations already fit, and failures are limited, detectable, and recoverable." + - q: "Can I customize an agent from a marketplace?" + a: "A marketplace agent can be customized only to the extent that its publisher exposes the agent’s prompts, tools, workflow logic, permissions, and failure handling." + - q: "Are AI agents from marketplaces safe?" + a: "AI marketplace agents are safe only after the buyer validates their publisher, permissions, data handling, dependencies, behavior, and operational controls for the intended use case." + - q: "Are AI marketplace agents portable?" + a: "AI marketplace agents are portable only when buyers can export or reconstruct their logic, data, dependencies, and runtime outside the original platform." + - q: "What should I look for in an AI agent template?" + a: "An AI agent template should expose its workflow, required tools, permissions, inputs, outputs, failure paths, version information, and customization points." + - q: "How do I govern an AI agent from a template?" + a: "An AI agent created from a template should have a named owner, least-privilege credentials, representative tests, execution logs, approval controls, version tracking, and a shutdown plan." - q: "Is Sim open source?" - a: "Sim is open source under the OSI-approved Apache 2.0 license as of September 2026." + a: "Sim is open source under the OSI-approved Apache License 2.0 and can be self-hosted under that license." + - q: "Is Sim free?" + a: "Sim’s Apache 2.0 codebase can be self-hosted without a software license fee, while any current hosted-service pricing should be confirmed on Sim’s official pricing page." - q: "Is n8n open source?" - a: "n8n is source-available under the Sustainable Use License as of September 2026, but that license is not OSI-approved open source." - - q: "Is Sim better than n8n for AI agents?" - a: "Sim is better suited to teams prioritizing agent creation and API, hosted chat, or MCP deployment, while n8n is stronger for teams prioritizing a mature general workflow automation ecosystem." - - q: "Do AI agent marketplaces support self-hosting?" - a: "AI agent marketplaces do not share one self-hosting policy because hosting rights depend on the marketplace, seller, agent, and license." - - q: "Are marketplace AI agents production-ready?" - a: "A marketplace AI agent is production-ready only after the buyer verifies its reliability, security, data handling, integrations, maintenance, deployment access, and licensing." - - q: "What is the difference between an AI agent marketplace and an AI agent template library?" - a: "An AI agent marketplace focuses on discovering packaged agents, while Sim's template library provides editable starting points for building agents inside the Sim platform." + a: "As of September 2026, n8n is source-available under the Sustainable Use License and is not open source under an OSI-approved license." + - q: "Is Sim better than n8n for building AI agents?" + a: "Sim is the stronger fit when an Apache 2.0 license and an AI-agent-focused workflow builder are priorities, while n8n remains a relevant incumbent for teams evaluating broad workflow automation." + - q: "What is the best n8n alternative for AI agent workflows?" + a: "Sim is a strong n8n alternative for teams that prioritize an Apache 2.0 license, self-hosting, and customizable AI agent workflows." + - q: "What is the best open-source Zapier alternative for AI agents?" + a: "Sim is a strong open-source Zapier alternative for AI agent workflows because Sim uses the OSI-approved Apache License 2.0 and supports self-hosting." + - q: "How does Sim compare with Gumloop?" + a: "Sim is the stronger fit when Apache 2.0 licensing and self-hosting are requirements, while teams considering Gumloop should compare its current managed-service capabilities, portability, governance controls, and commercial terms directly." - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for teams seeking an open-source visual platform, and the full market comparison is available in Sim's canonical Best AI Agent Platforms and Builders in 2026 guide." + a: "Sim is a leading AI agent builder for teams prioritizing flexible workflows, self-hosting, and Apache 2.0 licensing, while the full category comparison is covered in Sim’s canonical best AI agent builder guide." + - q: "Can I start with a template and move to Sim later?" + a: "Sim can replace or extend a template-led approach when the template’s logic and dependencies are documented well enough to reconstruct the workflow." + - q: "Do I need developers to build an AI agent?" + a: "Sim reduces the amount of code needed to assemble an agent workflow, but technical review may still be necessary for APIs, security, data handling, testing, and production operations." + - q: "Is a large agent marketplace better than a small one?" + a: "An AI agent marketplace is better only when its relevant agents are transparent, maintained, secure, customizable, and suitable for the buyer’s actual workflow; catalog size alone does not establish quality." --- ## TL;DR -Sim is the better choice when you need a customizable AI agent that can connect to your systems and deploy across multiple surfaces, while a dedicated AI agent marketplace is better when you want to discover and try a packaged agent with minimal setup. Sim is an agent-building platform with a native template library, not a pure discovery marketplace in the same category as services such as [Arahi](https://arahi.ai/marketplace) or [Agentmarketplace.ai](https://agentmarketplace.ai/). +An AI agent marketplace is the fastest path to a standardized use case, while a flexible builder such as Sim is usually the better fit when the agent must reflect a team’s own systems, rules, data, and governance requirements. -## Should I choose an AI agent marketplace or build an AI agent with Sim? +The choice is not simply “buy or build.” Teams can install a ready-made agent, adapt a reusable template, assemble an agent in a visual builder, or write a custom agent in code. The right starting point depends on how differentiated the workflow is, how much control the team needs, and what will happen when the first version must change. -Sim gives teams more control over customization, integrations, deployment, and ownership than a dedicated AI agent marketplace, while marketplaces optimize for speed of discovery. - -> **Choose a dedicated AI agent marketplace when:** -> - You want to browse packaged agents by use case. -> - You want the fastest possible path to testing a narrow task. -> - You do not need deep changes to the agent's workflow or interface. -> - You are comfortable evaluating each listing's vendor, license, data handling, integrations, and support separately. -> -> **Choose Sim when:** -> - You want a pre-built template as a starting point rather than a fixed product. -> - You need to change the workflow, prompts, models, tools, logic, or code. -> - You need access to Sim's 1,000-plus integrations and 266 first-party blocks, as reported by Sim in September 2026. -> - You want to deploy through an API, hosted chat, or Model Context Protocol server. -> - You want an Apache 2.0 platform that can be self-hosted. - -Teams comparing the broader AI agent builder market should also read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), which is Sim's canonical guide for that head-term question. +This guide explains where each approach works, what buyers should evaluate, and when Sim fits. For the broader category ranking, see [the best AI agent platforms in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). ## What is an AI agent marketplace? -An AI agent marketplace is a catalog where buyers discover pre-built agents, compare use cases, and obtain access to an agent created by a vendor or independent builder. - -The marketplace model is useful when discovery is the main problem. A buyer may search for an agent that researches accounts, qualifies leads, summarizes documents, or handles support requests without first designing a workflow. - -Marketplace listings can differ substantially in quality and terms. Before adopting one, buyers should check: +An AI agent marketplace is a catalog where buyers can discover ready-made agents, agent templates, tools, or workflow components published by a platform or third parties. -- Whether the agent is configurable or effectively fixed -- Which models and data sources it uses -- Whether it can connect to private business systems -- Where prompts, files, and outputs are stored -- Whether it can be self-hosted -- How the agent is licensed and billed -- Whether the seller provides maintenance and support +The word “marketplace” can describe several different products: -A marketplace listing should be treated as third-party software, not as a universally interchangeable template. +- A catalog of complete agents intended to work with minimal configuration +- A template gallery containing editable workflows or starter projects +- A directory that sends buyers to external agent vendors +- A component marketplace for tools, connectors, prompts, or models +- A services marketplace where specialists build or customize agents for customers -## What are pre-built AI agent templates? +Buyers should identify which model a marketplace uses before comparing it with a builder. A complete managed agent is a different purchase from an editable workflow template, even when both appear in the same catalog. -Sim's pre-built AI agent templates are editable workflow starting points that teams can inspect, customize, connect to their own systems, and deploy. +## What are AI agent templates? -A template reduces setup work without forcing the buyer to accept a finished agent as-is. Teams can replace models, edit prompts, add branches, connect integrations, introduce human approval, call custom APIs, and change the deployment surface. The guide to [creating an AI agent](https://www.sim.ai/library/how-to-create-an-ai-agent) explains how these building choices fit together. +AI agent templates are reusable starting configurations that define part of an agent’s instructions, tools, workflow, data flow, or decision logic. -This distinction matters because a marketplace agent and a builder template solve different problems. A marketplace agent is typically selected as a product. A Sim template is selected as the foundation for a system the buyer controls. +A useful template reduces setup time without hiding the decisions that matter. It should make clear which systems the agent can access, what information it receives, what actions it can take, where human approval occurs, and how failures are handled. -Explore the current platform and template experience on [Sim](https://www.sim.ai), and consult the [Sim documentation](https://docs.sim.ai) for implementation details. +Templates are best treated as starting points rather than finished operating systems. Even a strong template may require new credentials, field mappings, prompts, permissions, model choices, error handling, and tests before it is safe to use in production. -## Is Sim an AI agent marketplace? +## Should I use an AI agent marketplace or build an agent from scratch? -Sim is an AI agent-building platform with a native template library, not a pure AI agent discovery marketplace. +An AI agent marketplace is usually better for common, low-risk tasks, while building with Sim or another flexible agent builder is usually better for differentiated workflows that require control and continued iteration. -Sim helps users move from an idea or template to an editable workflow. Its core value is the ability to build, connect, test, and deploy agents rather than merely browse agents sold or published by other parties. +Use a marketplace agent or template when: -Discovery-first services such as [Arahi](https://arahi.ai/marketplace) and [Agentmarketplace.ai](https://agentmarketplace.ai/) center their product experience on catalogs of pre-built agents. Their catalogs, availability, pricing, and listing terms can change, so buyers should verify each service and listing directly before relying on it. +- The use case is common and well defined +- The required systems are already supported +- The agent will handle low-risk work +- Speed matters more than deep customization +- The default behavior is close to the desired behavior +- A team can replace the agent without disrupting a core process -## How do AI agent marketplaces and Sim compare? +Use a flexible builder when: -Sim provides greater building and deployment control, while dedicated marketplaces provide a more direct catalog-shopping experience. +- The workflow crosses several internal systems +- Business rules are unique or change frequently +- The agent needs custom tools, APIs, or data transformations +- Human approval must occur at specific steps +- Teams need to inspect and modify agent behavior +- Deployment, self-hosting, or data control is a requirement +- The agent may become part of a business-critical process -| Buyer consideration | Dedicated AI agent marketplace | Sim | -|---|---|---| -| Primary purpose | Discover and obtain packaged agents | Build and deploy customizable agents | -| Starting point | Third-party agent listing | Blank workflow, Chat, or native template | -| Template availability | Depends on current marketplace inventory | Native templates designed to be edited in Sim | -| Customization depth | Varies by listing and seller | Visual builder, Chat, code, and API access | -| Integration model | Varies by agent | 1,000-plus integrations and 266 first-party blocks reported by Sim as of September 2026 | -| Deployment | Determined by the listing | API, hosted chat, and MCP | -| Self-hosting | Listing-specific | Supported under Sim's Apache 2.0 license | -| Ownership and portability | Depends on seller and license | Workflow logic can be inspected and modified in the platform | -| Best fit | Fast discovery of a narrow packaged solution | Custom agents connected to business systems | +Build with code when the agent requires capabilities that available builders cannot express, the organization already has the necessary engineering capacity, and long-term ownership of a custom software system is justified. -No marketplace-wide assumption should replace checking the terms of the individual agent listing. +## How do AI agent marketplaces, templates, builders, and custom code compare? -## Which option has more pre-built AI agent templates? +AI agent marketplaces optimize for discovery, templates optimize for a fast start, builders such as Sim optimize for adaptable workflows, and custom code optimizes for maximum engineering control. -A dedicated AI agent marketplace may offer more independent listings, while Sim offers native templates that are designed to become editable workflows. +| Approach | Best for | Time to first prototype | Customization | Portability | Governance effort | Ongoing ownership | +|---|---|---:|---|---|---|---| +| Ready-made marketplace agent | Standardized, low-risk use cases | Fastest | Low to medium | Often limited | Must validate the publisher’s controls | Vendor or publisher dependent | +| Editable agent template | Common workflows that need adaptation | Fast | Medium to high if the template is inspectable | Depends on export format and platform dependencies | Shared between template publisher and buyer | Buyer maintains its modified version | +| Visual agent builder such as Sim | Custom multi-step workflows and internal processes | Fast to moderate | High | Stronger when source and deployment are controllable | Buyer defines approvals, permissions, and monitoring | Buyer controls iteration | +| Custom code | Novel products or highly specialized infrastructure | Slowest | Highest | Potentially highest, subject to architecture choices | Entirely buyer owned | Requires engineering maintenance | -Raw catalog size is not the only useful measure of template availability. Buyers should also ask whether a template can be inspected, modified, maintained, and connected to the systems required for production use. +The table describes acquisition models rather than absolute product capabilities. A marketplace can contain editable projects, and a builder can also provide templates. Buyers should evaluate the artifact they receive, not only the label used to market it. -A large marketplace may provide variety across sellers. Sim's advantage is continuity between template selection and agent development: the same environment supports editing, testing, integrations, and deployment. +## When are AI agent templates sufficient? -As of September 2026, template counts and marketplace inventories remain changing claims. Buyers should verify the current catalog on each vendor's own website rather than relying on an undated directory or third-party roundup. +AI agent templates are sufficient when the workflow is repeatable, the integrations already fit, and the consequences of incorrect behavior are limited and recoverable. -## How much can I customize a pre-built AI agent? +Examples can include drafting internal summaries, categorizing inbound requests, preparing research briefs, or routing information for human review. These tasks still require testing, but they usually do not justify a fully custom architecture at the start. -Sim allows deeper customization than a fixed marketplace listing because teams can change an agent through Chat, the visual builder, code, and APIs. +A template is especially useful for validating demand. A team can test whether users benefit from the workflow before investing in deeper customization. If the template repeatedly requires workarounds, duplicated steps, or manual intervention, that is evidence that the use case has outgrown the template. -Sim supports several paths from idea to working agent: +## When should I use a flexible AI agent builder? -1. **Chat:** Use a prompt-driven experience to turn a requested outcome into an agent workflow. -2. **Visual builder:** Inspect and change blocks, connections, branching logic, prompts, models, and tool calls. -3. **Code:** Add custom behavior when a visual block is not sufficient. -4. **APIs:** Connect private services or trigger the resulting workflow from another application. +Sim is a stronger fit than a fixed marketplace agent when a team needs to change the workflow’s logic, tools, model interactions, approval steps, or deployment environment. -A marketplace agent may also be customizable, but the depth depends on what its seller exposes. Some listings may allow prompt or data-source changes, while others function as closed applications. +A flexible builder is appropriate when the agent must mirror an existing operating process rather than force the process into a generic template. Common requirements include conditional routing, multiple model calls, custom API requests, retrieval from proprietary sources, structured outputs, retries, and human review. The guide to [AI agent workflow builders for multi-step tasks](https://www.sim.ai/library/ai-agent-workflow-builders-multi-step-tasks) explores these requirements in more detail. -Implementation details for Sim's current building features are available in the [Sim documentation](https://docs.sim.ai). +The builder approach also supports progressive adoption. Teams can begin with a small workflow, observe its behavior, and add autonomy only after the system is reliable. This is generally safer than installing an opaque agent and granting broad permissions immediately. -## Which option has more AI agent integrations? +## How should I evaluate an AI agent marketplace? -Sim is the stronger choice when integration breadth and workflow-level control are requirements because Sim reported more than 1,000 integrations and 266 first-party blocks as of September 2026. +An AI agent marketplace should be evaluated on transparency, customization, security, portability, publisher accountability, and the operational quality of its agents—not on catalog size alone. The [AI workflow automation platform buyer’s checklist](https://www.sim.ai/library/ai-workflow-automation-platform-buyers-checklist) provides a complementary platform-level review. -An integration count should be interpreted carefully. Marketplace listings may advertise support for popular applications, but those connections can differ from one agent to another. A buyer may also have limited control over authentication, field mapping, retries, branching, or the sequence in which tools are called. +Ask these questions before adopting an agent or template: -Sim exposes integrations as building blocks inside the workflow. That makes it possible to combine multiple systems, add conditional logic, transform data, and place human review between automated steps. +1. Can I inspect the instructions, tools, logic, and data flow? +2. Which models and external services does the agent use? +3. What data is stored, where is it processed, and how long is it retained? +4. Which permissions and credentials does the agent require? +5. Can I restrict actions by user, environment, or approval status? +6. Can I test the agent with representative inputs before deployment? +7. What happens when a model, API, connector, or publisher changes? +8. Can I export the workflow in a usable format? +9. Can I run or rebuild the agent outside the marketplace? +10. Who is responsible for maintenance and security updates? +11. Are versions documented, and can I roll back a change? +12. Does the agent expose logs, errors, and execution history? +13. Can I estimate operating costs before production use? +14. What support is available if the agent fails? +15. Can the publisher revoke access or remove the listing? -Because integration catalogs change, verify the current count and available services through [Sim](https://www.sim.ai) and the [Sim documentation](https://docs.sim.ai) before publication or procurement. +A large catalog has limited value if buyers cannot understand what an agent does or safely maintain it after installation. -## How can I deploy an AI agent built with Sim? +## How customizable is a marketplace AI agent? -Sim can deploy an AI agent through an API, hosted chat, or Model Context Protocol server. +A marketplace AI agent is only meaningfully customizable when buyers can change its logic, tools, permissions, prompts, data mappings, and failure behavior without recreating the agent elsewhere. -These deployment surfaces cover three common requirements: +Changing an agent’s name, prompt, or model does not necessarily provide control over the workflow. Buyers should determine whether customization includes: -- **API:** Embed the workflow in a product, internal tool, or backend service. -- **Hosted chat:** Give users a conversational interface without building the front end first. -- **MCP:** Expose agent capabilities to MCP-compatible clients and tools. See how to turn a [workflow into a reusable MCP tool](https://www.sim.ai/library/how-to-turn-a-workflow-into-a-reusable-mcp-tool). +- Adding or removing tools +- Editing branches and conditions +- Transforming data between steps +- Replacing proprietary components +- Changing models by task +- Adding human approval checkpoints +- Defining retries, timeouts, and fallback behavior +- Separating development and production environments +- Restricting credentials and actions +- Inspecting execution logs -A dedicated marketplace determines deployment at the listing level. Some marketplace agents may be available only through the marketplace's interface, while others may offer APIs or external integrations. Buyers should verify that deployment access is included rather than assuming every listed agent can be embedded elsewhere. +If those controls are unavailable, the product should be treated as a managed agent rather than an editable template. -## How does Sim compare with n8n for building AI agents from templates? +## Are AI marketplace agents portable? -Sim is focused on building and deploying AI agents with native templates and agent-specific deployment surfaces, while [n8n documents a broad workflow automation platform with AI nodes](https://docs.n8n.io/build/integrate-ai/) and [a template library](https://docs.n8n.io/build/ways-of-building-workflows/use-templates/). +AI marketplace agents are portable only when their logic, configuration, dependencies, and data can be exported or reconstructed outside the original marketplace. -n8n is the incumbent many buyers already consider for template-based automation. Its strongest use case is connecting applications through configurable workflows, especially for teams that want a mature general automation platform. +An export file is not sufficient if it depends on proprietary runtimes, unavailable connectors, or undocumented services. Buyers should examine portability at four levels: -The licensing distinction is important. As of September 2026, [Sim's official repository](https://github.com/simstudioai/sim) identifies Sim as Apache 2.0, an [OSI-approved open-source license](https://opensource.org/license/steward/apache-software-foundation). As of September 2026, n8n uses its [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), which is source-available rather than OSI-approved open source and restricts some commercial uses, including offering n8n to others as a hosted service. +- Data portability: Can inputs, outputs, logs, and stored knowledge be exported? +- Configuration portability: Can prompts, tool settings, schemas, and workflow definitions be exported? +- Runtime portability: Can the agent run in another environment without the original vendor’s hosted service? +- Operational portability: Can another team maintain the agent with documented dependencies and credentials? -For hosted billing, [n8n's published cloud plans are based on workflow executions](https://n8n.io/pricing/) as of September 2026; buyers should verify current plan details on the official pricing page. Marketplace agents do not share a standard billing unit because fees are determined by each marketplace or seller. +Open-source licensing can improve portability, but licensing alone does not guarantee an easy migration. Deployment documentation, standard interfaces, and access to the complete workflow also matter. Teams making deployment control a priority can also compare the [best self-hosted AI workflow automation platforms](https://www.sim.ai/library/best-self-hosted-ai-workflow-automation-platforms-2026). -Choose n8n when broad workflow automation and its existing template ecosystem are the priority. Choose Sim when the primary goal is to create an AI agent through Chat or a visual workflow, then deploy it through an API, hosted chat, or MCP. +## How should I govern agents from a marketplace or template library? -## What are the key facts about Sim, n8n, and dedicated AI agent marketplaces? +AI agent governance should apply the same security and operational controls to marketplace agents that an organization applies to internally built software. -Sim, n8n, and dedicated AI agent marketplaces differ most clearly in licensing, self-hosting, and billing structure. +At minimum, teams should establish: -- **Sim:** Sim is Apache 2.0 and self-hostable; self-hosted deployments do not carry a per-execution software license fee, although model, infrastructure, and external service costs still apply. -- **n8n:** [n8n can be self-hosted](https://docs.n8n.io/deploy/host-n8n/) under its [source-available Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), which is not OSI-approved; [n8n Cloud pricing is based on workflow executions](https://n8n.io/pricing/) as of September 2026. -- **Dedicated AI agent marketplaces:** Dedicated marketplaces do not have one common license, self-hosting model, or billing unit; those terms must be checked for every marketplace and agent listing. +- An owner for every deployed agent +- A record of the agent’s publisher, version, dependencies, and permissions +- Least-privilege credentials dedicated to the agent +- Test cases covering expected behavior and dangerous edge cases +- Human approval for consequential or irreversible actions +- Logging for inputs, outputs, tool calls, errors, and changes +- A review process for updates to templates and dependencies +- A rollback or shutdown procedure +- Rules for personal, confidential, and regulated data +- Periodic checks that the agent is still needed and performing correctly -These facts should be rechecked against official vendor sources whenever licensing or commercial terms affect a purchasing decision. Teams comparing license and hosting models can also review [open-source AI agent platforms](https://www.sim.ai/library/open-source-ai-agent-platforms). +Marketplace availability should not be interpreted as an independent security review. The buyer remains responsible for deciding whether an agent is appropriate for its data and operating environment. -## What are the risks of buying a pre-built AI agent from a marketplace? +## What are the key facts about Sim and n8n? -A dedicated AI agent marketplace can shorten discovery, but every listing introduces seller-specific questions about control, security, maintenance, and portability. +Sim and [n8n](https://docs.n8n.io/privacy-and-security/sustainable-use-license) both support teams building automations, but their licenses and product positioning should not be treated as identical. -Common risks include: +- Sim is an AI agent workflow builder released under the [OSI-approved Apache License 2.0](https://opensource.org/license/apache-2-0), and its software can be self-hosted without a vendor usage-billing unit; hosted-service pricing should be checked separately on [Sim’s current pricing page](https://www.sim.ai/pricing). +- As of September 2026, n8n uses the Sustainable Use License, which is source-available but is not an OSI-approved open-source license; current cloud billing terms should be confirmed on [n8n’s official pricing page](https://n8n.io/pricing/). +- [Zapier](https://zapier.com/apps) and [Make](https://www.make.com/en/integrations) are established automation incumbents that buyers may also evaluate when their primary requirement is connecting business applications rather than controlling an Apache-licensed agent-building stack. -- Limited visibility into prompts and workflow logic -- Unclear handling of customer data -- Dependency on the seller for fixes and model updates -- Restricted integrations or deployment options -- License terms that limit modification or redistribution -- Pricing that changes with usage, seats, tasks, or seller policies -- Difficulty moving the agent if the listing is removed +The [Sim repository and Apache 2.0 license](https://github.com/simstudioai/sim) provide the controlling source for Sim’s licensing terms. The [n8n Sustainable Use License documentation](https://docs.n8n.io/privacy-and-security/sustainable-use-license) provides the controlling source for n8n’s licensing conditions. -These risks do not make marketplaces unsuitable. They mean the buyer should evaluate the agent, seller, and marketplace as separate dependencies. +## Where does Sim fit between an AI agent marketplace and custom development? -## When is building an AI agent from scratch worth it? +Sim fits between fixed marketplace agents and fully custom development by giving teams a visual way to build and adapt agent workflows while retaining access to an Apache 2.0 codebase. -Sim makes building from scratch worthwhile when the agent must encode proprietary processes, connect to private systems, or meet deployment and governance requirements that a packaged listing cannot satisfy. +Sim is most relevant when a team wants the speed of a builder but does not want its core workflow confined to an opaque marketplace listing. Teams can use reusable workflow patterns as a starting point, then define the tools, data flow, model interactions, conditions, and controls required by their own process. -Building is usually justified when: +Sim will not eliminate the work of designing a reliable agent. Teams still need to choose appropriate models, protect credentials, test edge cases, monitor executions, and decide where human review is required. Sim’s role is to make the workflow easier to construct, inspect, and evolve than a code-only implementation while providing more control than a fixed agent listing. -- The workflow creates strategic differentiation -- The agent needs company-specific data or tools -- The process requires approvals, auditability, or custom error handling -- The agent will be embedded in a product -- The organization needs control over hosting and deployment -- Marketplace agents cannot support the required sequence of actions +## What is the best way to choose between a marketplace agent, a template, and Sim? -Building from scratch does not require starting from a blank workflow. A Sim template can provide the initial structure while preserving the ability to replace or extend every important component. +Sim is the better choice when control and adaptation are central requirements, while a marketplace agent or template is the better choice when the use case is standardized and speed is the overriding priority. -## What is the best AI agent builder? +Use this decision sequence: -Sim is a leading option for teams that want an open-source, visual AI agent builder with native templates, broad integrations, and API, hosted chat, and MCP deployment. +1. Define the outcome and the maximum acceptable consequence of failure. +2. Check whether a ready-made agent solves the use case without excessive permissions or workarounds. +3. Prefer an editable template when the basic process is standard but some adaptation is necessary. +4. Choose Sim or another flexible builder when the workflow must reflect proprietary systems or operating rules. +5. Choose custom code only when the required capability or scale cannot be supported responsibly by a builder. +6. Test the smallest safe version before expanding autonomy. +7. Reassess portability and ownership before the workflow becomes business critical. -The complete head-term comparison belongs in [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). This article instead answers the narrower decision between discovering a packaged agent in a marketplace and building a customizable agent from a template or blank workflow. +The best choice is the least complex approach that satisfies the team’s control, reliability, security, and portability requirements. -## Where can I find related AI agent comparisons? +## What related AI agent builder comparisons should I read? -Sim routes each related buyer intent to a focused guide so readers can evaluate the relevant product category without mixing marketplace, builder, and automation questions. +Sim’s category guide to [the best no-code AI agent builders in 2026](https://www.sim.ai/library/best-no-code-ai-agent-builders-2026) covers broader questions about choosing an agent builder. -- For the head-term builder comparison, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). -- For current Sim product capabilities, visit [Sim](https://www.sim.ai). -- For setup and deployment guidance, use the [Sim documentation](https://docs.sim.ai). +Use that guide for a category-wide shortlist. Use this buyer guide when the decision is specifically between acquiring a ready-made agent, adapting a template, or building a custom workflow. diff --git a/apps/sim/public/library/ai-agent-marketplace-vs-building-from-scratch/cover.jpg b/apps/sim/public/library/ai-agent-marketplace-vs-building-from-scratch/cover.jpg index e0a0e138be6e9ef8b4272c4ce8ce29770887f364..4ff549bb786913e798fba7827da4f790dd326f58 100644 GIT binary patch literal 30662 zcmeFYWpo_Pk~Y|4w#Z@zizQpk%obV9%w$0eESAO0%oa0R%*@PeF*8~GwO-nLXJ*fy zd-m-9n(FGVuB@!g49LicC#qf-UY7yF0O&tA7#L_cc+dw45di@a>HRxo(C0l7Iwm@3 zC&DAa!vTF6DM(2v7&$)Df8-G10907;9LP!tFj4?GDi{PR z*lP~}9{>gh27q|``Fn$bhJ=EJ0Drss{$Gm!y8OBXK!gATfTKX5fJ#cd|Gw)#Kf*p^ zpZ>?UgT;4ZI_RlS)0hDuu$|MDhLyfFwz)MVX^#YCJ7>0YJNp|}`Jayq+fv>$%`V)} zJ1G5U1A~!CUE3aE;E&D6{MMTWfGjSVO|mf3Eu{4b+szg(VX_#6Z|Q6$^)S4J`!S`S zA6HZ4^B_D*nfIR&!eqcx8PV)EjgH5sssw;w&QHZ_Q|KD%`Msqr)Kywj5VEV-(lc*8^gPz4Io7(sYD18m_Z4j%&e18;hEh$aAbo_ojtI!?eR?tQ_@mEBw4QDv<_>=tl0 z+TzdSg0~)~CV=5XvOby(igrSWWly`n>~;Ph=aEhgWIH$ISCV6vH3#D&zy?mk;`CPl z9i*kOMl_~pAEPPH{h}jLXIGU2rK0hbabfIDJ*>;)>}|7eGq|c!d@}Ot-3}^B;^>;NQ2Jh&kX#gk(wG*Ci*n&)pAC7(M*KfnUHK zQ^B-T%Uou!k)(_ShxOR}1l#(2h;-`?+*Yr24}hcH3tv?4<<~~C#HK6lrxSsHVG!6< z$3C!YwDQX4x~%^tlj&yd^P%4@h>{H7kh|s-|+@ylZr{!mcPni`1)mGUF)lloE z*T%Nkk{>(;TuDW>!wHE1p@aiRHZy<~5|(}AqifMucIs=t z>a*aq6;zDqE+Pdp%-ef!#dXhz=f?d4T1j|uhBh~m_%w7&_8UK7}J&S#Vph{de<`JRkU1HNwB(DbP zEqYPr7t{V6IFkC&G1fg~TJbb2#IaNgeO`>cabe`Z6%A`I@8YeLHrPM<{PKK=CQ^P_jiPL6Xv zmvzf@_e{!VKgZ^2y^tR8rUfMX{Cuv6$z!Jz%+dSF1b*EEmQ(1HQICELztmcmb7?c& zcHM`J5ysnmc&3%&XW>yO{}__PW#SdjmUHCG5V`NVlEX^(PVOBSDEf#Hq+WTt$%G=!=y4W zr!^kX#ZBk2IxMyZU1^(^r)I=$_B?d!iiNI>N&E}1U_{r#oO%FW{5g0_&fSIkb-o*S zFpA5FI_-x520kg0lD+{exZ z&)by|jqfurp>pq82loHFsrZ{L#E6E^{PEwT=Dq(e`L`pYQ;%MK*1r|~`w7g})u)V% zXqN_8)yLgcK?~c)ECATk{SAuZROU)O`59k^MNlx(^7-6MvHusEdvD0zxaeY=3Y2~jg?+@Q(*6?e* zShk>OufxbElyjT}g3n}JIync@sZQ)P4mj0+J-3n59KkoGpi@|)oCTX-wF>E|)pA-h zhRzL9iv`aP4jZ^cGME{En)dG6Kq$s5rwU5fI|2A>>x*}#@-^Tc0h z_}uX8mYmkbdoBI=p=Je}O8|_v)5dg%GZ2?Ym0Y!!Go5iaCAYlrOVikRDbKxWnEKcl zf~!v=a>ekWP%1*KbI)-x=OysJ+F{W~`ym&%YH|7d6rV|dD+KKW*rs|8k1J)UgrraF# zhwxf_?H`Nao4U$#?@FijBE=e>d_(uR(>_(10hiFn6F?Ej@vr(qYTIcOhdu^wgNTfv zF!fziwBbF=jnNVF9g&BBNPJi!xdT8>C9+Vqg55JsKhOe zWxpB!1W3d1!7VI+%a|`;R$S%(C_Vr18k?R}pI%&%52EftG<%jJY2o_cLj z`)~}A!I#V{2o!o#```C#dx)Qhfr>63<%P%K2oUFquyl7Png8yFN@SBcQL4BOT zng}$4q&ebbs)yS_1Y^nD4u;A35YO1P8_9i(X+AmFbEkcf(D>v+eyeKJ7at|YvCJPV z43oCc; zc@Q59-#UgU_tSR&4NQ?i<&+OpU%2z?lG*39mN;N0QfEBSd#my120>YRIL^1SiN7@w z91wTRsZ~+=;$_?qj2SpZPB?Xm<)sea-A1kpU}KwY_>lB_+^&i2B>%v+Key_MP@VC~ zv(2QffU=ZGOI(JJwOzLY_Q!$yqAlYhlKbz{?T7N7vjaZs{J5`yHCK)Mgmc0=fGA@N z1&FB@J}hZ*&-T_I<9WiTkW;{{XUya>AszigCc>Fhx38&GuKXtqg_$G#>O8DG5vpnm zZfERkkv*>fU3B{2Paloc-S|RG%$o=kP23A7<)S@3`3+I)_Xp^2;g$%!rGFOCR&RSi zrqt}ltmk&vNb1D!KC75ay8-UPRqUM*)mbHp3^>P zX%a}}h}aMlwie1|h-e4CeFaM5?;do&B*vbJ2=Jni%wu=<)rwk@<9TrJCA4oFxwAK<2Nwn}lN!qgPTxUW5NO0= z5^jE?vC9}yspU8cf>v<(N@#umAj~Fy1bj!mAIb8dIr>4-&1Ws+Yp%=@#T|OxP&BuL z*po9F?saLuE7qIJ9r4_QJQ^r&3zw^SmzVUoH?y-A1~NN{?^9|hc@QMghSJQzN^Umo`UHpJST-YG zn7whqVC&C%`sKE+KN@`xDaF=WVd~Ii*$v}(F}gZ!ka?7^8kr}9+ab?x+8k<<(*oLw zT@g_~X}q}GaA(9m4_J1<8Oza#wegSj)H`j!ADAf`rtSyb1W6sDT_NvBaz7A@(Ovk4 zn}%s2v(8;5Lrj{lW6wM5*j3;+q*qhuHb-POFxlx!3QDP(L$2eE+(Hn$@;}Zx2t%efj~kpoCtFr$3|o zqgN}K*C1Oq3U`;ov$x+3MNug}{`nPPoC;C9BXHjv`r-CKayEn2h5H}Fj57iiB=B-4 z&z1EIv3HG5m%c4jziiuSBNEW^j-Oc46=5LK+}Y=Y`U#BWr&?HJub)KJtt^6V^@736 z&(53H^FfZUe}x^24mQw>Ta_i$(2NEi4qZBiAH~>pvzh9n`k^JTR{3ZYY9}mbN_>vt300s#TfPjI3 zfPe%Cp_TwhFmMPc05l8=D=PYX3}O~GG)ydJ(J!9_^hij_ZONEEct>@B5KdSS?g<79 zcmtRh-mTn6f#TAMM}9~5VAhssp6m=w>pip z>PF6823{NN^vEq<1$wo3Wm+}UDP7Q^cn$KT&D9nbZA$LFBwR?clpAMVYjQe-yHmZ8 zh+lhc@m!rnN)naLl<^wbN><-2^UQkkYYq`jh9;u&n~o8@T)eo|tkV-$R(2A6)nrijTFhueK#yqsNt*l-ah#*zdgRLV_BU1`&D&4 zbY8TC=~~K*Ad68a;)Sa$hs8!NRYfc7Q(0SSU5te9%frxZUEYP*TOF14InB?%h4;u7 z(TT`juII1LWW31Sj10+bu8+$)i;NU-Kutflrq6na+ij7pP#L*lukW(EA!AwiGP9&2 z!&z^vjtB9Lwa0%HFS@b0B|0QJ=syVQlRk|c1uy;RRbvbKXCY-(cW8{ zx_AsIRPgqkK}mAs4SY1Ayy5UYK6(X|G%YR6$Du>!wKtM3HQ{*WfW&eB@f+n8bUZdm zj8d!&3Az+ZL`ixr<6@~o8=OJPCj|WsL`6%j#@C=Z?=Tr0yTWRo` zmbt!wSySD@J|OI4rha#Yn-#31xv}~j23CuGW0)(Auoyh1Anto(KyKbX^-&kaBi-v7 zf43H6Ot@3kgDNB>sAxEd@3NhWYgM&K7OMOKqug>S!sbn3zq@tvqG6NM6wv{N*=-#6 zl|@9AFMd7i%97?pWAMJ?py0&a<&9y*AK!3Zez#51X=K-?j)O;;8@_ZoQ!B4gJUda! zJuwLO)Bxhuy~4F?rtG@@FX{qIpmtH11#Cdlq!DbjXsl4>X&5=oC3zAAp#o*oa%8pJ z7LKyAvI;mQuGlbGv!DtqA!7N1RjIPZ(v8*vyLiqTz0&lUbSU|~VvYJNZ{boiW0{}6 z@J`zFhj^b4gxw!lRNe_&$yCJ~f23l|p{0l*RlAkw<4pj5PEdgxtd{6G{+db`(T+g? z(t*o%YS&&_ZQq@_{IC?T=Ta6JqZlUa?fAWft1zm95g7j&y$=gmHRu#6BtK%zE7;NH z+&1CHRB0LqibF!;QtfW0s+oSIkdlO$)wHEj-IC`Qb1pcx#)qzKHh7%T0pFA?Juj)2 zlqZ|6Z1g9CkjY$ONMrhk6plqis|em$D`U+xViPqyy&D%7T{}aSYul#VZm4Yr>Gic+ z7~Si}Gq0Doj?%9eg^GXHS~QQuFdYOfm0&>kDCE^GSe31vK7v=WwTVQ@SeiualtAA9 zrGB)usKq@|+17P#+^UV7;^fi1M1;)^sQ8m`ro)EIq{Tpglx5~)or;g()m)9Q0JW60 zX@dSI1E$;G8XTzO)onj(fi1G|`hpTMNM+V-2DgyT=c%lLNh^so=J78~8?oGMAczCZ zk*i*_@^bk*g^rfst>aUXE3aI^x5^8ZGK@Io7gG!D;nW>3dk2_L;A>zx?+1*HE<{;e z+B7(~g*CCahLkP@k#yo!=%f|lXUu}b)e@Dq6_Rp>hOF637y^HCn;DT$`?tx0bxDda4OrXYAy2{zffl1(f5ht();N{RCyUHq9|OD zs$tqQIFcbmF?=NQibfX2>RkOj_FIoouZz z{t%d=9Hj9IfT~n;F&6ZrFjszg_k+^OhV_M_%)g#>8JVdN&2fzQk?TdNm9SvKT;7ia zN|I*^on?A3OZAyCQSWK-=&3KMaGE6)8B1|vH5GC!<8<Rh^FC#O$xfz1uC=j!Mpf-sW}>;A03pqK54RU6F#s}} zLw$!SwflLoJ*c?FKGCf!!%I!gQMXKb29GXWw(oA* z)PtR#tGi>R7q9w(UDVNH-i*06!LO^OelTSy$Y4i(+VaZUFu8K!c4CI@cs`^(;_70} zlY6}1z4ReI!1nH@j1%v8cqhxj-ri<1H4*v2ye)(Hap)=U2!p3W@bbt|Oa^TzoKVH%L2KhVZ$G8utKt4A16!b*vw@%bC9$>hQ<>>CB-be`Og8l{63@cz8t#n1*zcB zQ`V8A)?HiO$jn;wF>A@#l31^*(Va;^uQ6S5%U^S6*-3Jc!r}r3+h@m$EI3M5wC2T?fAw=x03qm7~cI6Jd@2@$-*|9N{n{|s54ZPYOF(Iy>_GxH%ZtL zngpIj1K*mT--k&%)LQV4HAXw&;x2$Z_z$!C3%^U0w^jwYq!POi(rgYwOk%OtszfOo z5kvNB)7a18a=b{*6xmB-I=XG7O#?=pl0sC?SY(W-H3V$3Am^{eQjUQb;^hb7LO}vl zwBdW_1c|MFNIF7Ty~w|2O)6Q&(Xyy#P!#?2<4A2aVIg&JDu>7MG3WW?;sWqZ$+oF- z{CI>TnkpwkJ^iStH=pBY{wAQzJ! zt$ia~g!`1MZ5cQ%vExZf8;_cMvx?qrqFPHFmp$cD1 z3{o>uJa#-4ZbC9NVfhSYt>98V5mLSMogS5vvyLu9HurnZ&BNgDH#@!1lNc+_?0dJZvJUZ3K< z(jsfoA%*)yhsfr@qTXYL9%wPia2brA8ZS9wT9m;tHhF@K2_u=@R%M&0!XeD#*`5ri zdtBEv5rpiO9? zdTtmE8Ehl@`4as-cYV(qQ(bt(;D!r}xV!v@bEbvKSDo*xyKV=uR_@ltCu4@`F1aKf zwJkJpwmy|E*9L3TB?yo3*uaw6 z4zpOU;li2XPYZF*v7z*Dxa!7HnigM9a{`#!dU89n^m00Q+{=pm8$+l<9?@cn0~mh< zz_vo0&+Nu?GO1eB)*4qSN z4qrD=FjkiM+1+Ai-<97s8`MocVUcz^r+QFZ+bg8Vxop{?Es5xll&=^f0x3QuLl`68 zdao!=z{1qSpfM9_&kaL+{8>7Jtry>}RvtfOo*-`;`%V(?n>##>hjP=ZjAD!gT@kHy zbQ^us2WE21X}+G^p_DHNhAE6LnhJrA_>A$lA1JvsY|HhHK0F6Y#&?LS(L(u;S}yH2 zoG`@Ib6&jg5*>E1EPY9vZ$(8#MaQ3OEjJvVDuM@Q^i0q^dd<}}!T%sb@c?->tmVz> z2`RJ6)g`++gpoI1S-8p^HdZmDWb(Kx_1W-e-MU2vT1HM&OWU_*Mv6aF5-=pOv!02FWwW-F z*1Ec}b4luUPsXnwV~G^{4;=jR0n4`fDQd=!q2Qq7Bm$e?^s9eoobbFLrT_#g9^7r<%Eh! z|15R|6{BLPwguwuHgan|hYX`D%e(CZ+XA`&{P7xS0J##+sKay&;=QU3nt)q2i>tmG zOs0UF<5u|bgv5&rpvCX!M0(Y&o3>QGRQ+L*{m9{$V9Q~NeR&WJQS0tAJdJ!L{36;z z z0vLCV)25I#^?7t%t#C>fFie%YJ~z|o6-Q7Yx3VG5%aHaZF$FKR@f!~-|9-ZieJ<2+ zR3oLdZu>CQ2tDZm5v@V7g%Ly%;$)VNdT+|&tFj=6wU!EPaSNqB`zfY;D2XN~*ca3n zP7{m8mz|{a>febph0Dh0W1=!{#)K#3;VTu+5&*Y8aF6^4=fAG>pBBn6#t7f#cgFck zTYn)RAGf%NN-%!~kdNmrcoW#S(z ziy$SCg=ue_3V8XRQ@=Gs+9pgQJ1IF8U!xX&D(FsTIB;#OJAPigZYS4)>dJUu#r@|i zAVSq-_TrI2qoV%^Xi#_QI%sOSc29WM7OBLxlm5^inDdNqtvQt{#U&>DCJD#MlSXp_ zc%DU9j7*w`X?j?}9vkG(YLB{>3dWp#_{CR76)Xl@mMOQjCoAQ;8VKUP)Ymz5B>GTt z-TRNV=X>+R(A>pYJ3yJ*G_89g@|Jcq6Q(1oiMTAZ$~KWd&CntzNnltOQZ*o3 zfJ#FWcYVXBF3Nx_bM*rZR|5>^DZ{9EH>kQs)>^CQDZ9NG+YE;HwjkHbzG?q<(-#LC{f;hA(D2i0}NHz`VD7;{7OW zCKzuF<%X)1EiGrAg;d-Q7MeQfDx2O>bJ_@0yXGe(mpRaki_Kmle3Xrz$Qc76NkJjS zL2>b)6n)<#juFS5E^VL)uBZ#YoBSi7rCSg*_UEi@X7Py&5$ijBi>n0Sut~z`Tvk4MUBzd?%f22JL;R{7QZ!XzPe*a=5x67xWIV-C!%!W=LjE#2?R zWh6)u7m8%l@i!fo`@Q32txHtB^M+w__>jfeA_@$CGY*r2q(K++OJX*!j%Z5DSefJ^ zyFAKSwx8m9?^1nD%clVfQ5$VMP}54dtUYbKn`p5SKW`b8a~P6=cW9C%2{bmXb!qf# z{ZG7rI)mKn0j%tIuvSW6>;*VMYEg!Qec2s17&^31AC^YylJDAkLrx#qJ)0^1u$#L! zEgw0YANzkpckAxd?FuCSP`wk30baTI%n2ICK=WRJa0#c|ZzEPFGCkxESdCRxG!q4J zD&dOcd1YyTu;h}xiM?9v`JmDiWIHi_>Lm0=NKfq>#|3Dhrpj0zl?`?xkfV_i8?Bd3 z@%8czJeogF)qvK|-QkJSd(aT@%STJLZnbJbaQx1!CjC(XY7g?KnqkNrJfwJbOK~q0 zSrw;GkLcF0>O-UG+58Mmy!G^CKg3f%3f1fsJlQuV+)d&P)#T6bv}$4g8X~V_?VXLe z2*J-5@tR|~!_4ig8bvW0VH9Ql$7H2=4IH)d{28JGtDy0S3(zcaH!X&v6wK@J;FawK zUPuN?m5kGi>b9MCj%I+`xN}?B3k!YBZ@|=K^UWf&+&zIav&V)@cl#QD2b+h5RDUQ7 zQRH!T3mX6>n;8RDUf!CexI2CdQ7jmUR^c#8@jP~{u?kPMwn%3X!a-&v4_v&duAsz} z;3rx%-Fl>Z@-mJ-|6QweM*FikDs7Qc1hfr*&N}InA7U1+2|n7;pFAUXCF`6K1wa_Q z-7sIgTTpHms|o!GzXch2@thz(Kdh=krLf}rAzoj<|M$x<_4k;%N`dX+G6#hG>>=4; zJM_ecK=GF7fynado(k+VP?sq?(lfKCn1=m!dMV7Hk(jiqvho5&QWn@3RI$muK=eek zX*5*{C9-iPg8*$h{>bA>4lNF!j?FPdOb=tGOKtsfHiI(F-!MeaEx>d+IP9^fG zirLjr-7Nk?%VSX#Jv}1`lW{T8ZSa1Wtw#p8clp-6v7rHxa$MyTipJYwfrczs11W87 zGFTC3uYeKfrocM06H-DK{}k$tigjUkYx8Xzr7BjplIE`R0ZEg&qIDZXH13zE7>?cW zi$UujtvdG>yYnCQl*Ri6ooQ~_nnfzpN9}&82CFecW4+usL|3`theU`dy#ml2s%k<1 zLHPdD&C3uynF0&HVM7(Lc@_k&s}*Fxczl^Xt^3{>MpHdOt3PGOU)``QUclNVsoW55 z;SR?9fVXJ;P7Mx9G~D;3c;OYWzb?!&3sXnLTs?~0r_jlOb8o}u*JJAwiP4&-&`T;z zOBq6~9y2gt7^6CBXUsA!l|q`Lw1>|3TV7a3BT8jat*r;T`7!mm&;sOWalNedYOilY z#xS=P($u?w;m_eic_#>0eg}>m*9eox=vS+Qk}F^LnDQr&uej;{b)O8Q3pK#udypq{ zk+VZej~PW{(9sqZ<6$1Whvff77jA3|g}Eu2)VM?H;ii^xc9?u|z(Qw17BS%+9q}cy zf~GPtOr^B0$P#|u%lPl|;Zt8dgG&Zg^8wGPAq3C0~wG=9i+R43)5vny`TmE&t8bh3;s?*w>wKIajVi?Y1 z%zL4oSY7XuZbFc>I7eRD+Sywem<_MJdnu}HgtyxBRrWT~6nsa40mZzKOmQ;K8 zZG+pn4fv807R8hGl%{xxPI&f;EvOsuiconRtsuO5CsllFQea6tZ}4Vd+A;)#QSWgYSeu8 zR&eQEZ1A|J@@F$tW?gp$1?5t$3M#0U5b`J#(Q*P`gv=vNoFB0sj4QZkDgk1#NUr$e+1v=cakRDr)R+lE&x3F)2bmG$ojgad& zgB0Ws)HFRuk0P|=4&p3|T*vbm4UT=GdIhLa@vVo=*gUzxOpk;9k9iNxnrN#vF+ zixgZfMsV0+f$%_AiOz12B`kuGg_b%;Jl3*6U8U<1yT6US_IUPGWwVjsQ4 zfip)nlBJ5Mpals=?Ue=|tgXj)JKw%4WXP|pAGh7di!rHkGsHwuQZV91{_MTye&LVa z=NMFW8;L02dM7QWp;7s=xP057$Z}M<;V_(n6uag(H)uL-#c4*Mmnn8j@)(I{t3_jL=8uCdZLJ zF`Bs z)J!gqN;?=aasGk-(nr+J)sjczrR;K!IvTIWOC@%}^4`k{C+w`QEohPp5Gu$ zl?eo?$KsGz7}<5_>I~!Z&SB&$kFQ|8pX%6g7X)aB9RMOH&opsc$s;6;-0jU(Y|tjN zxzpbq`uvqS6(1!NkJ*a7RgD*BvmenXhQL4b4$=#>rgbGr5)7<4MZ_WYz+;54x#FPw za@$;+glg6a6%kkwwRqT^-tJkoSHR_xxs#sQ!4F(CN>Zx6a{VeF=65ffXKLU(Dyjte z+OXwnyFPYnrI}ElvhpOw=GZD&?4iw#h_hpabEB^$$*awjtsW?AV&j%RdRa(K)#bt` zB@n+ksuy9Ph=wg9%0eCo@6*!xv4MmXnMgsro4$KC6guuDTN7%t{MXO;%fg@W1eCAr zQ-}X{lEqMbVC8AaBU2>FV}(k>Ni)Wzl|)5FV@1wR;5=28A(YX2K=}HOCp3TPC;3Xd z4NTZ9Up9R;f*n8k16h^zjwB@-iQJCNNRv@Fm6)(aNoQ_Gl*292UQtmcZ9sb|nTYmN z+xZ2Q47nV-)1ayode0P}Qh;hFiy~v6h<8yVT}6ue2QOsbW4|jbj2w%OLHk2Ay_{zr4H^HupRF9NXUo z9!{hT2N#!(=1N~!OhE3ac{zMoo{7Br(X6uXyG;8@4RT^hml@>mPW z@6V&w%TX<8yQpHaUQX2v(nAKNXUs-KmD;j1!V<(T_M+4`GHCDlT0Q$Du4KgWTd++3 zOe0JZ*eHb#T=kVc#$O`ISeTmHs@3?3_@ZVx`nY_3c9Ncc5!W6z zQvOUUtKmv&S^M%&`shwB3L`~#0O}>6hv<7#r9Vr|6Ynyq=H-+1XxKp(%&L#Rh1uvQ zp#b2*{@O0@?P?P(-mnI8rve!69!k4L1ua*+7w>xc;G08T{j5w)?;@&d%43~C^r2z= zkA0N1twTzIQ1(p(mixhy%EMmDmmL?V%JS%gUzz{eCEiEV1nKi7XL$#Bpr;ooU`MEE zD~?N;1}6Nh_xa|}oCwNQ1;{)w+i$!AWY`nlTKnSgxm@clkpc1@p0VCsfC#{c`GiaN zgiM}!c0Q^V?e&1K<$)8b1n$BFQK(L-sIV3?SY0wn9kDYD96?^L zJ{%_(JVi9e2d%L{h>40?6f*?s4GJc~*e5!{x(SIS2}5y_C+Ym1-TPZSy&x<@=KX#U4S(hW?Y9vgt-oRfiqacE%N(RzrW&9A zxjo}--e{D6+i}4N_KW|!3d0kSj@fZ5Z%?cso9D6~;mOO(V{lq=3(TB2pd2+joMX|_!5VgG!P6L|a zpKI%=c6oD!xnr2-=zJO!tn14sP2D5FQb8ppWzmwh>H_8S$_qHpjKQ$=;phl6KU7Yj zB+r92`UqnJy2*~%d6>(2kUAh52Li4K(*rg@<4h49VFQBY&~y%Pzqg6pQZb}cr8pi0 z#@`CLjZ!2;0j^KvQS2^Jvvd}bz}~A-T&@hPT@FxC<4E*2q=Ue3xEs(K4l?dQN@?gE z_HPlE%Xz-ZzYhXWOpV9uV%_V)@$(ij3Zq0Zd_YC4Tg#`x>VP)C9xthIsQE`Rn%uoh zr(OgY)Zc(!7&Xf2MwrdQJ$qw;489T6#tTTE z0T#~rXZ_N>diJ1f51+b23gs}jHSA9JnW=v4Uyn=eG`YK%D-ocTB)+W#DvmU-0MWNa zvHK;(rr@vSyXH1SObRvLcbjQpNO^Q|_s7Yw1lpL=AwJ{k+{MVeS3uuF{Ze2Rrc~Lg zKhmcx)^fL$xue;!lA`x6{0;T9N7l1bEu*-bP|5@kCtFxmdb2h`N+Kl_Tj8S#LsU5d zY~q+w?-hG~=haQD-?V8oI+iSGFKpasZsG^UY3SXY=~L{kiyB)=*u1MzXu%z7b8&f3 zwHrybt-fE-amUwwl6tx}0wGpQGOBQVZM%zxsoE~RfPLDNi>X)q9oKD%@eIm}2_tW< zV@G1S=$JvLC-097E07ANVP@Z+Fj=mtzSI)!xN5N3 zw9OPeXb|!(^L^bCXi&rd*Px`L(PE{rVJdxVj3HBLg8T45?a{!*eeQrsjStsz$|>Ag zFsl=89D`dr*;1_%p07D9TO@NT46>;bv1m;)>QIPTCh12`DCw^o6`M5twaN6bXH<^#? zhJ-l=(K@ix*NE9>@MtLZ7K?Ox6eLMTAYGP6*!|%~CKZIgC zyC&b0ePR|6{$k*hU89#XwP_1vA@}zE=V4$te$d0fXEY)qPzI<#2uwV+6oJSz1p!u5 zhVdu(FU+G!z=Q`n86JLBIpo=ysAsjJIgx0cenzELRc_yF?ETrkpI9b$^m4>_($;E) zI}keNQ(S#ju^4C_$Ff$5P+U;oTRGj(w8z4Bxkf){Tu5UGG?FjCuS&DN0&XD`$%7o3 zKIxy~Fiwc9sMzAIYW@gx;)>&FZ+`Zm3qq4VS6HU+R$Gy~onvsZlQBSvuSP1sCRx`hZYG^g+fV@v8f~q z$quj2Y!U6-Vnhcvy8O^2xN{*2x_+(qM%id0i{)sIdqN&f_mpmSw;?&=N4of9wFKN2hX1WCf>NSH4K9WZ!>brNmAW^o|C~?|xk9D(o#*(@~!BxskpI zCmQt*dsgC6N)}Zz$^@Q1??PUhS(<*b(KZrmuyR11q^UTn-_zn3SOUG7m?o@`F0kb2 zGf?(@4dI)dC~!ssSL|?xzl{WP(EO9$aUHSz$mz~f!_YeFg-Fc6yRg*i!BvHp48amcbErt4?FE(SB7U+7(l@X9jRL@ zBNNh1NM=g)TSly=ziSAbqX?p(bXDfh&N)J3=bwjeh`gRXg}3m{aC;2HW;0o?_%)ev z!7d_L(hWX&hHNK@z)^gowoce6l|JpnvAV6xq^Dd*XWRLLx`#05bbCKHhRJ76nxm>% zrUs5gERv3Z-Iqrzb{Ol^iqq9-s3mrRpC&qZ^s?g`+5XZ-0%{h?fU(bFJjErmEIAwi z;-zdAHovHsn7>b!%r`}mI|W?7Tkbd4D0)a!H;I= zs&tk?Yh3u=+!(!3|a+qE25vuRC>UD$6$yi;` zu6wQGW|9>HU#FUgkYihd;Rz>YFHC$*5-k-GYrd1jcX%zIC7J65_?|635EvhB0YHoy zX&&?ZdMj{!QvU0>$i(`QncJQC0c%!|ZRgKQkq`Uo0ZH4qlf^PGP7@z!Q&XDXYrO)V zykd4~>ns*9$#39$`Da-_Ymu4M&J@8cZ-1884k?SPi=I_0c*xIjh{u>XY^rRy%pD1) z2U;Gn&CtGo1t`sOnu?AQ7ikl&FYg+eDhwOZXE~}+zZ4bQp9ZH1Dm1a$-a`)+t4DPF zSv$n5*n(ZPb{j%a;6o8V3Q= z>qpk~(WOaoBQLd6ZQ^vRi=1hMU_0#Pv5oeuIrOcA3m!y22}qE>rp*SqsGE*L)tN+6 zMt^+@7#MGDH@)^BbI6uz6*X8W08C;#F>AsVN1CztKH_lh!^T$@RnWQfoW(Gc|kdMWZ(Us(I z3J?br>29))W}ku(w3HWS5-a3iuJDnhNWHVQ8ODD~+QDo@+roQ}GHIO!T{m5Z9lv zR$#=5vE(0ej}UD>$`IUoPgg=hydd_7&n#_ZlA=jXS_uig^+F4wnmfFp(SX4`LbobA_+>N;us87Z}=GHE|RNd4xhb{ zpFiXU;9&soF{M?fC?_K?SVLkIy>cDi{U|i*<%)PuvF|z;Cms}rbjjb+`o720rPfgq zea=R#FV8DzvXR5=v1kE#FjoG6<&sI{zE^Nw@qM7rdo}?sr|q6W)b>d(WVIP{6&?+Q zC{~W8H{(;T&V9)n?Fd$u3QCv$@b#e@1YaNdX3tHAB%R-*l)5boR57`Z%3zy}SJC=s zKr-xrOcqBZVG0_FNLa20|vm_rSQY{cYTId!ShhX~xv%SL|>D#)@?b^;MtZ2s8`E;Iv4N z`YbL$*~oWSLwTILvZBzgEj~)FfdyAl_T4HSeLSN5w6tAa)V$`#TdEDpER}nngsI*T zWIYWAOkO6Dey-tKd%Ps0Z>_!(_qZK8V&nGCi>&F~_WI8K>R11#ulJ5>Y74(bLy_Jg zARPkIl_tG+g7g}ihu%AeqV(PZgeD*eN(&H>Dn(R;P^HEg0YQ2ZP^r>AH{iMNci($s z+eP|;%Bc=Y!Ddf0ZX~E^Dfb`wY;;o zNFzuAXv96R;AbqRTOpXLMMu9&`fJQUQK|@P zJ}1n8J0yUa8^ge^RUg#!dMktA7X#~oyAM{Q=81prw6gPAe?~F{5xO<>{CV)bcOyWe z->>Kyna951UZWDfUo<@;FWfNP$^WuL;c8Z89sG>V3&^-OL)E=8@-0YMvXfAi)c9@P z{My6(I|d5Mj2fB<@5B%9!PtA01|ug3RD+y|OK%MtKJ>pTZoSGf99e+^x?| z2Lu=^e#L0~FSIz%UDmDm{69Z zksh|Do*_@^h?QFAv4xhp!7>`*7vmEKlgZM&WK@Bl{toF2t~injIVnES>Umx?fjp1> z@r6OTOMdT;6wcFE1<6lnPYeA@owE$V3biWPC^=0h@Y%F*Ml}b1<&6k&y%!!gT(j`C{fN9@IQaXLcF0!>@P94rySr^#d z=;nU`j~B9jtNgNU{xfuSGo79A(2|n$6PL}bC7TtQ5X0aTo!q}}#hI&xl$_tOE$xZO zgEQ6LSoc?Y*36S_Q>`m@?~fW28^fLxjSSa7<-mIJ_v3BiYqf+Gnx}|zz?E54HTk~E z>u>!svf8v?D7&4>x&Hk4Q8eV*9|M{Ps}H1K&XWm}b&Fe?A!(6jKV*ZVJabJ_dv}Pc zzM;@{?qB5xVC^0aIBM1T&+!}1-RN)sXaAo@e9OzYoc}tXTW=6_chPP``LbtLVLYJq z6)M3MBZfXTG^$Ap;Ur5t{udC`J35lXO){V}o=J^UGgT7t%Fj}~@9j)^%8t5DiTV=fC>(wbJBg-ecK|NlJ_j#^}s zI+VQYh>ebw%D)cymR&9)L?3#U;+~tQah2a5CT!jzYL=aZsjNau+y~3{k%p41MyH=f z_eLw}*67$D$20`$vLXYrJgU3plg!%KLAb#FcjjNQxj)Wo$(v}ZbX{|4QUDNQ`I+}? zF1_h9PCxyTN*#>Ai7Z%hLhTQw)wqdzw?VyTG^1*S0kr@0WH7Sc_Hg`FbveHcoES1f znc3(TO8g;e79_+TR=`_*9+Q5teL?JMr@&F(R>hqGOr>g~YI$sE=*yH~e8_9?{o+BT z@w?yOg7)*aKvmSNFv3vwWZwES6NkxiG0mCBJNThP4rx>03Cq^yc8!Kiua(vn_`%?y&^`~z{Mc&Sa<3@x=cgot_)W=ShW9ew! z{)yf~^p!I&addY!(8Q?^VM0jMxpz$RxEMwqN6 zQ!&c>Nv*Y|rWIadAraFjyY$fMmJIz+<@sylVO2D2eCy?JC`EB}uKSe=Oz?o4$UE{D zi!U~aTsBMY!f2EBZytX}I>DR!#d5KAiepfR^*47gt*j5UO#Mev#NmuDCA@b) zyIik?Z&CLVTMqT8yJ;F8x(?dTIqJwp6|s-`%*x{=wmkh0#2>6yu1197z!y^GPq0zA zx__w+#@CJqUl4tQgsmr*DiFqF(=2kpBvr0~1~FdtN5? zz$wqyYG*iwsX~1_|xzJ+MVOn?5YDjP5 zm^5|c+_$4jOu)1p5OkxI7-{)k_tv|#I?l&29tBR1aw~I9Hb$I#P;=OS0lI6O5~r|0 z&@VYTrk^8|oV{HP8AEajSG@8tBVMCHZK24O4jsBJS3oTs`FQBREXJAZ&4zo8=lzVh z?@8*>4M!9u`-+s;Ig7e;03r5VMG5ZHYn;b__TJd*x*5Mv;&Y{2Na|C zx;ZcqQjnoL%0^BTc@}={ab!X)C^D<84PR2l32pElzVNL-P2=tR4SwqNV+Z(**OcF; zrK!sOLvwcr9S-QMbB7I9tD9rLE4?9&j;wQsPrMF-lzAdXmucpSI2(9k9=Vum-pMgc zjMU-#?qmT4@OqOKtp^?0FQkjN`bv%G_Mn;IM?x^S>H=o-Y`q*@H1I3=z2&o5SP^Uj zPlB~(bAA)MsNjEfk^hlDN>e4*aJ|B@7K}En@ucAH{i} zw|Gf(YC5Q@yyc0n=hbR$o1}xutTxb*$PjvPr7Z}O{DE32|0ZfS)ze0iXfc1bF@Xq#Qf zxXf0f#7`{_K`>|AotAWpR|1+5OSCavHKdRsC9MfRcvxW8=LRtd6`+){mLQU^8+jv^3tlB^5~X@&Xt zw`5d~GDtrm*jmGEn;)}87!9hB-7~3_U^@&B0$fNq7Cd--xArL_4YiWz(_Pk5gHnXQ z4%(F1xJ=G=C6W^A%D5%Xqb$g%^(O9K^=iBzieiKB!$9(A{OqD!`m>V6K9M-JLLWao zPG4pIRn9~&Ei(B;fEptg`W^cuZ^U8(Ryjx>KdpaLPwD~pgjau(E29-*J+((|pbZI^ zdqa;LC;fg_Uq*uNhi8iMdBcQ0*1;qUr)3#Y%oa#k39@=ja9`#2r4ZOgieDx|4n^=m zk6E{w(2iF3G(-`87)5VebLahW8f>3R&XnvDlo&JoB0n(&v=k2(A>7lPQyQam7T1I# zE(E$@q+DHHXn#ALe+%k+At)-$JnsCZ*Za{LdD9?i`Drp+E%SSlasYlWMUqL+=EjB? zN9lVqAA{+5@=mN$5y+ay(an;HsIr*+3dsF*6Ed$s8878cfzWML?oe{UvLS2i;3Zfk z=)F77r6-IhDkEZs@dzRDh%}x4uC)~Mz2XJAyZ-`^rpnwyLTdjc7z6Meh#(X}r9qm( z;0AOaWg=cJU)dR$?7dJb4-6`lj(6!v4zx9_uzzayiKaJ>WJ(Rl28mJ5wf|>M%kf1* zSyk1HIr2d6jU;O>xVSv}Xd9XnRklCjQ@YXR_$Pm&Uk_T?JI3#6GvxU7HM2szr2-&#iWxw<_b-5bfpTM7$nBT)d6tLe_sw5MrA0j(i(gBk59tv*(zRwQK%ofr zDx;LBj2I#F(?l^IkC%WsFWfg8oqXJpLfR>h8S~nGG6(JH3x+hRJ}td6n*aF~?g;fU z6KLU&W+SW}OjIRX7-p&w@>@zdNan2(iAwpJp`ju!SZeK4lxXjB@;`g9EGPHzglq#H z>at#Wm7e*Ic5VO5`ztM6@+7n2$19sGVF_m4#kPo1?2OVM&}jNGJ#uBGlwz2Fk4 za0ASEN;GIRC1;T@`GC@?L)|~Y9J!sC8BES8`>iv5g;^H1qHHAnv@a<5=}(4r&E^(H zmVuo|e5YD^Xn_!)4MzEBr`1;p#(aX3he)3Y%w3c5<9xP7>zhAc2syFQm75lEke}r* zpGoW>)$bE&(NPR=c#f*(ZW+UjjrE;d^p;kDFwU_~E4yp^g7 z?9P2YJ+W_2DiP34F?30($4(XBZrG=I4U`lZ;$;1nr%ZEo1eTTkT`JtU3}xo+0wG|a zyzp7W$0CEZWf{3>$Z+^;Iu`{c-*Kv1NEU?0 zYhA0@RXS8TuB;8a)qTW8x&T?_TDDeo@S1X7AyI!O(R2?_`0)V$=euGO;M;Cow!^ql zwraq=Mi4HMuoyajf$-Onp4(waw9zsYIC1ruRKx4Ez@)5LWBhMesqWw> zx{g2ZGZ0DBBRWtjSq*FbrN@txl7{`ZpMCwEtF7^pS35uUym-_y(9b1bc*d3< zI`z)E-WKt#(J z^j;eVUG;U9{2|CK$mycb5jnwz;lnDi-!(ZM)U^bR21&poef*& z((h3Z{?A8zf{4{5qYMyNL(-Ap;v5Nr!C132NtV+gahVI{Zo@AN0G% z7j0N+IlD;wDp5U)ylk~8HD+FYMZbGpy{+)irj>iIs?C9}th&QrA`}X{zu5c>;4rp` zCFG#-J~5@C`i!M+GkdJ6U+S!3(gdsngC7o0y%(}4lPNr(C@6p6@A%_iKtudhkyryr zTcm~iFm9I%dtdwOomE&m-q%)*w zPyCkx7A6QhxZenKfzpeqfB5C1mhC2@@kTy^?NWy;RA_#VJ*LSg;IHA}0z778O9uCA zEi*h)Pi47ieKRf~UMnFxNAr)+*=abF(Sh7J4?jpTQ~p#Z_ahjWvm2v5iB8R#%M#1D zr`_{=K@YBV#1=E-ZgUaO)*~inszSc7A*)J-odUnzH2i_v4_o_!$ce|I zz3&wp8+SQTRTNhJI2i8GF-1zpmt)(V-fb_5l`-ufPlJ9L&Dn!s`Mh`+lrL(L2e9w* zi=f#)ckowVex8?`tZkI#R3k{jQDQWC$-Cpuk(3<%xeP6c6FfMc5?_nc|2p}_q3C(| zB+W$Gyt0WRG+vhW4rUNia}_=5BxYP5J2JMp$5zsoKYUYW{1-o1aFCaGo-+MYc+_sm z`bA=}+$Ayp-8(`2{5V&`ZbxoWvWq5XQ^QB0g)!&7K|wtKV$#9Hx^2PNIKz)G1f7 zCg{)R9+}%xbXy3)n}f^LxF<|D#HV<_KFhsDk+9c7@hj3!i&tGWVNnJQL=LNM3#lE* zG5*2<#g|DGVxu31SKl?O7KEGVSj#D=)BdbJ%ydap*5L_1tW6|BpC1IFtw2GK7Z0QY znNSjetSAo;7jceuxxTjoRaOKo{zj#2Y0QS&th-xIurRPM;RVtE>ygjJ)gbVe;JlwV z0GGztbFrG&dH1Gey?t3yey8Qhi+#dHBUxOTSLt6W3DX*ksrYL%TDW`e$o-1~<$0J+ zt90mnsbQ^uW%KMBCB7ZS%!1ad78rW)hk&cq@P|y(7ALn|YiKxI26QKSa5mii*9{3b zMeZV6#_4#Jk-n*amO7)KLmzK)w)F*~=r(r^^I60;69%Nk269CzY$*A;m$k1=A22MJ zws}l3P^y~JGN={s42Dca_Fr=@up)@q&P}wsjwsHv=IJys{Nq8~w0BOjqn-tJCuv<` zc3m^}Xy$Z(v*Zndyc{7uf_~E8QBsn+y=+sr;$V*fH&4+Gx`FzKW(4I=`;vfDURp(K zu)-uSM_On!j8c~c*UtZ%<4Mv{Lq; z14a|&sq*^qyGX+e{{@|Znua}(%*&iuZuPwh zHg0!!vSZD-BHc@4BmmiYTe2`yO?i2^8IqR_Rwi+`mvB;kM-_JMp|#%2HrXBXs3QGX zFX}Wy=p^l>JgY4{oyTvrlMODv?=H1ooqThzqGY}H^Fs>&BSu8+`pnwT&+l;MWBkCw z6v23|MN-G?FQ1b>jY%3)UVT8J;VrAo`YE+JNPX*$qpSMLH;?}=E-ASn1&P%sFd&4- zLGib>u6u0C(*!U6UT+{#|A52eF2YiV@(Qy($q_jVW!t=)qp3|Vyl@O5xV|qF!XYwl z9uuLvkiwOJHXKN|fmD~0%=x9kYCN&BYDjI6a(WU_CKBsNiFw~`xoG5 z1P+*v|MG0jsoXq27O%BY1b{TBfNw5X$#2S$ok?mwn>dTO_&)i1qHfSJ1q{rp>WxSV zZrJLa`iSvnVyJC-P+`d-j&`vbX!NHD*gI}USjv7eZ3I_nuqCEG*-OKX(wlpH3}!41 z&rhZy2xgwi?ndZD5b$G0Kj1);`auF_TKVejSBY~;(T-PFE|nF4+()#CPdEd{@npDA zD+fSNXRO`lFX{y7Pyh15<9U#>!ji{vsB6dp;&A@pSR48NREvP8*3$GIiIl`fslr&Y zm1}*_rc&bM-nR~6It7!wee-m}2%$;^j2g|8nIHTK{{l8W88~t^Y5(NH)n*J>vsPCl zQBB<_s>TAO^GnW!!C4r+*FN%+Tn{CG(*N-}G zQ9U5e6}!b1KyORV*Zt;goK6V27NYbLy-H~pZOTK1O zzVl&i576xU>W$!2Dt&W!-czw#fIR}+7czaq*Tt4!!XgRMJclu5$PM7lb(No@ds2FE z*86Pp>t4fDvIYT;58V7&wj`ecPuiXs)+t6JrKJP;1;^k9t9j)4h*m3G&dMQgwxnM3 z=)B?kOm8bVVYhj9t1dM$0eU@LwPpwP{V8iDJG;->vNcz9Khm8nNhS0r7Kg3!y4w*u z2vUKdQ=`?R$<9)aVzD7KYo{%8o)CMGfk-r-u^L$bK6yKs{5xj)4-k-$dyYXrpVlYK zCgIWL!qCd*NTA-HQfj9R0DX$vplD&maI7UJIIqE>X>manf4`m$ORz%N_MUZ;{UVa?EA76#2F1%>T8mHWS?n0Iotpeek03`N51haAlDQbSdIfJ z@F1%FdCzuDO#6sF+(z;S@ss1$2r85hfN5P9!fZ|id(!G8w_VKqB+_Ljh5dtg$n{=s z$|QPm!C&>1R_>J)ipZRpi@(TJEx&WB6(A@Thl@@mD`SN(c|-3U5tOsU37Eypy0`n& zy&`exNQ%vT#1!@n@|`%rltRDX_TluaRUaVc2*%;o zkw|R-O3_J!9VSeT~M{;6CCD9OcO8Hz*i%NfQmQpS(N*|fg)r)8MTYd zNv%_qiayI4_N1V3>@5kA9Ct_;cf{wry;P&9r9Uz_5ftqsnE6)^hBZ*s+hnw&9;|d` z)}lsFm$Gys2Vt4wO3Lye(AkNDH|TE^mHnS_4N^L_owmuk0vE*nj=q+0RsHQTZR&KT2&-1Cljm?B3Z>cO zzbOU$@%@d5){ay+g3}=d{_E|nu)`r*A>C-xPE8hmxm7Ay}M`yeTW<3m&fpbXlO{8q4XC9%P8C9I;r2ul$xCh zUui$>6({L_P1`oV$kf+BttL{|O9e`I%TDnUbRq1t=!bP&9u~H~VRQ-uqYb!3OPtax zGJ?jMmeP|f>_uz9CN1Tp(ct!0Xfdu*NSsXp`0`~l7inLKlFKaT+{(x6B#yZd0xxe- zmb{lQ5OB7@3oaA@K(R`8yBSPpX_k=G;&y%bxy?o1DNlAgP)Ad& zK&h6JTAF0e$n#K(TAq~A$nnsCQ(ngEVY^Jl*$bkpE4eO=-_P~Jqvoy`-{ny?t(9vl z555Vb;h%|ZSxs+cN&%?b)&L?!Ew+w<)t#E}SG|s$VI{``Z*SCgRj?gqIwu#ODOQ{E zhEGWyhWM1S9-5PH;fSgulPPt*cl35)JsP}@>p`K}P*A-0aY=^8sHV`9$A7G4v;k)? zsub+ovETb`Uyw^`4k$S5t;vYB4*Wzb>kRWzsA0G7HOv8%=xbQ$OAo%7$ctf-WM$9w zAD|XTA8|Az&~;95{|fVX4;-Wg4P<=-irdmOIkL0Q)H+-)@_Edu#?0Vf82j!GbekDK zRP55RDBON>Qs#yeO(=!yb_hPAaE8Jn9zcd)^ zm^r4h>+rRp7gv>W`oLqvD))ZItKt-;$~`=zv-^s8N|M;mJPPbSgBu;np~PlxH17_< zMM5;pew2-(NGr)`mbEqKSGIqoM~2z1>Cd$T92I{+efuLSy-zq*G~rK-wSLu?+RJTu z{?ygqS+b!fosz(Zqv0=;id;$LbJ`!!^P}#nk5$G=(;A)i^^^XT*;`|D%Zl*pQrs;8 zKN$!v?9mXt0U_Fhtrc4QLWDGZ4dIf4+gz4wtI98Sf&$BokGo_i5syDOacd+vu;|X> z$?d%E+z(4K^eLT{o7yX8lDauQU*e8Y@-vj&=IG#OKBk6*I|iq_k%Jg`-R><4F~Gwb zUnEbIi^*Z8S*a7Ys;Q7&)16byo=E3>$T5zV|8Ls(<#5a$hbYa+c*W0=8zD)L#!W2M zXP_5DL6~{0xZGgvU9u_&z4;PO%pE^V3K~C>M2Y_`IKqjN&nW1^emdpuDgonz;!jc} zKYf$X+gddf#F3$eD_j`JpXO+mfa&feOO_OsHXA!LYH2R$qjo*CngYcy^AN2t~5~-!;Pd{BqHQW&a3^ZiXIiAL>NOQhAE@bSnRz0zhP!hyII<2PHMTg|C3kxVhoHb( zybFlcu-a*aI3}?Aa_yeIHYTE}IhIq*u=o)y6qo`w-mCr9_%{b<4ugkj1>V^!$*DmJ zVxGsZ!Kj&G-Ba_|Q95Azv9_8UEs@`+vmvgQ;jg&0Ngn1JoHcXjERJc_pS09*#2DtP z_mbn~riz9&1~a9&{xz+N^O=V#DK>dYrfN9;+3}s@e71DOvvm{$T=fm@qpqdST2oE= zA`*wlmp9Yy%Ne2;Opz+OpmcVd(FJdCh~>76G=6iO423G)+0$)##JubHQ%0Q^y`^`} zS_c$1{gY38aSq*Kv~MXsKy9?Q$3#3T3M6(>CSjPl#XW#<_sIOo7Go%Mzv z_n1sL?NjsIpqkgZ+P_pTE}b%hi)mtrT811LmZvciO&B#K1s(lS8V&>_PBPaHC>nqx zMSbO{w>KU=Y=~~Vv(dEZ#4=As`$-*T{x9Ge-v<9-Hg(|UV4SWI`&UKG4b$a) zzo5VO%Y!R9;u(oMdxz~`&v;c}dbpduZHd=9_&T+X)7hK|&^Kr)6Yap#!fm}T?Nnn^ zj``oSZE=U}He{EpfI&5^ME*jEbTv0ol|m`0++&2$AMMSqYQmcvWT5V%om%10-OJ0G z1r-(=E$WqS*3c|;J*fpXl+oJ}AoggygEB)k=M<$l+;Eb>rxWemvf*<-E_2wU&hsi$ z9?{Nw17A-X$dO<8qbf zE7B~zqp$j$xMP2d0~1`F>%hJhwOVrSZX(yjKpwI)-I}Uh?RW7i?gDack+TdE^7#&7 zYK)**@f>}l-tRcT{ZneY%#!!5RI8aTq|#bi!UyuL;3sd$GI}zx7Azyx6Ru1lb3@>d z93i2liVqqup(A1*U4leGmg{Fy zTX2#%iNF;c-8ZMaTaq671y_hDW?qvTIA;+4ANESTy)R&kh?DRO%YK|)ne`7U7I7!0xbebo8x0>(PZ+O0cm0g_N#Kt(Y&Cf~?4=G_^D4F%hD$h~ch??K1%2HtX@ z-1U+Qo43aG2q-qb<|5x++#%sSi+3q?(|<0GbLev@x=x8u#dCzLyV9p2l)SRIpG&t{ zT%mN|ARj_saLc2}7~fvI3ITMo2+)PfW73#HRUc{<#G&D zFCz)czjBk@lAo0LZq8Y6_Iphg-O3HfmJ*XnG-|N11+$SLrDkFaMAor2W`c8+@=iQU zOeDCK;^%vGnnfYdSXmoO_P!0ACLA;_Nu864w-!&hX_s=J13i^a({t;#;E9yqL!Oxg zEMB%RJkGo?gtOa#^oRVo02u&43Bb!GE3;{*(GCX|jIpY+1Ch1(O|{4MCJCWsvVae5 z5Gslk7E2c{I-+W#>;FISO9j%5YO zFXaHdc#kGF4t=$gh7;voCiM90?!IyhIOI^t+ZN*%VTqC= zfmLQpWGu}AwkpT-$VEn8B(DfoplZ~{{DUpmSCrBN50D&3J)RH&PG7MyIw@9b`WBTd nnds{+Y-#^zXWPHqPc+?Z@PLnprvd<%{a?<&ZHjwy#_azB#)zZe literal 28873 zcmeFYWpo@(k^tCZX0X5(TFlH$7Be$5wwNtuX117_nVFeoNfuaWvBlT^X1=+(-M4pd z?%um!Th(1%os|$7k-=H-tM3~C5dic*4j33{IC#*BgouEEh=hrZ0y;4XF|aW}A3}U$ zd|c3%iJXLloQacxj)7B@k55!hQO&_28qeSJ=;{i;_{TNO_J)3h;#U<{3h17!nO)& z+Qr!};wcL0|7;l1A>LN;cCTU@{2Ur;02pqf+TbO1ViF#en1@3$S%b1R8P+`S);vX( zrRj3OL=|n_6d&DxItW_ec#aN<#vj1M?{0)D*Mh`FGKg;M^X4w0W8GnSzyC)dv+)?> zB>MkHP5!s?|6<^OG4Q___xO+~3KkjHFFJen<}2z65_F5p9xqO|{}7fp>G)nA7-tR(^Q`{RT^r)Rc>g z-lv#}zPIeNEQ(*$R;;2M3vbbZ6=czcrJ5&7a|+$fpoAxrs69KoTr^u?Zn7Dj zAV%>^_joW1O<1k`T=2R$vaG6RLQ>2Q!KG&PzfBX;H6`t5WP3uc=+suEkbWUSTU|xM z8w5T%w5r<>sU_MAQkS!H6tb~u{Du>ad9n|1uJebpDWCmKo557srO4i7SqV&#p5n1& zh4x{MfQcdx&e24F>`&vHTi&*gLXXz^#$_=gXI(JcCCGVv)8*jPFM4r_6hirP_z3G? zdt%&HtuHV%Ym>#Ks>jQ`r-FAO8@`y@c1NVD&@=4W9wlR401tL0ZtJe5vgxKiU)yIl zQEG;$h-RxVBYEp4C7(jGGs%jFZEY?_=P0|%0%w2d6&*7KDoSWOUpe|N8*9#X{&Y{E zMZB8Yf>W4~IazS&WM^v{lK4h3FcyHw22Ja6fc=uOM%sH5$GK=e=NVB{3f-0+9Q{M~v* zW$No#9?{0??2gkQ&=V3)49#+Mw&r;^oj)mCOTceiYw927G}fOAuM}55QFbyKIW`== z9U-h1(t8>meP+^BQ7NG-pMT@(zYi80`;h%r;uBo7nx(WGeP=_A<4--i#8iFOTMpjT z%Fx`0;ss@_c0}r@zM4q#k~S)m>-iBYvv#Y)1{O zO^6G(!fX;b#ogLS4;iRY+td`ieZpe1M%}&qVPE{Mn)7BagGgPv{wv4;z>9#M$kM?H z001$0>m6H?n+JixF!}oYNGLBL_@_3?49n6FK0SSjzgdFOSJwZ}gFgqv#Oy1m3hC0# zEn$v%DazOT--HH)6bm!60EnB(J6~V_zpDMYFam4k+_)K!c(=o1%11n zAwyz23d9qa<#C8pg9+VXvK$iU?zyYQs33#?1RK`C(9=Xh?5>tw`m^lCx zIPO&J%##j=MXB0eJ)K=ed`6(fCw4Z5mIjvQhnbzMhC;LkNBgG zSPGb%wCa_H#{V^Q9IjOgo~^~0vt}YKT}#>qqxr;KZi&ox(SO#b%FIY7Hwe*fOpa4? zU(JSA%*cuQh5$gF$VhLC0j{6a%%|LTdaVAGdNLNt%{)a{0%iqz<=~8 z_2-{n3TiNN*||>#zm8F49eZA$Q87)gI8*3JTWMR>^Vf1ScN6*bwCqk}`|`TCe?=>4 z=SDOfJ2}3Zdk5$|0&|m+Qf(|6busc-Gt!+!gMQmylF(^v$2)Z`>BM0;MQ$CjUo<^% zoubrwY))=&;2Y=ESPWJkUE1TjSNv+H$ewu1s&_HVnTr??lG2U=K8ONcmQ~Z&@#Dg$ ztR%}copqU9F$b-r8%eqK5{nUF@FhBFq)&odA}$01XNM^Ez5681d7TA#@%1MPc=Io} zt&c@a{>C0yV0HkoM;w?vFYU~ECEVS8-0}O!F8=U2DdDE5MC}u<%;w@43na4(RP(sg z@57C>6eTq=S}mdwoyDn1AP7OZ;sl-lkS1V$OTo#Vg~cJP>dMW3qi@!W&ur3~-E-g{ zBzZl&NY0OpG}n;-lEt&9U|aD=b!hPJAq!o?^Z%k*tt_Y3^}oKV+%LM>CA_EME3|VK zInpYf^LVq$6Dv`Fdc^tT%4C#_kABS*m@ukiQxC@Q$hreAuv?PmwUdN$Bb;<3XU49R zZ&03AC$*2E@H-RoW1nTPe4-7awgQvzm9i!=b`&x%;j(ku6Du;ZFW2#j<*HzF=Fw3i=(V)A@Z18Deb*kn=^^TA6F8u zy~C#BhUO#lj6(cnU>-qzb&I4otCW%R6-%t-nB>=l4T)2q@&*`Utd>8b=k_cn;nWKa z4b7Q{YIEz+dDSXgyyUNH8-u~p>aRlMwoKw1nTA>eHFeli+wcOqvm~_*Jz}I*b^+;a zAtBu@LqKBy9(TqaURRzlF^9#=D}09w2mJaNs%&GD3ifZYQ~Xs0mQ#T7yn5-#C_MGy z%-?hT_g@GUMs*za($L9&FZ_SF2P0r4il5L3ngCj3>!zY%VxLiZ7yS zHm6?tr;k4eDGwfE((ehM8OauQJ;|(LUEB^IvK%=Qx&Ddx=*INO{xF^fCEHHfQpv{_)+T1NEK>izawlFE zdi)nvq!AMpGc|9n5g~C)xl#Y@Y2~jNwZojrDtlwv|2D~g6c7bnYJ)d?5wo6k{#kzm zvVN*EGS+Bh&1N)gS?ANhynC1;@Yag$jZ&JeGFj7STYC7hy9?|~S)FihdjoVkpZ#*3 z#r9{48W_T)dohP*N^sw;9L#tU|ol~BMGM02rCh|x`^_hWtGg1sHrQW zc!%Y-^RYu+L%NoW9sQrK8q<>>a)m6U?!V95jo(vQ8R5SJ6#j>4qO}nnwFv%%_Z#2a z&R9h%3s6g$Mt0P)reg>skMH7_BE{Q#pgFam9_0`_mI?%a!s8Qt{$5VEDf?A)Z>JqR zzpsa1_G1qr5t!Vj)^$e{V}q>ir&(Vziqf`bNbx>+3Hezq@S+80HJnY4Cq$V#amyzf zDW!M72lCs+;}V=93eQt^6XDvf+NmNjyBObxmY4D7^QR}0WEe3+%!`DBt&Fhq5UzO< z+Y>3L{Y|s@apU|tDRXJ33na`^>ei`^J)7SDVPgM37yk*Dai9Xg!NI}6Az;BE!5|jdAZwubJ#ikL>N|jNuuJ*E|g1spoxh3s#GE)Gu*fWyI$2%I>Ma#k z8vX)1c4b1YV4z`%)PP3C#YHjT^p>A~`|0hwTQ!NcFD}aK<8&+X^0C8YMGBzi*QGEi z8Y5^OWX1F^2z501H|ZR)Vs#3}A&jBBrZi`=)T1I*Z38@rX)3eIjVRO}oZ{nETqaH* zVnt>!!g%y#kh(-n_~p1q{B~S*lT`Lb!QKI6_a4CN;godY%rBP(PcK54zfPd0qwNxU z4I!>K)T-O8=$MpMy13g^eNwyEZd6}=TZ%hWsLO|T3GA`@WGb+Nlg52@SD@#c z;?RJSt^kX7K(|c?W!X?1#GCZljn;Co9PP*FcYuv_hJ+8{imggje}ZP{?L^W6ZJ3>| ziZp7p-Gc${Fhl7ofsRZ~{lf;oRZq?GEy~&lc{G8Bn}ME7!pmp+acV~s5S|jAd6%2M zDM!Kx(Eo+GJMIjX3&Yx1H6C44ZBxaFqdv2=Oq~GSZ)qQ#%9QH9OX>m|l7Ru!6FWAK zX@*eK8M85{3z6KTVTi4>ujEy7vcY$@jGDjqD=huT2ex^FsdPI zWno!$pFc6yGTFY_bwV$!LK*z)BY19n3^C5g$F@AVn+4AdE6C$l`Qo3~Py)|OL)Cvvg)N>SdUeQxXqsB$A z;N3s};`6c|S};+`ox?1E4O8--5WAJr`$0kPw@B)g3Y;qPS7uP-ctkw(FMlvv)u1l+ z3%Qt9#Ts^E+vwBt5CgtIQTCHy<}%OuLA`dCDGn%>ZXuW(m3zDcq<7Bv=BV@$g*}2w zqS*)LbfT1g_T0O)32mQrXLk?GDibgTk;@#RmV2r;Q*+>qvqYH_9@};BwanYs+eH#s z)j~tp4bOi!@Lt{i?ZZ+$!dCsP`J!ENdhQh1wd9AiV&44r>9SrVqvi~&=V$6NAx4v9 zmg1~owCdJkq<7l9#K!TkH#JqY_6; zDgt)NL#A2XG@aq5c8zx(#JU>RJuVpREyQE%B>47e4HZwQ;V-2h(UN_tF^K9qQ?kcp zGUzI1iQ2~<-tse7I_a0IakEggbEUK6sFc$T(5{*vjFHziYRV6=wT(1x#$QgpI-1Ue zfubEXPH_j#Fy+{jN`0D}XS=w7W+L#`}E{ZpBz&4EL%eGh9x z7DJmpxcg)&JJux>uW_vHL!L{YuvcYLxda#TzPIa2jGO$*`T>kr`OijMry7canv%}- zYVBSyL^H-1GPjOCBU#V2^GwKtOGl)=Ez1!kwKW>6N?diL)$@?MB5#kI;=Z90652L0 z(9MWjexDd*j%!yS^Fy0z7dx|{t07|=>LlGiG2O+E)QH8BFoACef-8rnZM%lo%vrrc zkAW&SH51%9U?gMv)P3+K9>I1MidH%aKddqwO84M?wCJ-VVndrg6Yn9y#fR@HEE!1Y z)}1CwKb!f?HRABaL+u*Omc+Jpr=WU`6SS-DO=*X~vs@xG`Zv6cOwvwSnruK+8RY07 zlE!gup5igh@9){a&qN7TX5-8++;%h-M?(Zhm1tUd;s2Q>Az1s+zvcb-H{nC^@T3^W zkd2suACuVL0f=R?HkV6p+SENrBm?u>rj|g@hKw;NG-Xx1-Q9AsJ8D!E)aj|x>ij{^ zj&NVRtc@yjV=ILb)g{4`E8pJsCA84^!>3}EWis^))uiq6QG}j{&u#K5UZEkSp@oWq z8bSAX-t()b-tOL$zqFxPylRxEq}2~*67RMpaAKG~I_#Kiv48e{Rf=x?iQ=gp1+LTK zwZ~rF5E;aeUsT!c0$St)CarKzA6(8uVbXD`7RoX8aLjF&6=Xklr&JCuchyXKm~U2c zNeFK3ntlxwsX-i+v&*e%8b^LmovJV$(zwn@R<+&Y@`tXj(h4@H->9mbtq$n>RzkKg zr$@9*nMyz$-J+q2_3DXdn>62o`tJY=qkSZX1*$JWnSA8Ea^VJ#oj7=3xt{&}R% z=A0liP4Ouc9Ib+ag=_KWQaCuS-G*X?qmifTv@;&s zZ~>~V>RKr6d~~#zmGnEn1?bc^szFlWu$6P0G>_=A$WLaM#?r;Ys4VD^Cuxmfif#QJ zGe%FEr#n+YhWwAg+mIz5|@3a|nHHL*-HH zV)<)$XL%Ak+t2QIB)iAjKUWeg&#T18&jQhhmKvd|lXtC0U&MEb%-zmJ51V$ZxwS7! z3%I_Dw`ec(b?+mub>U7$fJi$()20EGM}(>=O1p7>^{J*FQ#PQ54I&O_#t>*P#MTzVu|c}#hT(%Wdn^}oA`+YxvCGq&tUQh2LMi5#K`E-~;M*{p)vf+%~R zgpW&p233c^C}Xav>R%ZNX^Q5+ZsU!x+O5diF`U1iZOg>!`K_sNX3RG#|3vF&&MN3c zSIAFtz`$~f>H`gw%e;LJjHlq5QO-@TQ9G*M3(aq1-T%gSm_M$xopXmhtHljY@EzaR z3rc?2`b(p>WFt~sWAT7<3aNiVPYaw81F=C>#;mFS)#OKOzzAtn{BoK};KF4DQ-T25 z*~do)gWX=$7=sTQwZEg88&ci@Ys-Hm-4lBxO4w!-=eH{DH9BZDP_?5mFBn>y%nIX$ zmTwv!V~kNRf^vFR$omS-NV-Y+!+#0tIXk@ZVd5OJ-&T%Gg_+TGe2zsnR+yi?(jP&Xw>Y7)QdL;PZ5rER>?Rc5SwO1)>Q~epDrqja-C`qE1q8MWf$)5TQeG5 zik9#_Jv|CMJ~#cPpK z8?Bd2<^%$XGzrq5ox~^Q`HL=IoT?j$MJ5M61B7gn;_Yo>i~|4_7Oo|;_Mg0%wpOgP z&%mji08DZ1v0uLj2Vw(Xd~XbYT79piq`sTl9NJvynRShiu%ShRhzPf$ zshdwPv2A|D#>k4nxU>?cN&KQ?F|_N(A69M9a05Px&P8W;yxVbiSPoXa`4YQJRCL)p zbr=~dK9vWr&Q(?!YG{eD4Y)@5uQ4!DlFu~S_e=?ueV7J_@ByREOJ5z9eZ~heYEJLE zIQ~}09dpn1zssKv7Juf3lR0W2jIRaKQJyP%pe=7#11Nh_HjuJM>S=$rUrF^=uG{bz zsR;#y`BOovKJW|;0QQO!+B71`_UyG32|mNQRn(&}MimZ3nR6|bvB-pYpZ`tqwF(Mr zC4(>+Lu_$FW52rItEkS`qMwJxpLgSIkV4Y;G}S2AT(cdTY-Ed0T=BVW%nB}lRgTJApg#HGZ#coo=6dAwy<&klnTK>1MLtdY~60pL&waaB^B_#NG zSnOm`hwT~7D!rZHj(Yt*K^E9cv*sAb^doW?`SSNNY5 z=PXGX>=Y`5>v02Yt(;{OL=#}?v>LHA+Jn)?m8RGL0sWf3yb;UNvSw;aRG6F^DDcEt zqG-O-E59?oiW$spo zC#jMGxv|(yv-X+;$}9MtR#F=(zwbjuo3zD%-eaZccltCp@HkSIor z2ZxV)$O;nGrsyd~6{L+#u}8O=xnqZa;WAQEBtyjpuYqMsLC3|LFI~$qN3~cq%pqf2 z))f+^c||_Oyhn^LrlGPf=e?AeJGaW{S4wZlTk!JGtyWQ!Vu5)=S`wsBNEoUws4vf| z9j(5AKtwb~3jTI;VfEB()e!pe9Y8F-cH+1CBY-7^j8YgPh7{!mENdTU8h8Yg;Lf@`pVQ~n(gRf1!@L5rz0%|uv0h6&7)Vu`vsk1B!nyY3UC-A8+~=j6Q!ocVJZ zoyNF-?wKc;ZqqcQz-rRGi9&da<>Vc9q&4Pzdij3M994H7%aSF5JMCa_#5v3@OJs-?Ylw0M@{!V&|cY*W=AZ|HaE=F|aE6@n>;*I&dQd&lz_{g+St zn_=OW0~;_Lu#>4X-*$MmQ&N~#2R8}~`d)x$tC|+e;rc4J7#NF^c%S~j((y149N1~F zy#1rB^DbTVl|e9w_8`fh+JL43x{))-nP!Q)SPX~zvdU0}E!0(+B%|mr#^h_Y@))(J zG3>}S=#X@9qt-Jj_c_QMABIhl(?USW3K-N-zVmFQv+xcuu09qQm?e}M%bPwrM zH04FI9k}Nqu=v!pt&NujhIwwixg_=7Am|9KUpaM3>m7jCAVeg1whvsnVQ`CRt?2Jd zKE4=|)#r`&Kfi^}As#-rjzg6hPhCREp{bhN8L%&66wC0eJQ_pcjr)}V{*9^#1kdm% zMInPpDnErACRIB=H&yMdq0$S_m!h-rSn^*lL++<1>M$_2wjt>Q(JnhZrIooU^Ut;r z7=uhv`<%lL+jHVs*UVIniXFU&K5orscweDe!;y!V$lw7`hJr+tL5`!QoeL;15E<=w zSjQ_#hQr&kXphm-z?FRlGqkR<`V^uX!j|gF(O-KUY$>8?3wBT0xIb*jPnHq-l4TsK zxiAklU9O~=_A>eK69=J2s>@HMQJWu$%(FkUork1zo2VPAt++!N+VRvXuS_r+;ivR) zc`-LwbJAkj{UNljsS8{A4HMmjepCYisY!4W3+r3=qAbSJ z_9=sLfccL*F5i@@Ynycf_asFI%--ov%po*KgQ_g_Y5wYQh%5Y-KkkmV*+6EIJb9Mr zra&VYkd41qx{za_XovS`8um*ZZ3I0SgzAz(^y+Gbs=sJsj^(Hx98mG9nrmBoTsoN7*ScyM?ZQn1thPJ(cwxuzK2hx!W>N+T3OQ z4iH&fPOcr?C~V>5>E|V-oMuXLqYEpa*exqgbcOC!CC~I_dVV3!@bAKP?peh;`O`3e zyVONBE_b->5kDf_94#8;;>`x$r9O&>GGm-&g>-H}Ryt%2J&?vLVR&M$X zx%6F_Y>+WTF>0T3{1sMhqrLX_zb(42fR)8U;x$?hR7WSL9L;Q5o%&~Hm*usTZ+fg{ zspMQe>M?CUy?^y`Od&C*TlGVAm<(N`Z zwi1n}Y8qp|w2zzQi$~*xG?;q$utq#C`u(+@2AwI26lY`*>RU3Jb;grOZJUtMKFGn3 zat~;I9Y1s|l4k0GW?BHvJIicCLUwF&w54>r?0g0u1Ki_lMb(*AczBtJGfm^m@x}S* zgAxoY^&AGBMq(z?XZnS(I@!tO0FJF= zc6G469a*DI*cLs2p+uT~nGrq`KK0nF8 zjBH2@!`-3dT}@G+aN}{1dbpf|n`ei*tbOa2{9`5z>A8X=hjsB)^*dRP-Ro`EEE=5L zY2x8U?Mlbwx7}^8Zmy1d5+Eq1m9>VSpppU|5INSIqD4)OnF8NC&6H)18xj#?gToaC zsM2x>O@f8)E@DEuGdPi@<5!+wDXU2FD{DA!{sa4AfgykOs2SJL^j<9a^M)17)#_zlgDE}JMk$FZ&t#E|bN5XS5CS80$Ibi@?%X3`; zg#;JdO5R3vNBoew|Acrr);Of}W|aGpg^6M4Z07i|ObTXDWf3X#2>xCzMGWR;r~ed_ zAu(noBOi;hBo);uZwBE;u#N{ZoPl)VANRe^;tngGq3mVf0S#4u3C=EaN3pJ1fuU~a=Ys#%HBhK|;`VnLh z?cu>93r6>bgbRfozv*bsuf3oP@g>qvdV2CPUq+PXdZAfhK|B;zCnj(a0x1 zp54bc)#85Jeeyk-;rfe=zaGxkop+x5H-B9|k8t=7%315)H)Lj+()o1m#W-RO=^QL< zL_NEMr1l)cfN~WGL2u=Mm|fj=Ww`!V?lj&dJKZ2MrAo!DOfxclX}0!EwXKFZZVWJL z-c~h=q!Vw=b+A4@p=M7%q=W-y^~~enDKpW?PocJ#79-W>NULJo6dOQLJYiKL$ee|5 zQqAlt`|nl%3lXQr^DQ)|O07GX%2*gDGw)<1(vi2;*2nnq6#gqO1&JzSh>beVkRf zu2{F`-NLwS3C`iFrZ_IM9{QixLTnk#yrpDq+q_908}?>oMU=fj{X!mYB4ZX`toKZ!!AZ+e;pc@#`FQRH8ZMt$n(BTr?xPD}%3bMxKhspcq6^LxP@n8p&;3AePkEgS!2C=+>XY%?5|2 zW=NRqimV#uUs(<3sRbKaSqIy#y2?pwMnb?cVU^qh>yQlbXj|qpDOZqJ|JEXW;EH=Z zU(pWtHB%P0ZZcfU*7X2}a78;7eP0;Xxzst8^cp-(zBJkp zd7M5oOw%yaM`Ijo^=?v`Y3}P0xpb`6gW@K4FCGj8EyW;9bluD?*Pj(`(@5+Y(@lqA zN4hcHuFmJx3?S?&lVv;=s?Kn5n^g|RE2}HDXG#zL9%hO%{UkT|3pRY2?VPKA$f~#+XGUF;_&PTb`Lej9ezD-PVm3#rEBVQrtQ2UbZjhCWTisj7hM%)^=ilT((Z34 zclqChRT_ir+bs63;3qXqy!Pr5L#KT(>&NDa5liu;hPpN(gUMPcAgatpQE{7_IYEnY@g%yH1u=bo14 zB-p#P2%Pte_QJg1@hL(1GSg4RYvoqIYO4sgtc|TQLf`Tq%%xJ-=~+7Tax_N=TF=0z zSu*&>ywHWr1!md3xRz8t_G<#UNBuuLYt=^ZS9RfK&8B>4F0;$7te#G3%D=n5^CfyS z_4^}@Zhi;fTRrBxwY~B%_%s~GuZFI;7Trl3jbh%LmU$yZ1HCx8BfkvAtR>kCep4N=Z&&8%|%0IA_bs%Y$hM(9c zquV0;!iW9Sq>0 zcdJjE2;t7;5_W*n37!#qsl~%*x$W+qvlWK^b*B88NSOB6uyv{>%gdND=6?dF>ATBo zw7ayfk&)=qyiiN^3xh)yXMT7edwp%ZG|E|$)y&=w+788_mfGhBgKI84EDDq+Qe3ST zE7&#bH|YtoYH77Jjr8uRA9zGda2g{gG|I_STzZMhGK>}+o81%p-vJo)_s^R(8yAX6 zj@(|scB%sio!3hfSjDy+G!l}ORaepsUfTz12NvPP*g2AP#dUs;bU2n8)XUsDX)@Xx zX6+s9=CsiKfWYrp*>tRUwLYT{I@=4~+&XQ@dlj zREzgP;YY8yHy@@xJ$13`J)2LF)}ep`y?+tUd)>xVTV@P8f5TR)Xfw9?7HC=r zMpTUu`o^M7$|-$Sr6S=~@QtaYwYmtVg}Ye|Bm0iy#p{=9I~D&56GWOk0a~o_N1$<6 zGAcH$%>#lq@fV!35~tm?U}t3{Pk#l0W4c8(zVius3zv}l18u9~vW@Y{V-$=TM>zMu z$K)4KE~M+veuycEPL>(en_N>-kNjcgjuhbh2~+S;p>eS}*UI9s-5)cfXJzT(_+ui>Gi}+Ya3=ps^_IW& zX)l2M)j!#x!}F}Ypw%HnW1jL(>EcVA;d_sf(6*UGO9Fwcu_LlBIq5G81-xW zR?!RDD1&xSdj@y7)W2+eHb@oU4H63?`2Dc2Z(kd2m8#=WrPVhacnxVd$C+cu2<$Ga z57L5=z&zxJhCTLB4%PoM@UJkQ1z3@P&2S9<@LXDpX)%VMaFLTSM;~Sh)d7c3FQ~Pw z@0;QC;3Agsr|?4TKUu+@OGmSGfQ<`;Jrvez`B0f{#Vuh(2uf73dC&1|K6X-(Q+2M+ z_e4yHZp4kUAVcC#V+o67jP>X}M*~43>&dPA0r{`o7RT+u2Xr7FTijwDbJ5|H3HN$P zQF5<%3@QRxZ*nj3ICRJso1>x0Z)+tU@dFr>^3%?xG4s6_w#>=-j;wfBz^s9RIp;7- z)`i=yr-G0yTcq(CEX2+;_vW8G(@&7c_b0OhG?SB8;-{n>OtLgHs1S51TJnVCKW(Vo(y9F_QbT*-d-AEW&4}n98w1{2gIML|#sIn~q*ni$bYeZb`Pj1%4=a zy{HVhade(^qTGo?wR<*pq&bl3w6uo#-fHQ2BR&H7c0|6`PA!#!|0Tz!fRbE8CAKEOcK_IFoAUa zDTp?ng+ci<&|)9p<%ArSg3vpyn$OVO=x&0+!d z$rk-_yv`bEpGL2?#4Lkl{O_xNH;5(d8{Hj*N~JWe`1N`-Za}n_&X$ZJi&Z)~d|hMA zpg@V&rNdF^PP*azdM);GvWVsqpTe;4__OWxUReIdjAkyRxzaLx(v{HzdxY>3ztRwm zG~VXK++!*af-qv6&>D4nd%`_@vKsweN!bzAmHL)6*LdaTdK+SN#M6i^ihr4y6ujt5 z+8-Cjs;5@%_D{YNvZw0T7)to5I0JWd4K=1eX z6(v)u)y6xZvE@^M$O9)TkHtzWoofVU_#y7b-3S3gjiPH-HNP36UCY~1$5{Nc8&moz zW_pphBK?>ce&babCWC6zMFLMqAKG(PQcj~a!Ir-MZ|Vi&$G><06cxxl1_mL(35G6H zYD@!}tlEE@UWyu_8lpu;tL5E(BTufZk#frDwCkei>`|%GbrK!J`cZ&8teD zhVn`z@AQ{NwdH3Y;xa1@+HuolT_yu*AH*fKR(ZB$YXx!Tgf3(NGjG_=f2y5XOtQB_U=V@a_5L?*KtU(6&Zt(6&Y}aBxU?XlPhyFi6nbzo0FS05nK+RAMGYC=3!IB^Af7 zb@j~3hEDMbM8d}3`uZ_R83hdz8|GL<6pRAqfe!xp1)l@|wbc<$0JN8GTO}M5evsq` zqK3VZB{;uP0$2m{#XDeD>1&QnDD#eLcuDA^^RIURePQUMQHdtHnAgXM;*++v9>bVu z9a|w+`+&u3u+HJ7h!_R-1MU%vbylN%jZfhdWPS1mwHzf3M$&Cd=`jmRC4;%lkrneb{%ai1wP#Vg1KR?al-w1%48C_-Nwrq#K|&BStL@V@S|H6X^_w&%vfme6!h}PD(YjutzoI z5c80WJeGBmePk+04O~raB|S`MdoC?<8ShJGCgaAcmD0)6H&+hoEp)%wFmK~%5sN6o z%B393L*HdHR-jQ)77VEoV&MVSiOG|3+$0hBhPZ*z?4Ubv#+^M3UgBBk~?m_ICk5IcM-#0 zV2_15I5IhReyOZ%R6FP+K3QB(2Bih-h88B?DO2ujtOS z#`u_3_~ml&&n)nF!|h0i+}sES=b-$rPw3GbKk-!z=tXfx-v%|0lJsZIv{{rW$I5fG zz&rJO3k_<>Z1Gv>k z&JhVAXP=%QHMKAAfRQqfqe&79ql5L~hAG~e^~qs>5S`3HbSm#@>UXqrZIqzcs?Wt7 zaCH0)-%oX{`eA$kN@j9^LC< zN^DzTT%=hX*H|2b2!)?%pTtrBCio8Y7l#ygU9-T`kBb&lxh z&^NWnH#*q4Q*5viU$SV3Y}R82iovKX2R)=s*9+gU`bGmRRYCo&zXL@50d3^1Y*Nr% zl5qhN{k0Vcus21y?AEq~ZM7<5w3DuY{2wcD1GijB8Q&Y=WUF@UF+b;uH^ z^_P+Qf`0L|XQYm*(QG@no|NRft7pE0*5f$4KzQBqA2Ur{-- ze=A!!NI>R=KVL*EoI+%$=w$uU24>hLSBxp71_Yw~fcgS*WhJ#8EJs z!sc>j=mR~i94YT{XK^=huglrl-#+OpPnW~r0O5aSA#@*#;Wj| z9b!hJG?VeAR~&5Ad!$F|3MH*oY;($H;1N2GDh4@H*+Td~5$#FK?oqg^rniBj%QGT#Gf z;&3RQy{s|6ino1cqKmhjKn2j+z}=q|O6tmHm@NlxLD)J(q8R0W-4$bEx(NX1aMbcD zTNGL^0jiu1qR*L*c1I%N6uv3Vb+B5EG)4?jT&E=KqZ&j_u|x}vUk6kYfXH8re6wXQ7rTAk@37o=H7IuA1|w?2Z2ACC#95tQz+EJ{FQES;cI zn?hor@1sRp5X z@u)MO2zqPO2)ISgCiXJHB#~j6K_~Qcm^-KvJc~W)F z*pXvTCh=*nVEoRQ>5=u}+h&LK!iNpbq9y7p`%C(9YjQH?cfgcObx4B|9K!|O$Gm{) zS4@0agP@T0ZE><_D%Dt>+Jr9%U6OMr%Kd3qIA%{$4- zruJdQv_UZ8G4mm5>5P*YO~_5vRgcjh+ar0-VGArv4_uFc#4r~n3>cK|Y3Jh}(FBrW zNl1fMY=PZGh6joKZnsO4&?|F&JFoSA>&;+>1t126p+N}^c-zmzv`I^%0*U73cBG|3 zV!olpM%u@H555Nuw$orrOYo2Wnc$nf%ixv-5@gx7WD}nXt3ZIzw}z_5C>4>kCWWC2KI62AuIaz&461 z7AyQ!8|A@AS%YT}nn07~$WK-!&erGSr+2^x#cQdisvyKFtJxkL31?rkm+j|^O7P+; znW(;CCig6}@Q0cQ%oX@f8$Cpte=Jj3>h#~udjt50n@j#ljR^o9r=itzGwo#@d9pUS zAE*3%UyPvz6C|9(yMh$COx2^im9|Nu6DZd<>coClU$Z^KFcS1Brjt3~h)TlygI+^K zBudy!Cq3k!({S_H_bw#*&Gr~<$}$kznZkj%uW-pKlQ=_kX~qsvp$W^1(dKW0af%rD zjN@N8JXR*T*&i-x7>Bmb746A9T+M@{FY#CxhkUq>;#g4_-{V=Iu|{jyfx>ygYV$i4 zE!NrEN;}W{+N-QO{G;Cw&Ww38%YrZbzS)HIa zS6&8V3^-8kpnvSnpI>&tA#03Ak!(+{hZhdM1U}DCQiM;pD1eC<*PBmJw64d*wC0Fc zHxS0o+NoTozYgXN z0XF5*tmX@?6c0{w0COWf(gRvS2BZK~fzx2(x^)nls2F@XJAfwWuu^1vyiLf?P#f`! zFLZm$A=}A^sRBDmUNIN@Nt%Bt5L|ce&u5)RZ7f0@eQ^tZvEX4HrUWFx3;b>m%9=I0 zpFDT-1?Ek4#lAKg+}|wD^>T;@4%KL6iHv50o6E6OiYxJuR%{HC))fgFd*PZrdNoQhh=)O=B{M%;qImTu=?i!q9$e*J{zr|9r~YzIi2iQQcgOKxA-*(ZIck zf)Qhe?}ftW5Y#3<1kCcry9r(XO3$BstUgY=_HcA2gtZ?vwN0Ab!r4C#=>YH}zN20CV{YFEc! zb^A+h)bhv36_#6x0wt0mj^rJWouqIyE{F~>cMY?dSw2M!7_!0F`{;SYqo(16UAg35`*Kc@EFLgXZ+|Mh4o+$sW|;*+q` zvTUJ+!SJVu%@iW7cS-!u^~?exX3P~cK{nyXP*}B#)1V1Vk|t^J>lVTdY}KJ zukVhAtBu|ry=4SZXY?6m5YdKc(R(DiA&52%6TO5;)M&%#EkrPgHW)QSB+-c$L>MKa zN0bOjL?n?rzVBPVd)Hle-G9$|&sqCDXSe6s`|LMv!5c+J0yKC%P$ z(1%RYuxXXN_cf|sm*{5Eq-ODk{?s$&^#kXO?NG{s3o$vuD-Gk4qkZas0V}6iI4;;d zo3JYoR(d^q#J(?rFSh=a*zJeryAuhQ7y>c1$P(BgPdK3*A+-HLENGem#gl@@_svya z!Uc*BXBjZv4U|sizNO!{Sg9?8NQKbYzlWvh7Hd?5K>j#x3@ zi%kFEqaD9&z*rvc=HirU9={Lqn(kcnz5HjW9JO?45z(lVv5s1(xFd`cRXq}+%2*@n zu!4dlIsoRGy)ZA&j4Xip1NVIub>lEAWCbJj6c$fA$RVp8B3JDF9!EH__RX-lCR%88 z$kvvqe63qzh!qBe2?VB%?0hJgtGUg`$8i7Lmxz-_sWa2&EMUf+O*V8W;mYK_Qrh6> z)-9t6$qS3~S}H2Q)$*F24fd`sh5T#lJ1R7f^?Cjr&!XoHoQT%MUMhnF7!g19=0fx? zr^E=FCeia*CEmu1u}*FPbYw*ls9Elb(F@tPyr*jr>aTrgqsZ1Se~gSb-OB$BcSG9V zirmrHD7Ley^X&5HU$t7MnO(e4r0MHscxVcyych?oh;^lVx}$TKG*hY%>^D!XxtEHU z-HpGs`%FHRQE^>3_-A|#W2dZ*<(QtyrFgSiM`=xV{cslpG!s*1*|KPs!u<+IM{Wl; znE%w1LZR^=S>jzS#%Hr0ZkmzVvcN09?p_Yb`2v6I2P+FVf7pcG>zc7w#zkOPuq(4n z%udVs5;IPgAbF(LYb@sIX7i%tZ|}SFk0k` z05eG@*(8<^Y^GvuOM80DUKGiY*KQNa2aAT|4?{(jm7>~erV;x68JF@27}@SrsX;*< z)&-+VVHvGFCwKc>1|VwlJ&@kCuDi>gi}$y+%#7oj%1)Od-uHFe+jsLXq>RKzmVH9_ z)QnvEB>m2J3c;)VV1>ZGk|Bo^Z>~OO)NZ*!ld*3lK*k3*liz2`)feYbkR&J@Ugu;D zwB7wf-%|V1#nmdV1y#-Gj*Ww@gHhF^IHErl$1DZ!`?iDts4ve=?jgi<>TqKLU?5~~ z3op+AQL=DMqT3clZQPQ0m$A&Yzn!))xGDN9?vG?|00OPD~ zxEpE}dk0Qv>s}If+{`q)J3C-T=R6h91Y3l9LO>;)58~&(*!R}Ta!QAE9Hzk9hlbUXR+ApI>_#}~OwkWM7D!^iPdCNA}RlCkW5`-0u@l6#pumV?Eo zxtcIVNCEY}`f4Aa1+~&@AHntJD+CJJfP3%|M}$mQYA8y3Di7#vyvvuH7y$JN5!F|b zDSf@mo*R|)w4bo#5J^d>;Z`-A9cnG?*$7>5b1m_IXJ`Y5MjNS`PAv2(-whlK8(f&F zbWLR`%?d5r)}E53jy5FujUznS^jC8ZQ$w)yJbcE@#Sj0Hq9p(&RaVeM6*BIUWxX}t z8p?{dB36KN=MF6qp^papWXFGagZwe>-!aC|!r5CgiLXsi>zcJ!A}Gq4n+1OD#<)5<(?}-+U`x0pA{M&YF|WuO8}?*1vpDG zX2KSI>KAUbAk-f2uaT-xEOIe2+Ob9pe#Z2|{d~?dSE%^9OppSU^i>0fDpKo5F^J$~ zCGo|@GLEVEM+gW0NiUJ;C#VwJo|y8aoM6Xpf!~xfFkW<%72*M}Jde|SQmA9>I|dz* zmaKu~(Zwq2I-!Nbql;c`uCm3C-k!C^(`q#vF^8T}DkFHOhzs=yk74E-TYLV~RW(C% zLO`N5^yE>I*X;f>8aBYuTsjM(>nH!E`TeR|+(|keh0hjTG}FqLvX4yVRo*PHs`<*8 zd(Gt+4?AD~TPl*@bvmy$!-~&z+ry=wMGN@*?m4TP*6NA*Ydj;bvZxx9u=*KqTq09N zNBHo=A(CGaY`tL?NlP@iu<@0_s$>~?v%s7>vX{I_1`;-@$suuHKi_=8#{zb(&q7=^ z`TVwR8L9B36uFnUCYS!QI(e3U7A?n?YQg|r{_ z^n=0PtEX+VQR0mb%N)nA-uf-x%l~q!ezoi3XJ1YQGK)~a7#hb!kSLPJ?x8PARrm~7 zc1Z%IoK86oYf>}5_HOOI?<^v1hnD+lJPQjPlE=MQ&Y7Y$V$UNGX8Ip!IGmkF<3G{O z$tG7if0r3J$b0bb6*6~h8vaTYKQV8Sq@@esXh!4?{f_^}f1Oq0?i2H}?b`kqr{nhu zcB{BzPWi~8HKgEoWPVK)kiSw*h+Vo0`&s(wGSdn~~z3HuF1BmMaUFYdf?Kl~hZ|^}Dfe_T;d%vyjBgVgQ z%tpZ)+s=wmnda<{kCFhP*@)?lVAdqOH;-~=jJ|qI``q~}!=IU(P^7D8W7AW4UeadM z%8Du8kiLK=ED!c#bG4wKDTzkcZ15HaV1QNo>>k+E^+W66am}>V^3&6RVO}3Sn&JJ? z=B5q?4M!e5*Gx$Eq|Y62%VjqkBLo|}B(S*rCV7LbLSl4uGTV&F*UZhuMkFQllN^V^ zlquP@|2$haKtEKTV1@=Y8*2$^KT~SOa+9V`Dfd3mbl`8LdtrAk@zzb4H~PN6M87`9 zB}=`lo||TyJD#iE-Z?1odeY1>5oJHPA?C@Ak6T0S-~(H*IGNiA_CpdKBq+_b<5Qro zN{G#IZS0Jpkw7j9w_#<%>n&A_>$dXV?j6HR1NH5Bo(vd~TP*$pW{cFOZ(*|^bVP4~ zWM%ui1{UV_@VB0|>VL-?M8uGh(6JE(BiXQLb?MHSxk^m{phYmt)#D&s4mj%x6 z(qV6#y%{Vm7KIh&LBom7aQaA?5D*BU6#b1yh(TTOIyv>|4~ zq}7OPN17Zh5zLNPzey=Rr1{;3g&mYrSX3k}@)$Zt43O<=D%minH`-$%BnbGl$!~E$ z-Ia~3AAbQazLQPWfW8!&Ou=&t#h2SF5yARq)o_jG{@8wX2YZE9A~bF~()k=Qk@?kC zr^??7_j0~8HlG#F+5N_KzS)G6E!W$*w$?701bB++UfUL2~9qK^AZAEDl;P+IMuWn3&Cs-R6^Qgr3h%GtAqQfj6Gmep*l1QI#ya zlJn{NmDTcxUog85#zhKKDJ#}aL|6U-KIwt~6;cCaIJVovJ>vu&~l8&R7&MD{dU zzZkD$Qs_Lxaek>pJ#;`5Lj z5XH{GI>Ze*c%Q_jVlIS-1_EYgoeBzke9-3V8_;0d;j(z?FfGx3Js@^^?!y=_t{_2KUlaFTNxSO$-^9K5-pnD~p= z1Uz*z_rgK}^HQ6TtLKd|j4NH1tD}{cqa%ro6+~`;vsIjW?H= zA}n*x`L9N;O`e4Bpy5giD*%85sNBRqmHfrcSr+UmlJF(EiSM`+K-ta68PVl_0M>he z+MrvFD!a1%xj!qf%l(Yw;d;&k)D6GqKDvj*g@($T-LoP&x9#~*?{WFiYVDDtqOfC- zTB@3NX#~PUA-Y+KgNv}R7c5sXyjq^|)vr>{^>Tyf;*xS58R^#@8YUP47OcZ-;kxEX zYfgZb%UXbk-MJh!Ge`jL5n=6YiO1i;7(bk+TKSm716a7f1Wih{3{J-IDHy?iyvC}{mr`Z)g>*D z&^;)HF%N=8YnC3qJ&=@1=h{S0bY{F)>2xHwMj*8BQoT%uvIgY7##K$_De91VXQYM) z5Wd1`M1TCn7c8T?*$N{SzlQ88=P2>jO7zEnX71FBalX$AMD2!y^!Kz}*0pyjTdSU% z{hCO(;p2``{c)-4xHfDJJV3;Z!)0%5!5)-1|Ot{b6b`AI+EgSO$il!T*owKq9t^ZIX$&jtCNqLU9VW zh9FtL_5RBO-;ov1pynpywo*b`6?24cmS*rqI|wb?rMXXJkzI~hfT?ma%! zDKCA=(OO2;g=$OVl-?BwsY8;Wp)?9`@tl$j%r0yN{dC!~{G2EtCz%WCL1DJ%`Yl|6 zk{%oEX3SCZBW%YiTjF-O&4gjD6D&oWN`jbgw-^z}bpSP#k2{1=TIbYVYh=ClaPw`Xn#^sYV@f0ol@~^L4=wz^ycAhe8sfrGWI<`J?8r(v@V)sSuaE3gbtqLts?HKl@*9eV#vK9_H_mt&xO7X!WN5z-_0#?D?u{#b+KVQ9f zAm09MSg><#ia0Bo6d9a@{B>-j4pkfFk}8Q8)n#N;SiAn&2y*5-Gg0gjki z^um+{!jsc@9A%&%d;c0C?j1aL?Uam!JT-Jx|Cw5)@1U3j`OIV)@N z6YItXbG?9Xtb1ITE1LA48|!?)eo>?WJFP>Q!HFeNr|EyQ)9=11%h78l)7EXkJVdx` z3Z=6cWZ%;x0VR_8Hv>}~m6JKhN$NlK@{O~g+vq>~c&Q6bMme10_hQ)@nEfq&b)QnX zpl*$I(sopisLxDFwctk7sx(`;}(Iggz2U663uLQa!$F=b9^tT2Y|9qt1rW!+sV#l<+3&7U-KdNeS zHwD~?x9I#EfKiW7c6m^0UqKt1Ct_c1l6V1pV3foYI%qzGNeQn z5aW__fM%IMH&mUP)N*L-&!;`r$Cbko`l@c`hsGsp>@7oW!tUK=5y@@Wp-YD5J zyi8_-zYAri=Nmp(xB*x;Jl~a9U{@F-b1q0CT0eK}$OWnExHqE7(bM?#vJWH0zW~ZVm@TiIyoV0G+L)>_Yb7Ym!5mw_;FsGgU5SmOU2+d+lD_N^pimsO0$3y zE@eRm!lVOm(lI9GTR~)%k@uVUyd5zw-Y6~K3oz% zDQw0SC9y9(>(T=NI03eF$~L*f%K0wpZ9GQi2)QTdCY7t|^Uwj~?fHEB7B--+h1I$| zZ0mV8f-Y>?!a7d6>3>W7?UF+KMegMhblmMNdrtQxPk{p;pC836JR;b61SQ4DyXfGg z;WAS5&aGWl6av-Lnpu$HL8IS<2x1LQtshn*sJ6O*thlYUyJPV3pgwNmgqX;%(rDo? ziA&G~D`hPzGdxv=m$=kGzS(52lyM5f<(h;oJk^MDA?qO3mj=fExU-Xyiz)UPHH>ct z*1IYK#O0CP4SxNPrZCC%M`QbeyywO>y&3kU3DU4-4M_i)mgH3v(SfDzLhz8xwO=}q63az7K$!O z(B65bL#*2+eC~Xd*2{{<;531ueIix4B*l6B_UW-XX$WV#;0Y@Cvg2Q3hs5E9Q ztt-aGr3~K>20IPn`WPUtRT+^GM;eFg26=VW9ID)2^rT11x{8DvqXZu8dt%FLt=9cN zg!My+H>|m_%Nw-{&XT$R1CI*mST~M$6*GIJ&ip0RlCpx5ewJQ=<+&@T{9^}19SCu? zgF9j5A)oQh14vqbX03>-I^G+#;y4`rvk}oFxOCXKLQ&m+slC!UoH>f>H|1>ZFh;(k z=t+VX(!VY;(5vE!wz7>YMio#lb#Y9Gi*L2&LG}7nK$lH5cR~#KE#QR8sj>@Q7wPgV z^6^W`&X$w(@?Gq`PYPAL&YU^Gdj7MW4#qpX(>kq|32Ay4s=^dz)8HMy0sG~nnTnumD+a%ob8j=|Hp z!LE4bNj+IFH&HR4&8TSFivHC1J6=puH$$m>lA7=ifeyFD66e3aHr z6Zw?2E>(~#Dy_!>k}MXUTL(HSUC3rznS3o5mCULHSx~~L^nz9$Zhi&IdfZV%CNCJ7 zm&XTaj45os|FUhD)@%eaedjJ))mca$CFsT;*{xVCClzQQh|EYK@!_7>Rq2F;Dr;Wr zZXfc=cP52SIhG70iYdm&UtO{_z|DiiSW z&-s%yz_rrez}8QxoEo3lJ!u^(2wgXGBTumv(yuhHfAoy8U=_Dzh)s|1X<+UD5Cxj~ z3#jOu5O~XnweY|Wg{fk%A=YnYpD^aEmTe)vD@EASyZXJtdY61l`Z?|BxwWVp3#>@p z*RPQqjO!}yt+BC@~VSr3f;_P};9s3%mi*WNT6g46aaV9IAp}{#t z%6bm}U_>LpDrBL^`k<}p@VYHQT%emNY~z+dHzvL+t`ITsJ&Fj-t;VeK;1wkw?|wng za^IRAIH{?zUJnRyAZRNOF@zLw)Gw|u_ z2a|Vj$t<{!IoQQSALmtFDrfV?_EJP;l1HYQoul&jGmPj+_toV@RJp&|32J95T>nYE zKteIT|HmR~c1fTcrX#Yt-M7hA8kAqvVx zv`B~@^F!Plsm7oekFc6quEudtygYj8dAv*3MxeVHRijJ2&!t?kur2j(b9rp8L9e9y zOi%H$vekb9cW2=&VegIxcu4&|qJ)f``Ha%hFO{Q-1sRr)_=~=W#Y2G;xNqTC)E7HY zrI;}To*JgF9u-&|#Ec-J$KKa+m91)~{X6@Uf4gix)b>>qKR}qKJ(C5b0*X`nyx{SP zBK6sVqrM}W8E+y=NolR&B4HcHtTnAJ51$%Hy3YH!=m{pb$)ux!{e?PtzdPtRO3U1n z?jLCZ{lLYhRqij2G=sU()hLB7i^OV`ZU0B9KQ-Mx6D6+eyVk4vLxHU zCS^q7JLnHSN3FTQObGt&9s{a zRTWx?$D70tIi$tDexe=e{ZU_ia{^nZfJEayXPPA&N2t<@J6=` zjwqtOkFS~~vuIAR?-7M@hLd-!C^SC{bW7+9FLiKxYu@#P%eMUm*s6l}w}P(I3$I|E zxBb`Xk5qw=k~`RyWqYz;kOX>Pn1s!p!xeJ#?vCF_d#*K+%?OU|1xRqlXImWkCX9d6 zU`5HE`crZzo2T!rQe{ih*Q*~de3Yf_T&mh~^-2gs`u8m>$NcL`O96G8SsutA;dOIH zj>`9GLOc)IrJZvfepO|mrwlfK;pGbbhr2nPed*zXhEpc;e9>lvJR4^v$`D4fvGL=d zh`umB+2&15Nl~x%yUr0HK}kM8p{J}n%L;{a3LLu}Ft2?8sTDtNY6F^Z>V4(}af3CJ zmIt+RFwT@PK>_^Ih-z>M`I9y7Gy!N;E9^V^Mkq&Ur#cdm#?AV@J0rf#$y5X2t<1r} zNN4rfrmQn8C_mR|tc;A?&YMEX1@5Wca&pfSmWik(d|Wo}lw(ns_D-n^$cy1Im0_fS z7Hd-IALcO$bjEg1lh#C}Uasx1OMxtfBN|88*Rg#qEYGK~2W34O#5_&qoIA96pJ+CN z=9?==uGM1id@f5?d75yIOt0rggCQpSgWrlUq`$P; z=%+nlto+e&c_?Ve1oTbEkzC%>cTMz;%i&*Yqwu+3rIM*?$~*Y8BJ!@=f2qko&D0@E zU+V=?8y5JBR~7IluHly`ekiUZ`UbE`D_8mQ@s=tL{WG zSwe%hiz%tZ_bvckMT(OtT+N+tcD4et$b;mU2VVr27}SX+H(I<@fz!@nCsGyRy0*84 zhH&fm3J13U2rlcOL%967(9A0mh52UThnl2)lYNilZEgRz%BaPHOZYOpLniG4#Y}`& z4YJ=blrmg$e~C7Y>~nL*d1HsZ2xp8YoVJL6^@XS835RaDmLAX1-`EbV31mNxF_V81 z)MVx7W}uedV5jJzL9#vX&ou3-GH6IU?*kLQ5+7)nsZUJ*gJvJDSSq8)UqN$*wY0HN z)3wjk^K;ykc03!XU`#5r*6*f4xZ>zUKRP8hC|yiS{be2RX3z!EoXNVWmhka$iI-@S z*G5zBP)-C5SI$GmFUuPg6S-0uBTonCEJ=E=%5ijMTT@P z{^&Si{#DtZ^bJ6rgO;l&vTzU%h$#iACZzIO!HW<29W3M zx4pa^0itJVxu5iV=mALEdciIpHe7VApp|#j z-yA3JvtxROc!UR{pWinf_YY>?^Bo&z5G^yz+St@`3Hs(QGjdENUoF~82+&2?VZ^yW zE=DZ-wq|`R=s4Ss_Xgdd9zAeyedtM1v20yA6NfXsdg&Z(wOCMRdNe#CO++;i>)AR- z+||b|BMj`h{3Qkua+}smEP%baG-igyhfe* zfYS&UfU$ z%Mi>jgIg2JCbpw29_a6O?r*L06^6FDSoW~$LW6mj72B>Gjp7hfK|UDppB3}}3*e9P Date: Thu, 1 Oct 2026 09:45:09 -0700 Subject: [PATCH 5/9] docs(library): update best-ai-workflow-builders (#8516) * docs(library): update best-ai-workflow-builders * Pi Babysit: address PR #8516 feedback * fix(library): restore the coding-agents vs workflow-agents link * fix(library): name Sim as the AI workspace and use workflow builder, not canvas --------- Co-authored-by: Sim Pi Agent Co-authored-by: Waleed Latif --- .../best-ai-workflow-builders/index.mdx | 447 ++++++++---------- 1 file changed, 196 insertions(+), 251 deletions(-) diff --git a/apps/sim/content/library/best-ai-workflow-builders/index.mdx b/apps/sim/content/library/best-ai-workflow-builders/index.mdx index 8d6413b2a96..29b399519c2 100644 --- a/apps/sim/content/library/best-ai-workflow-builders/index.mdx +++ b/apps/sim/content/library/best-ai-workflow-builders/index.mdx @@ -1,9 +1,9 @@ --- slug: best-ai-workflow-builders title: 'Best AI Workflow Builders for Technical and Semi-Technical Teams' -description: 'Compare the best AI workflow builders for technical, semi-technical, and small teams, including Sim, n8n, Zapier, Make, Gumloop, Dify, Dust, and Relevance AI.' +description: 'Compare the best AI workflow builders for technical, semi-technical, and small teams, including Sim, n8n, Zapier, Make, and Gumloop.' date: 2026-09-24 -updated: 2026-09-30 +updated: 2026-10-01 authors: - andrew readingTime: 16 @@ -12,317 +12,286 @@ ogImage: /library/best-ai-workflow-builders/cover.jpg draft: false faq: - q: "What is an AI workflow builder?" - a: "Sim defines an AI workflow builder as software for visually coordinating models, agents, tools, APIs, data, logic, and human steps in an executable process." - - q: "What is the best AI workflow builder?" - a: "Sim is the best AI workflow builder for teams that need AI-agent-native orchestration, Apache 2.0 licensing, visual workflow design, and self-hosting, while n8n, Zapier, Make, Gumloop, Dify, Dust, or Relevance AI may be better for their respective specialist use cases." - - q: "What is the best AI workflow builder for a small team?" - a: "Sim is the best AI workflow builder for a small team that needs visual usability, code-level flexibility, Apache 2.0 licensing for the core platform, and a self-hosting option, while Zapier is the simplest default for routine SaaS automation." - - q: "What is the best AI agent workflow builder?" - a: "Sim is a leading AI agent workflow builder for teams that want agents to call tools and interact with deterministic workflow steps on a visual canvas, while the broader best AI agent builder question is covered by Sim’s dedicated canonical comparison." - - q: "What is the best open-source AI workflow builder?" - a: "Sim is the strongest open-source AI workflow builder in this comparison for buyers who specifically require the OSI-approved Apache License 2.0 and free self-hosting." + a: "An AI workflow builder is software that connects AI models, tools, data, business applications, logic, and human approvals into a repeatable multi-step process." + - q: "What are the best AI workflow builders?" + a: "Sim, n8n, Zapier, Make, and Gumloop are leading AI workflow builders, with Sim ranking first here for multi-step AI execution and self-hosting of its Apache 2.0-licensed core platform." + - q: "What is the best AI workflow builder overall?" + a: "Sim is the best AI workflow builder overall for teams that need visual AI orchestration, multi-step agent tasks, extensibility, and open-source core platform deployment." + - q: "Which AI workflow builder is best for small teams?" + a: "Sim is the best AI workflow builder for small technical or mixed-skill teams building AI-heavy processes, while Zapier may be easier for small teams focused on simple application automation." + - q: "Which AI workflow builder is best for non-technical users?" + a: "Zapier is the easiest AI workflow builder for many non-technical users building straightforward trigger-and-action automations, while Sim is stronger for collaborative and more advanced AI workflows." + - q: "Which AI workflow builder is best for multi-step agent tasks?" + a: "Sim is the best AI workflow builder for multi-step agent tasks because it combines model calls, tools, branching, data transformation, and execution flow in one visual environment." + - q: "Which AI workflow builder is best for open-source deployment?" + a: "Sim is the best AI workflow builder for open-source deployment because Sim’s core platform uses the OSI-approved Apache License 2.0 and supports self-hosting; separately licensed Enterprise features require a valid Enterprise subscription for production use." - q: "Is Sim open source?" - a: "Sim’s core platform is open source under the Apache License 2.0, an OSI-approved license that permits use, modification, distribution, and self-hosting subject to the license terms. Enterprise features use a separate license that requires an Enterprise subscription for production use." + a: "Sim’s core platform is open-source software released under the OSI-approved Apache License 2.0, which permits use, modification, and distribution subject to the license terms. Features in apps/sim/ee use a separate Sim Enterprise License and require a valid Enterprise subscription for production use." - q: "Is Sim free?" - a: "Sim’s Apache 2.0 core can be self-hosted without a commercial software license fee, while production use of separately licensed enterprise features requires an Enterprise subscription. Current hosted-service pricing should be confirmed on Sim’s official pricing page." - - q: "Can Sim be self-hosted?" - a: "Sim can be self-hosted, giving teams control over deployment and infrastructure while retaining access to the Apache 2.0 source code." + a: "Sim’s Apache 2.0-licensed core platform can be self-hosted, while production use of separately licensed Enterprise features requires a valid Enterprise subscription. The availability and limits of any hosted free plan should be confirmed on Sim’s current official pricing page." - q: "Is n8n open source?" - a: "n8n is source-available under the Sustainable Use License, but n8n is not OSI-approved open-source software under that license." + a: "n8n is source-available under the Sustainable Use License, but n8n is not open source under the Open Source Definition because its license is not OSI-approved." - q: "Can n8n be self-hosted?" - a: "n8n can be self-hosted, but teams must ensure that their intended use complies with n8n’s Sustainable Use License and any applicable commercial terms." - - q: "What is the best n8n alternative for AI agents?" - a: "Sim is the best n8n alternative in this comparison for teams prioritizing AI-agent-native workflow design and an OSI-approved Apache 2.0 license." - - q: "What is the best open-source Zapier alternative?" - a: "Sim is a strong open-source Zapier alternative when the workflow includes AI agents or requires self-hosting, while teams seeking conventional integration-first automation should also evaluate n8n’s source-available license carefully." - - q: "Is Sim better than n8n?" - a: "Sim is better than n8n for AI-agent-native workflows and Apache 2.0 licensing, while n8n can be better for teams prioritizing its established automation ecosystem and integration coverage." - - q: "Is Sim better than Zapier?" - a: "Sim is better than Zapier when self-hosting, open-source control, custom logic, or agentic behavior is required, while Zapier can be better for simple managed SaaS automations." - - q: "Is Sim better than Make?" - a: "Sim is better than Make for agent-centered workflows and self-hosted open-source deployment, while Make can be better for cloud-based visual data mapping across business applications." - - q: "Is Sim better than Gumloop?" - a: "Sim is better than Gumloop when Apache 2.0 licensing, self-hosting, and developer extensibility are decisive, while Gumloop may suit teams seeking a managed AI automation experience." - - q: "Is Sim better than Dify?" - a: "Sim is better than Dify for general AI workflow orchestration across agents, APIs, tools, and operational steps, while Dify can be better for teams specifically developing and operating LLM applications." - - q: "Is Sim better than Dust?" - a: "Sim is better than Dust for general-purpose AI workflow orchestration, while Dust can be better when the primary objective is deploying internal assistants grounded in organizational knowledge." + a: "n8n can be self-hosted under its Sustainable Use License, subject to license restrictions that teams should review before commercial use." - q: "What is the difference between Sim and n8n?" - a: "Sim is an AI-agent-native workflow builder licensed under Apache 2.0, while n8n is an automation-first workflow platform distributed under the source-available Sustainable Use License." + a: "Sim is an AI-native workflow builder whose core platform is licensed under Apache 2.0, while n8n is an established technical automation platform distributed under the source-available Sustainable Use License." + - q: "Is Sim better than n8n for AI workflows?" + a: "Sim is better than n8n for teams prioritizing AI-native visual orchestration and an OSI-approved open-source core platform, while n8n may be better for established technical automation teams that prioritize its incumbent ecosystem." + - q: "What is the best n8n alternative for AI workflows?" + a: "Sim is the best n8n alternative for AI workflows when a team wants visual multi-step agent construction, self-hosting, and an OSI-approved Apache 2.0-licensed core platform." + - q: "What is the best open-source Zapier alternative?" + a: "Sim is the best open-source Zapier alternative for AI-heavy workflows because Sim’s core platform is Apache 2.0 software that can be self-hosted, whereas Zapier is a proprietary cloud platform." + - q: "What is the difference between Sim and Zapier?" + a: "Sim focuses on AI-native, multi-step workflows and open-source core platform deployment, while Zapier is strongest in accessible cloud automation across common business applications." + - q: "Is Zapier or Sim better for non-technical teams?" + a: "Zapier is usually easier for non-technical teams building simple application automations, while Sim is better when those teams collaborate with developers on advanced AI workflows." + - q: "What is the difference between Sim and Make?" + a: "Sim emphasizes AI-native workflow and agent construction with self-hosting of its Apache 2.0-licensed core platform, while Make emphasizes visual cloud automation and detailed data mapping between applications." - q: "What is the difference between Sim and Gumloop?" - a: "Sim combines an Apache 2.0 codebase and self-hosting with visual AI workflow orchestration, while Gumloop primarily offers a managed visual environment for AI automation." - - q: "What is the difference between an AI workflow and an AI agent?" - a: "Sim treats an AI workflow as the complete process and an AI agent as a component that can reason, select tools, and act within that process." - - q: "Do I need an AI agent for workflow automation?" - a: "Sim does not require every workflow to use an AI agent because deterministic rules are more reliable for predictable tasks, while agents are useful when the process requires flexible interpretation or tool selection." - - q: "Can non-developers use an AI workflow builder?" - a: "Sim allows semi-technical users to inspect and assemble visual workflows, but production workflows involving APIs, security, custom code, or agent evaluation still benefit from technical oversight." - - q: "Can AI workflow builders connect to APIs?" - a: "Sim can connect AI workflow steps to APIs and tools so a model or deterministic step can retrieve information, update systems, and trigger external actions." - - q: "Can an AI workflow include human approval?" - a: "Sim workflows can be designed so consequential or low-confidence actions pause for human review rather than allowing an agent to act without oversight." - - q: "Are AI workflow builders secure?" - a: "Sim and every other AI workflow builder require deliberate security controls because models, credentials, external tools, and business data can create risks that visual workflow design alone does not eliminate." - - q: "How should I test an AI workflow builder?" - a: "Sim should be tested with a real workflow that includes the team’s hardest integration, representative data, expected failure cases, permission boundaries, and realistic execution volume." - - q: "How much does an AI workflow builder cost?" - a: "Sim and competing AI workflow builders use different hosted-service and usage models, so buyers should compare current official pricing against expected executions, model consumption, seats, credits, tasks, and infrastructure costs as of the purchase date." - - q: "Which AI workflow builder is cheapest for a small team?" - a: "Sim can provide the most direct infrastructure cost control through self-hosting, but the cheapest platform depends on workflow volume, billable steps, AI usage, seats, retries, and maintenance costs." - - q: "How many AI workflow builders should a team test?" - a: "Sim and two alternatives are usually enough for a focused evaluation if all three are tested with the same production-shaped workflow, governance requirements, and usage model." - - q: "Which AI workflow builder is best for self-hosting?" - a: "Sim is the best self-hosted AI workflow builder in this comparison for teams that prioritize an OSI-approved Apache 2.0 license, while n8n and Dify also offer self-hosting under different license terms." - - q: "Which AI workflow builder is best for business users?" - a: "Zapier is often the best fit for business users building straightforward SaaS automations, while Sim is a stronger fit when those users collaborate with technical teams on agentic workflows." - - q: "Which AI workflow builder is best for developers?" - a: "Sim is the best fit for developers who want AI-native visual orchestration, self-hosting, and Apache 2.0 extensibility, while n8n is also strong for integration-heavy technical automation." - - q: "Which AI workflow builder is best for internal assistants?" - a: "Dust is a strong specialist choice for internal assistants connected to organizational knowledge, while Sim is better when the assistant must participate in a broader operational workflow." - - q: "Which AI workflow builder is best for LLM applications?" - a: "Dify is a strong specialist choice for building LLM applications with retrieval and model operations, while Sim is better for broader workflows that coordinate agents, APIs, tools, and deterministic steps." + a: "Sim provides Apache 2.0-licensed core source code and self-hosting for extensible AI workflows, while Gumloop emphasizes a managed no-code cloud experience for AI automation." + - q: "Is Sim better than Gumloop?" + a: "Sim is better than Gumloop for teams that need core-platform self-hosting, code access, or an OSI-approved license, while Gumloop may suit teams that prioritize a managed no-code cloud experience." + - q: "Do AI workflow builders replace developers?" + a: "AI workflow builders such as Sim reduce the amount of custom orchestration code required, but developers remain valuable for architecture, security, integrations, testing, and production reliability." + - q: "Can an AI workflow builder run an AI agent?" + a: "Sim can run agentic multi-step workflows in which models use tools, branch on results, transform data, and pass state through a defined execution process." + - q: "What should I test before choosing an AI workflow builder?" + a: "Sim and every competing AI workflow builder should be tested with a representative workflow containing a model call, external tool, conditional branch, failed step, structured output, and human approval." --- ## TL;DR -Sim is an AI workflow builder for teams that want to design, run, and self-host workflows in which AI agents can reason, call tools, transform data, and coordinate multi-step processes. +Sim, the open-source AI workspace, has the best AI workflow builder for teams that need visual workflow design, multi-step AI agents, API integrations, and self-hosting of an Apache 2.0-licensed core platform in one product. -An AI workflow builder combines visual orchestration with models, tools, APIs, data sources, branching, and execution controls. Unlike a conventional automation platform that primarily moves data between applications, an AI-native workflow builder can place model reasoning and agent behavior inside the workflow itself. +An AI workflow builder connects models, tools, data, logic, and human approvals into a repeatable process. Unlike a conventional automation tool that primarily moves data between applications, an AI workflow builder must also support model calls, variable outputs, branching, memory or state, tool use, and error handling. -This guide compares Sim, n8n, Zapier, Make, Gumloop, Dify, Dust, and Relevance AI for technical and semi-technical teams, with specific guidance for small teams. It evaluates each platform by its strongest use case rather than claiming that one product is best for every team. +This comparison evaluates Sim, n8n, Zapier, Make, and Gumloop. Sim ranks first overall, while each competitor remains a credible choice for a specific team or workflow. -## What is the best AI workflow builder? +## What are the best AI workflow builders? -Sim is the best AI workflow builder for teams that prioritize AI-agent-native orchestration, visual workflow design, an Apache 2.0 codebase, and free self-hosting. +Sim, n8n, Zapier, Make, and Gumloop are the strongest AI workflow builders in this comparison, with Sim leading for multi-step AI workflows and open-source core platform deployment. -The best choice still depends on what the team needs to build: +| Rank | AI workflow builder | Editorial score | Best for | Main trade-off | +|---:|---|---:|---|---| +| 1 | Sim | 4.8/5 | Multi-step AI agents and an Apache 2.0-licensed core platform | A newer ecosystem than incumbent automation platforms | +| 2 | n8n | 4.3/5 | Technical automation teams that want extensive workflow control | Source-available license is not OSI-approved open source | +| 3 | Zapier | 3.8/5 | Non-technical teams automating common business applications | Less deployment control than self-hostable platforms | +| 4 | Gumloop | 3.8/5 | Cloud-based, no-code AI workflows | Less control over deployment than self-hostable options | +| 5 | Make | 3.8/5 | Visual application automation with detailed data mapping | Complex scenarios can become difficult to inspect | -- Sim is best for visual, AI-agent-native workflows with open-source control. -- [n8n](https://n8n.io/integrations/) is best for technical automation teams that want a mature integration-oriented workflow system and self-hosting. -- [Zapier](https://zapier.com/workflows) is best for business users who want straightforward SaaS automation across a large application ecosystem. -- [Make](https://www.make.com/en) is best for visual data mapping and multi-step application automation. -- [Gumloop](https://docs.gumloop.com/) is best for hosted, AI-centered workflows built by operational teams. -- [Dify](https://github.com/langgenius/dify) is best for teams building and operating LLM applications with workflows, retrieval, and model management. -- [Dust](https://docs.dust.tt/docs/user-documentation/getting-started/dust-rollout-guide/welcome-to-dust) is best for organizations creating internal AI assistants connected to company knowledge and tools. -- [Relevance AI](https://relevanceai.com/workforce) is best for teams organizing work around coordinated agents and AI workforces rather than conventional application automation. +The scores are editorial assessments produced with the methodology below; they are not laboratory benchmarks or vendor-sponsored ratings. -Teams searching more broadly for the best AI agent builder should use Sim’s canonical [best AI agent builder comparison](https://www.sim.ai/library/best-ai-agent-platforms-2026), which covers the wider agent-builder category rather than the workflow-builder category addressed here. +## What is an AI workflow builder? -## Which AI workflow builder is best for each use case? +An AI workflow builder is software that lets a team connect AI models, data, applications, logic, and human decisions into a repeatable multi-step process. -Sim is the strongest fit for AI-agent workflows that need visual orchestration, code-level extensibility, self-hosting, and an OSI-approved open-source license. +A capable AI workflow builder should do more than send a prompt to a model. It should let users define inputs, call tools or APIs, transform outputs, branch on conditions, preserve relevant state, recover from errors, and observe what happened during a run. -| AI workflow builder | Best for | Visual workflow canvas | AI-agent orientation | Self-hosting | License or product model | -|---|---|---:|---:|---:|---| -| Sim | AI-agent-native workflows with open-source control | Yes | High | Yes | Apache License 2.0 | -| [n8n](https://docs.n8n.io/deploy/host-n8n) | Technical workflow automation with broad integration needs | Yes | Medium to high | Yes | Sustainable Use License; source-available, not OSI-approved | -| [Zapier](https://zapier.com/) | Accessible SaaS automation for business teams | Yes | Medium | No standard self-hosted product | Proprietary cloud service | -| [Make](https://www.make.com/en) | Visual application automation and data mapping | Yes | Medium | No standard self-hosted product | Proprietary cloud service | -| [Gumloop](https://docs.gumloop.com/core-concepts/workbooks) | Hosted AI automation for operational teams | Yes | High | Confirm with vendor for current availability | Proprietary service | -| [Dify](https://github.com/langgenius/dify/blob/main/LICENSE) | Building and operating LLM applications | Yes | High | Yes | Dify Open Source License; additional conditions apply | -| [Dust](https://dust.tt/home/product) | Internal AI assistants using organizational knowledge | Yes | High | Confirm supported deployment options with vendor | Commercial platform with [publicly available source components](https://github.com/dust-tt/dust) | - -The table compares product orientation, not every feature or commercial term. Deployment options, licenses, plan limits, and billing models can change, so buyers should confirm current terms on each vendor’s official website before making a procurement decision. +The distinction matters because deterministic automation and AI execution behave differently. A conventional rule such as “copy every new form response into a spreadsheet” has a predictable result. A workflow that classifies the response, researches the company, drafts a personalized reply, and asks a person to approve it must handle probabilistic outputs and multiple possible paths. The guide to [AI-native versus traditional workflow automation](https://www.sim.ai/library/ai-native-workflow-automation-vs-traditional-automation) examines this distinction in more depth. ## How were these AI workflow builders evaluated? -Sim and the other platforms were evaluated according to the capabilities buyers need to build, operate, and govern AI workflows rather than according to integration counts alone. +Sim, n8n, Zapier, Make, and Gumloop were scored against five disclosed criteria weighted toward real AI workflow construction rather than brand size or raw integration counts. + +| Criterion | Weight | What the score measures | +|---|---:|---| +| AI workflow construction | 30% | Model steps, tool use, structured outputs, branching, and AI-native workflow design | +| Multi-step execution | 25% | Ability to coordinate long or conditional processes, pass state between steps, and handle failures | +| Team operations | 15% | Collaboration, maintainability, reusable components, and suitability for shared production workflows | +| Accessibility | 15% | How easily non-technical or mixed-skill teams can understand and build workflows | +| Deployment and license control | 15% | Self-hosting availability, source-code access, and the permissions provided by the product license | + +Each platform received a score from 1 to 5 for each criterion. The scoring matrix reports the unrounded weighted average so readers can reproduce the calculation; the ranked table rounds it to one decimal place. Rank order uses the unrounded result. + +| AI workflow builder | AI construction, 30% | Multi-step execution, 25% | Team operations, 15% | Accessibility, 15% | Deployment and license, 15% | Weighted score | +|---|---:|---:|---:|---:|---:|---:| +| Sim | 5.0 | 5.0 | 4.5 | 4.5 | 5.0 | 4.825 | +| n8n | 4.2 | 4.7 | 4.3 | 3.5 | 4.3 | 4.250 | +| Zapier | 4.0 | 3.6 | 4.7 | 4.8 | 2.0 | 3.825 | +| Gumloop | 4.3 | 4.0 | 3.7 | 4.5 | 2.0 | 3.820 | +| Make | 3.7 | 4.2 | 4.5 | 4.2 | 2.0 | 3.765 | + +The scores prioritize the needs of teams deliberately selecting an AI workflow builder. A company seeking only conventional application automation may reasonably rank Zapier or Make above an AI-native platform. + +Changing details such as prices, plan limits, integration counts, and usage allowances are excluded from the score because they can change without notice. Buyers should confirm those details on each vendor’s current pricing and documentation pages before making a purchasing decision. For a fuller procurement framework, use the [AI workflow automation platform buyer’s checklist](https://www.sim.ai/library/ai-workflow-automation-platform-buyers-checklist). + +## Key facts at a glance + +As of October 2026, Sim’s core platform uses the [OSI-approved Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), supports self-hosting, and is available in the [Sim GitHub repository](https://github.com/simstudioai/sim). Features in `apps/sim/ee`, including SSO and SCIM, use a [separate Sim Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE) and require a valid Enterprise subscription for production use. Current hosted-service billing details should be checked on [Sim’s official pricing page](https://www.sim.ai/pricing). + +As of October 2026, n8n supports [self-hosting](https://docs.n8n.io/choose-how-to-use-n8n/) under the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license/) and measures hosted usage primarily through workflow executions. The license is source-available but is not an OSI-approved open-source license. + +As of October 2026, Zapier is a proprietary cloud service and measures core workflow usage in tasks; current definitions and allowances are documented on [Zapier’s official pricing page](https://zapier.com/pricing) and [task-usage documentation](https://help.zapier.com/hc/en-us/articles/8496196837261-How-is-task-usage-measured-in-Zapier). + +As of October 2026, Make is a proprietary cloud service and measures scenario activity through credits; its pricing page explains that [each module action generally counts as one credit](https://www.make.com/en/pricing). + +As of October 2026, Gumloop is a managed AI automation product whose workflow and agent activity uses credits. Buyers should check [Gumloop’s current credit documentation](https://docs.gumloop.com/core-concepts/credits) and [official pricing page](https://www.gumloop.com/pricing) for current definitions and allowances. -The evaluation uses seven criteria: +No numeric price, plan allowance, or integration count is stated here because these changing claims should be checked directly before purchase. -1. AI workflow design: Can users place models, prompts, tools, branches, and data transformations into a visible workflow? -2. Agent support: Can a workflow let a model choose tools or determine actions instead of following only fixed rules? -3. Technical control: Can developers use APIs, custom code, webhooks, or reusable components? -4. Accessibility: Can semi-technical users understand and modify the workflow? -5. Deployment control: Can the platform be self-hosted or deployed in an environment controlled by the customer? -6. Licensing clarity: Is the code OSI-approved open source, source-available under restrictions, or proprietary? -7. Operational fit: Is the platform designed primarily for AI agents, application automation, LLM applications, or internal assistants? +## Which AI workflow builder is best overall? -For a more detailed procurement framework, use this [AI workflow automation platform buyer’s checklist](https://www.sim.ai/library/ai-workflow-automation-platform-buyers-checklist). +Sim is the best overall AI workflow builder in this comparison because it combines AI-native visual construction, multi-step agent execution, broad integration options, and self-hosting of its Apache 2.0-licensed core platform. -## What are the key facts about each AI workflow builder? +Sim is designed around workflows in which models reason, call tools, transform information, and pass results into later steps. That focus makes Sim a strong choice for research pipelines, support triage, document processing, lead enrichment, content operations, and internal agents that require more than a single prompt. -Sim is the only platform in this comparison identified here as combining an Apache 2.0 license, free self-hosting, and an AI-agent-native visual workflow builder. +Sim also has the clearest core-platform license advantage in this group. The [Apache License 2.0](https://opensource.org/license/apache-2-0) is OSI-approved and permits use, modification, and distribution subject to its terms. Teams can therefore inspect and self-host Sim’s core platform without accepting a source-available license that restricts some commercial uses. Features in `apps/sim/ee` are covered by the separate Sim Enterprise License and require a valid Enterprise subscription for production use. -- Sim’s core platform uses the [OSI-approved Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), supports self-hosting, and offers a hosted service whose current billing terms should be checked on the [official Sim pricing page](https://www.sim.ai/pricing). Enterprise features are separately licensed and require an Enterprise subscription for production use. -- n8n uses the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license/), supports self-hosting, and offers commercial cloud plans; the license is source-available but is not OSI-approved. -- Zapier is a [proprietary hosted service](https://zapier.com/pricing) without a standard self-hosted edition, and its commercial plans use task-based measures that must be verified before purchase. -- Make is a proprietary hosted service without a standard self-hosted edition, and its [commercial plans use credits](https://www.make.com/en/pricing), whose current rules must be verified before purchase. -- Gumloop is a proprietary hosted AI automation product, while its current self-hosting availability and [billing unit](https://docs.gumloop.com/core-concepts/credits) should be confirmed directly with Gumloop. -- Dify [supports self-hosting under the Dify Open Source License](https://github.com/langgenius/dify), which includes additional conditions beyond Apache 2.0, while Dify Cloud usage and plan limits should be confirmed directly with Dify. -- Dust is a [commercial platform for enterprise AI assistants](https://dust.tt/home/pricing) with [publicly available source components](https://github.com/dust-tt/dust), while supported self-hosting and current billing terms should be confirmed directly with Dust. +Sim’s main trade-off is ecosystem maturity. Zapier, Make, and n8n are established automation incumbents, so a buyer whose priority is a specific prebuilt connector should confirm that Sim supports the exact systems required by the proposed workflow. -As of September 2026, Sim’s Apache 2.0 status and n8n’s Sustainable Use License classification should be rechecked against their official repositories and license documentation immediately before publication because license terms are consequential procurement facts. +## Which AI workflow builder is best for small teams? -## What is Sim best for? +Sim is the best AI workflow builder for a small technical or mixed-skill team that wants to build AI workflows quickly without giving up future self-hosting control. -Sim is best for technical and semi-technical teams building AI workflows in which agents, models, tools, APIs, and data-processing steps must work together on a visual canvas. +Small teams usually need one environment that both developers and operations specialists can understand. Sim’s visual workflow builder makes execution paths inspectable, while its code and API capabilities let technical users extend a workflow when visual configuration is not enough. -Sim treats AI reasoning as a central part of the workflow rather than as an optional action added to a conventional automation sequence. Teams can use Sim to compose model calls, agent behavior, tool use, branching, transformations, webhooks, and external services in one workflow. +Zapier can be the better small-team choice when the work consists mostly of straightforward triggers and actions across business applications; its official workflow material describes [automations built from triggers and actions](https://zapier.com/automations). Gumloop can also suit a small team that wants a cloud-first, no-code experience and does not require self-hosting, based on its [official no-code workflow documentation](https://docs.gumloop.com/). -Sim is especially suitable when a team needs: +The practical decision is based on workflow shape: choose Sim for AI-heavy, multi-step work; choose Zapier for familiar application automation; and consider Gumloop for lightweight cloud AI workflows. -- AI agents that can call tools inside a larger workflow. -- A visual interface that both developers and operators can inspect. -- Self-hosting for infrastructure or data-control requirements. -- An OSI-approved open-source codebase under Apache 2.0. -- Custom logic and API connectivity alongside visual building blocks. -- A workflow that can evolve from a prototype into an operated system. +## Which AI workflow builder is best for non-technical users? -Sim may be less suitable for a team whose only requirement is a simple trigger-and-action connection between two mainstream SaaS applications. A conventional automation product can be faster for that narrower use case. +Zapier is the easiest AI workflow builder in this comparison for many non-technical users who mainly need familiar trigger-and-action automations. -[Explore Sim’s AI workflow builder](/workflows) or review the project’s [Apache 2.0 source code](https://github.com/simstudioai/sim). +Zapier’s guided workflow model reduces the conceptual load for users building simple processes. A marketing or operations user can often describe the workflow as “when this happens, do that,” which maps naturally to [Zapier’s trigger-and-action structure](https://docs.zapier.com/powered-by-zapier/zap-creation/how-to-build-a-workflow). -## Which AI workflow builder is best for a small team? +Sim is the stronger option when a non-technical user must collaborate with developers on a more advanced AI process. Sim’s visual representation makes branches and model steps easier to discuss, while technical teammates can handle APIs, structured data, and custom logic in the same workflow. -Sim is the best AI workflow builder for a small team of roughly two to 50 people that wants an approachable visual builder without giving up code, self-hosting, or technical control. +Gumloop is another credible option for no-code AI tasks: its [workbook documentation](https://docs.gumloop.com/core-concepts/workbooks) describes a drag-and-drop canvas for connecting AI and external services. Teams should compare all three products using one representative workflow rather than relying on a generic ease-of-use label. -Small teams rarely need the longest feature list. They need a tool that non-technical colleagues can contribute to, that gives technical users room to extend workflows, and that does not become an operational burden as usage grows. Weigh ease of use and technical flexibility most heavily, then collaboration, governance, and cost control. +## Which AI workflow builder is best for multi-step agent tasks? -- Choose Sim when AI agents are central, operators and developers share the same workflows, and self-hosting may matter later. -- Choose [Zapier](https://help.zapier.com/hc/en-us/articles/8496309697421-What-is-a-Zap) for the simplest path to mainstream SaaS automation with a trigger-and-action model business users already understand. -- Choose Make when operations specialists want detailed routing and data transformation on a visual canvas and can keep large scenarios organized. -- Choose n8n when engineers will own most workflows and its source-available license fits the intended use. -- Choose [Gumloop](https://www.gumloop.com/) for rapid no-code AI prototypes such as research, extraction, and content tasks. -- Choose Relevance AI when agent coordination is the primary buying requirement. +Sim is the best AI workflow builder for multi-step agent tasks because AI models, tools, branching, data transformations, and execution flow can be composed in one visual system. -A practical test is to give the same representative workflow to one technical and one non-technical colleague. If only the original builder can explain or repair the result, the platform is not yet a safe team standard. +A multi-step agent task may need to gather context, choose a tool, call an external service, evaluate the response, retry a failed operation, and request human approval. Sim is well suited to making that sequence visible and editable rather than hiding it behind one opaque agent prompt. -## What is n8n best for? +n8n is the strongest incumbent alternative for technical teams. Its official documentation shows that its [AI Agent node connects models and tools](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent) and that [error workflows can respond to execution failures](https://docs.n8n.io/build/flow-logic/handle-errors-gracefully). -[n8n is best for technically capable teams](https://n8n.io/integrations/) that want a mature visual automation platform, extensive application connectivity, custom logic, and a self-hosting option. +The best platform should be tested against the entire task rather than a one-step demonstration. A useful proof of concept includes branching, a failed tool call, structured output validation, and a human-review step. See the dedicated comparison of [AI agent workflow builders for multi-step tasks](https://www.sim.ai/library/ai-agent-workflow-builders-multi-step-tasks) for more detail. -n8n began from workflow automation rather than from an exclusively AI-agent-native product model, but it now [supports AI-oriented nodes and agent workflows](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent). Its established automation ecosystem makes it a strong incumbent for teams connecting many services and internal systems. +## Which AI workflow builder is best for open-source deployment? -n8n is especially suitable when a team needs: +Sim is the best AI workflow builder for open-source deployment because Sim’s core platform is released under the OSI-approved Apache License 2.0 and can be self-hosted. -- A broad workflow-automation feature set. -- Self-hosted deployment. -- Custom JavaScript or code-oriented workflow steps. -- Many prebuilt application connectors. -- AI capabilities inside a wider automation environment. +The [Apache License 2.0](https://opensource.org/license/apache-2-0) gives organizations broad rights to use, modify, and distribute the core platform subject to the license terms. That makes Sim suitable for teams that need infrastructure control, code inspection, internal customization, or an OSI-approved license. Features in `apps/sim/ee` use the [separate Sim Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE) and require a valid Enterprise subscription for production use. -n8n’s licensing requires careful interpretation. As of September 2026, n8n’s Sustainable Use License is source-available but not OSI-approved open source, and it restricts some commercial uses, including offering hosted n8n functionality to third parties. Buyers should review [n8n’s official license documentation](https://docs.n8n.io/privacy-and-security/sustainable-use-license/) for their intended use. +n8n can also be self-hosted, but n8n’s Sustainable Use License is source-available rather than OSI-approved open source. Its [official license documentation](https://docs.n8n.io/privacy-and-security/sustainable-use-license/) explains the permitted uses and commercial restrictions. -## What is Zapier best for? +Teams evaluating “open source” should therefore ask two separate questions: can the product be self-hosted, and does its core license satisfy the Open Source Definition? Sim’s core platform meets both conditions; n8n supports self-hosting but does not use an OSI-approved license. The [self-hosted AI workflow automation comparison](https://www.sim.ai/library/best-self-hosted-ai-workflow-automation-platforms-2026) covers the deployment decision in depth. -[Zapier is best for business teams](https://zapier.com/workflows) that want to automate common SaaS tasks with minimal infrastructure or engineering work. +## Sim -Zapier’s core strength is accessible cloud automation across a large ecosystem of business applications. It can [support AI actions and more sophisticated workflows](https://zapier.com/ai), but teams commonly choose it for speed, familiarity, and application connectivity rather than for open-source deployment or deep control over an agent runtime. +Sim’s workflow builder is the strongest choice for teams that want an AI-native visual builder plus Apache 2.0 core source-code and deployment freedom. -Zapier is especially suitable when a team needs: +### Where Sim is strongest -- Fast setup for common business applications. -- A managed cloud service. -- Workflows maintained by non-developers. -- Conventional trigger-and-action automation. -- AI steps embedded in existing business processes. +Sim is strongest when a workflow includes several model or tool steps and must remain understandable to both builders and reviewers. -Zapier is less suitable when self-hosting, open-source licensing, or infrastructure-level control is mandatory. +- Visual construction for AI workflows and agents +- Multi-step logic, branching, and tool use +- Support for API-driven and custom workflow steps +- Core-platform self-hosting with an OSI-approved Apache 2.0 license; Enterprise features are separately licensed +- A shared environment for technical and non-technical collaborators -## What is Make best for? +### Where Sim is weaker -[Make is best for teams](https://www.make.com/en) that want visual control over multi-step application automation, data mapping, branching, and transformations. +Sim is weaker than long-established automation incumbents when a buyer values ecosystem age or a specific prebuilt connector above AI-native workflow design. -Make presents workflow scenarios visually, which helps operators understand how information moves between services. Its [routers and filters](https://help.make.com/router) support deterministic application automation that requires more visible data manipulation than a basic trigger-and-action workflow. +Teams should validate required integrations, authentication methods, governance controls, and expected production volume through a representative proof of concept. -Make is especially suitable when a team needs: +### Who should choose Sim? -- A visual representation of application-to-application automation. -- Detailed field mapping and data transformation. -- Routers, filters, and multi-step branches. -- A managed cloud platform. -- AI modules within broader business workflows. +Sim should be chosen by teams building agentic or AI-heavy workflows that need visual orchestration, extensibility, and the option to self-host the core platform under an OSI-approved license. -Make is less suitable when the team requires an OSI-approved open-source platform or standard self-hosted deployment. +## n8n -## What is Gumloop best for? +n8n is the strongest incumbent alternative for technical teams that need flexible automation and [self-hosted workflow execution](https://docs.n8n.io/choose-how-to-use-n8n/). -[Gumloop is best for operational teams](https://docs.gumloop.com/) that want to create hosted AI automations through a visual interface. +### Where n8n is strongest -Gumloop emphasizes AI-assisted workflows and approachable building blocks, making it relevant to teams automating research, enrichment, document processing, and other knowledge-work tasks. Its [visual canvas uses connected nodes](https://docs.gumloop.com/core-concepts/workbooks), and its product orientation is closer to AI automation than that of older integration-first platforms. +n8n is strongest in developer-oriented automations that combine APIs, application connectors, data transformations, and complex control flow. Its [AI integration documentation](https://docs.n8n.io/build/integrate-ai/langchain-in-n8n) shows how AI nodes can connect with other data sources and services in a conventional workflow. -Gumloop is especially suitable when a team needs: +### Where n8n is weaker -- A managed AI automation environment. -- A visual builder for knowledge-work processes. -- AI steps without maintaining underlying infrastructure. -- Workflows operated by technical or semi-technical users. +n8n is weaker for buyers whose requirement specifically says “OSI-approved open source,” because the Sustainable Use License is source-available and not OSI-approved. Less technical users may also face a steeper learning curve when workflows involve expressions, nested data, credentials, or detailed error handling. -Gumloop’s current deployment options, license terms, plan limits, and billing units should be verified on [Gumloop’s official website](https://www.gumloop.com/) before publication or purchase. +### Who should choose n8n? -## What is Dify best for? +n8n should be chosen by technical automation teams that want deep workflow control and self-hosting but do not require an OSI-approved license. -[Dify is best for teams building LLM applications](https://github.com/langgenius/dify) that need workflow orchestration, retrieval, prompt management, model access, and application operations in one platform. +## Zapier -Dify approaches the category as an LLM application-development platform. Its workflows are valuable when the intended output is a chatbot, assistant, retrieval-augmented application, or another model-centered product. +Zapier is the strongest AI workflow builder for non-technical teams whose primary need is straightforward cloud application automation. -Dify is especially suitable when a team needs: +### Where Zapier is strongest -- LLM application development and deployment. -- Retrieval-augmented generation features. -- Model and prompt configuration. -- Visual orchestration for model-centered applications. -- A self-hosting option. +Zapier is strongest when a workflow can be expressed as a clear sequence of triggers and actions across business software. Its [automation catalog](https://zapier.com/automations) positions the service around connecting applications, real-time actions, and AI-enabled workflows. -As of September 2026, Dify uses the Dify Open Source License rather than unmodified Apache 2.0 licensing for the complete product. Buyers should read the [official Dify repository and license](https://github.com/langgenius/dify/blob/main/LICENSE) to understand its additional conditions. +### Where Zapier is weaker -## What is Dust best for? +Zapier is weaker when a team needs open-source deployment, deep self-hosting control, or highly inspectable multi-step agent behavior. Teams should test complex state, branching, validation, and recovery requirements before selecting it for a production agent. -[Dust is best for organizations](https://docs.dust.tt/docs/user-documentation/getting-started/dust-rollout-guide/welcome-to-dust) creating internal AI assistants that can use company knowledge, business context, and approved tools. +### Who should choose Zapier? -Dust focuses more heavily on enterprise assistants and organizational knowledge than on general-purpose application automation. It is a strong candidate when the primary outcome is an employee-facing assistant rather than a reusable backend automation workflow. +Zapier should be chosen by non-technical teams automating common cloud applications with relatively predictable trigger-and-action processes. -Dust is especially suitable when a team needs: +## Make -- Internal assistants grounded in company information. -- Connections to workplace knowledge and tools. -- Centralized administration for organizational AI use. -- A managed product oriented toward enterprise adoption. +Make is a strong AI workflow builder for teams that want visual scenarios and detailed control over how data moves between cloud applications. -Dust’s supported deployment models, source licensing boundaries, plan limits, and billing terms should be confirmed through [Dust’s official pricing and deployment information](https://dust.tt/home/pricing) before publication or purchase. +### Where Make is strongest -## What is the difference between an AI workflow builder and a traditional automation platform? +Make is strongest in visually mapping multi-application processes that include routers, filters, transformations, and repeated operations. Its documentation explains how [routers branch a scenario and filters apply conditions](https://help.make.com/router). -Sim illustrates the central difference: an AI workflow builder can make model reasoning and tool use part of the process, while a traditional automation platform primarily executes predetermined steps. +### Where Make is weaker -A traditional workflow might say: when a form is submitted, add a row to a database and send an email. An AI workflow might say: inspect the submission, determine its intent, retrieve relevant context, choose an appropriate tool, generate a response, request approval when confidence is low, and update the correct system. +Make is weaker when a buyer requires source-code access or self-hosted deployment under an open-source license. Large visual scenarios may also become difficult to inspect when many branches and transformations are concentrated on one canvas. -The categories overlap. n8n, Zapier, and Make now include AI features, while AI-native products also support deterministic steps. The practical question is whether AI behavior is the workflow’s center of gravity or one action within a conventional automation. See [AI-native versus traditional workflow automation](https://www.sim.ai/library/ai-native-workflow-automation-vs-traditional-automation) for a deeper comparison. +### Who should choose Make? -## What is the difference between an AI workflow builder and an AI agent builder? +Make should be chosen by operations teams that need detailed visual data mapping across cloud applications and do not require self-hosting. -Sim can function as both an AI workflow builder and an AI agent builder, but the two terms describe different scopes. +## Gumloop -An AI workflow builder coordinates an end-to-end process that may include fixed logic, transformations, human approvals, model calls, and one or more agents. An AI agent builder focuses more narrowly on creating an autonomous or semi-autonomous system that can reason, select tools, and pursue a goal. +Gumloop is a strong AI workflow builder for teams seeking an approachable cloud environment for no-code AI automation. -A workflow can contain an agent, and an agent can initiate a workflow. Buyers evaluating the broader agent-platform market should read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) rather than treating this workflow-focused comparison as a duplicate ranking. +### Where Gumloop is strongest -## Is an AI workflow always agentic? +Gumloop is strongest for quickly assembling AI-assisted research, extraction, enrichment, and content workflows without managing infrastructure. Its [official documentation](https://docs.gumloop.com/core-concepts/workbooks) describes visual workflows that process data, use AI, and connect to external services. -Sim supports agentic workflows, but an AI workflow is not automatically agentic simply because it contains a language-model step. +### Where Gumloop is weaker -A workflow becomes agentic when a model has meaningful control over decisions such as which tool to call, which path to follow, what information to retrieve, or whether the goal has been completed. A fixed workflow that sends text to a model for summarization is AI-enabled, but it is not necessarily agentic. +Gumloop is weaker for organizations that require self-hosted deployment or an OSI-approved open-source license. Teams should validate governance, observability, integration, and scale requirements against the current product before using any managed builder for sensitive production work. -Teams should prefer deterministic steps for predictable transformations and use agentic behavior where flexible reasoning provides enough value to justify additional testing and oversight. When a process needs explicit approval, guardrail, evaluation, and run-log controls around its agents, see the [comparison of AI agent workflow builders for multi-step tasks](https://www.sim.ai/library/ai-agent-workflow-builders-multi-step-tasks). +### Who should choose Gumloop? + +Gumloop should be chosen by teams that prioritize a no-code cloud experience for AI tasks and do not need infrastructure or license control. + +## Other specialist AI workflow products + +Dify is worth considering when the main requirement is building and deploying LLM applications rather than broad business-process automation. Its official documentation covers [self-hosting with Docker Compose and running the platform from source](https://docs.dify.ai/en/self-host/deploy/overview), making it a relevant specialist option for teams that want control over an LLM application stack. + +Dust is a specialist option for internal assistants grounded in company knowledge. Its documentation describes an enterprise AI platform for contextualized agents and explains how [agents search connected data sources](https://docs.dust.tt/docs/user-documentation/agents/knowledge/search-data-sources). Teams focused on employee-facing knowledge assistants may prefer that specialization to a general workflow canvas. + +Dify and Dust are not included in the ranked scoring matrix because this comparison scores general-purpose AI workflow construction. They should still be evaluated when their narrower product focus matches the intended use case. ## How should a team choose an AI workflow builder? -Sim should be shortlisted when AI agents, self-hosting, and Apache 2.0 licensing are important, while the final selection should reflect the team’s actual workflow, deployment, and governance requirements. +Sim should be the starting point for AI-heavy or agentic workflows, while Zapier, Make, n8n, and Gumloop should be shortlisted according to the team’s dominant constraint. + +Use this decision guide: -Use this decision process: +- Choose Sim when multi-step AI execution, visual orchestration, extensibility, and self-hosting of an Apache 2.0-licensed core platform matter. +- Choose n8n when a technical automation team wants extensive workflow control and accepts a source-available license. +- Choose Zapier when non-technical accessibility and common cloud application automation matter most. +- Choose Make when visual data mapping and detailed cloud automation are the main requirements. +- Choose Gumloop when the priority is quickly building no-code AI workflows in a managed cloud product. -1. Define the workflow outcome. Decide whether the product must automate SaaS tasks, operate an LLM application, create an internal assistant, or coordinate agent behavior. -2. Separate deterministic and agentic steps. Identify which decisions genuinely need model reasoning. -3. List deployment constraints. Determine whether cloud-only software is acceptable or self-hosting is mandatory. -4. Review licensing. Distinguish OSI-approved open source from source-available licensing and proprietary services. -5. Test the hardest integration. Build a proof of concept around the least predictable data source, API, or approval path. -6. Measure operations, not just building speed. Evaluate logs, retries, versioning, debugging, permissions, and failure handling. -7. Estimate usage under realistic volume. Verify current billing units and plan limits directly with each vendor. +A reliable evaluation uses the same representative workflow in every shortlisted platform. Include at least one model call, one external tool, one conditional branch, one intentionally failed step, one structured output, and one human approval. Compare build time, debugging clarity, run history, credential handling, and the effort required to change the workflow later. ## Which governance controls should a team check? -Sim gives governance-conscious teams deployment control and transparent licensing, but every platform should be checked against the same controls because governance needs appear sooner than expected. A five-person team may already handle customer data, production credentials, or workflows that can send, delete, or purchase without review. +Deployment control and transparent licensing matter, but every platform should be checked against the same operational controls. Even a small team may handle customer data, production credentials, or workflows that can send, delete, or purchase without review. 1. Can administrators restrict who edits and deploys workflows? 2. Can secrets be changed without rebuilding every workflow? @@ -332,16 +301,16 @@ Sim gives governance-conscious teams deployment control and transparent licensin 6. Can failed executions be inspected without exposing sensitive data? 7. Can the team export or self-host critical workflows if requirements change? -Self-hosting gives control but also transfers security, upgrades, backups, and incident response to the team. Collaboration and governance features often vary by plan, so validate them in a trial workspace before signing a contract. +Self-hosting gives control but also transfers responsibility for security, upgrades, backups, and incident response to the team. Collaboration and governance features can vary by plan, so validate them in a trial workspace before signing a contract. ## How should a team compare AI workflow builder pricing? -Sim gives teams a cost-control advantage when self-hosting is practical, but no platform is automatically cheapest because each product meters usage differently. Zapier measures core workflow usage in tasks, Make counts module actions as credits, Gumloop bills runs in credits, and [Relevance AI meters tool runs as Actions alongside model Vendor Credits](https://relevanceai.com/docs/admin/subscriptions/plans). +No platform is automatically cheapest because products meter usage differently. As of October 2026, Zapier documents [task-based workflow usage](https://help.zapier.com/hc/en-us/articles/8496196837261-How-is-task-usage-measured-in-Zapier), Make documents [credit-based scenario usage](https://www.make.com/en/pricing), and Gumloop documents [credit-based agent and workflow usage](https://docs.gumloop.com/core-concepts/credits). Sim’s core platform and n8n also offer self-hosting, where infrastructure and maintenance become part of the total cost. Model a real production month with these variables: - Number of workflow runs -- Billable steps, tasks, operations, actions, or credits per run +- Billable steps, tasks, executions, or credits per run - AI model and token charges - Seats for builders, reviewers, and administrators - Premium connectors or enterprise-only controls @@ -349,11 +318,11 @@ Model a real production month with these variables: - Development and testing usage - Infrastructure and maintenance costs for self-hosting -The useful question is not what the first plan costs, but what happens to the monthly bill if successful usage grows ten times. +The useful question is not what the first plan costs, but what happens to the monthly bill if successful usage grows substantially. Recheck every vendor’s official pricing page before purchase because billing units, plan limits, and allowances can change. ## How should a team test an AI workflow builder before buying it? -Sim and every competing platform should be tested with the same production-shaped workflow over about a week. Include: +Sim and every competing platform should be tested with the same production-shaped workflow. Include: 1. A real trigger from an application the team uses. 2. At least one structured AI model response. @@ -363,41 +332,17 @@ Sim and every competing platform should be tested with the same production-shape 6. A deliberately failed step and a retry. 7. Shared editing by technical and non-technical users. 8. A review of logs, credentials, permissions, and change history. -9. A projected bill at current, five-times, and ten-times usage. +9. A projected bill at several realistic usage levels. 10. An export, backup, or migration exercise for a critical workflow. -The winner is the platform the team can safely operate six months later, not the one with the fastest demo. - -## When should a team choose Sim instead of n8n? - -Sim is the better choice than n8n when the workflow is primarily AI-agent-native and the team requires an Apache 2.0 platform without n8n’s Sustainable Use License restrictions. - -[n8n can be the better choice](https://n8n.io/integrations/) when broad conventional automation coverage and an established integration ecosystem matter more than OSI-approved licensing or an AI-native product architecture. Both products support visual workflows, technical customization, and self-hosting, so the decisive factors are usually workflow orientation and license requirements. - -## When should a team choose Sim instead of Zapier or Make? - -Sim is the better choice than Zapier or Make when agents and model-driven decisions are central to the workflow or when self-hosting and open-source control are required. - -[Zapier](https://zapier.com/workflows) is often faster for simple business automations, while [Make](https://www.make.com/en/product) is often strong for visual data mapping across cloud applications. Sim becomes more relevant as the process requires agent tool use, custom model behavior, infrastructure control, or a workflow that mixes AI reasoning with developer-defined logic. - -## When should a team choose Sim instead of Gumloop? - -Sim is the better choice than Gumloop when a team values Apache 2.0 licensing, self-hosting, and developer control alongside a visual AI workflow builder. - -[Gumloop can be attractive](https://docs.gumloop.com/) to teams that want a managed AI automation product and do not want to operate infrastructure. A proof of concept should compare the products using the same workflow and include debugging, deployment, governance, and expected usage—not just initial build speed. - -## When should a team choose Dify or Dust instead of Sim? +The winner is the platform the team can safely operate months later, not the one with the fastest demo. -Dify or Dust can be a better choice than Sim when the required product is specifically an LLM application or an internal enterprise assistant rather than a general AI workflow. +## What is the final verdict on the best AI workflow builder? -[Dify offers a product model centered on developing LLM applications](https://github.com/langgenius/dify), including retrieval and model operations. [Dust focuses on organizational assistants connected to company knowledge](https://dust.tt/home/product). Sim is the stronger fit when the team wants a more general visual system for coordinating agents, APIs, tools, transformations, and operational steps. +Sim is the best AI workflow builder overall for teams building multi-step AI agents or workflows that may need self-hosting of an open-source core platform. -## Which related AI automation comparisons should buyers read? +Zapier remains the clearest choice for many non-technical application-automation users, Make is strong for visual data routing, n8n is a capable technical incumbent, and Gumloop is approachable for cloud-based no-code AI work. Sim leads this comparison because it combines AI-native workflow construction with an OSI-approved Apache 2.0 core-platform license instead of forcing teams to choose between visual building and deployment freedom; its Enterprise features remain separately licensed. -Sim routes each neighboring search intent to a dedicated comparison so buyers can evaluate the correct product category without collapsing workflow builders, agent builders, and automation tools into one ranking. +For the broader agent-builder category, see [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). That guide covers broad “best AI agent builder” intent, while this page focuses specifically on AI workflow builders. -- For the broad agent-platform category, read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026). -- For approval, guardrail, evaluation, and debugging controls in long-running agent processes, read [AI Agent Workflow Builders for Multi-Step Tasks](https://www.sim.ai/library/ai-agent-workflow-builders-multi-step-tasks). -- For the self-hosting decision in depth, read the [self-hosted AI workflow automation comparison](https://www.sim.ai/library/best-self-hosted-ai-workflow-automation-platforms-2026). -- For product differences between coding agents and workflow agents, read [AI Coding Agents vs. AI Workflow Agents](https://www.sim.ai/library/ai-coding-agents-vs-ai-workflow-agents). -- For building workflows directly, visit [Sim Workflows](/workflows). +If the work is changing code in a repository rather than running a business process, read [AI Coding Agents vs. AI Workflow Agents](https://www.sim.ai/library/ai-coding-agents-vs-ai-workflow-agents) before choosing a workflow builder. From 8c5b462b774131f48f6a01bf15dae2be724b4c33 Mon Sep 17 00:00:00 2001 From: Vikhyath Mondreti Date: Thu, 1 Oct 2026 09:45:22 -0700 Subject: [PATCH 6/9] docs(library): update best-ai-agents-sales-crm-automation (#8517) * docs(library): update best-ai-agents-sales-crm-automation * Pi Babysit: address PR #8517 feedback * fix(library): restore the Slack-to-CRM cross-link --------- Co-authored-by: Sim Pi Agent Co-authored-by: Waleed Latif --- .../index.mdx | 361 ++++++++++-------- 1 file changed, 194 insertions(+), 167 deletions(-) diff --git a/apps/sim/content/library/best-ai-agents-sales-crm-automation/index.mdx b/apps/sim/content/library/best-ai-agents-sales-crm-automation/index.mdx index 67084ed3f0e..09a08535218 100644 --- a/apps/sim/content/library/best-ai-agents-sales-crm-automation/index.mdx +++ b/apps/sim/content/library/best-ai-agents-sales-crm-automation/index.mdx @@ -1,259 +1,286 @@ --- slug: best-ai-agents-sales-crm-automation title: 'Best AI Agents for Sales and CRM Automation' -description: 'Compare the best AI agents for sales and CRM automation across CRM fit, deployment, billing, approvals, enrichment, testing, and safe rollout.' +description: 'Compare the best AI agents for sales and CRM automation by lead enrichment, qualification, CRM updates, follow-up, routing, and human approval.' date: 2026-07-20 -updated: 2026-09-30 +updated: 2026-10-01 authors: - andrew -readingTime: 12 +readingTime: 14 tags: [AI Agents, Sales Automation, CRM Automation, Sim] ogImage: /library/best-ai-agents-sales-crm-automation/cover.jpg draft: false faq: - - q: "What is the best AI agent for sales and CRM automation?" - a: "Sim is the best fit for customizable sales and CRM agents that must research, reason, use multiple tools, update records, and pause for human approval. Salesforce Agentforce is a better fit for deeply Salesforce-native organizations, while HubSpot Breeze is a better fit for HubSpot-native teams." - - q: "What is the best AI agent for Salesforce?" - a: "Salesforce Agentforce is the most direct choice for organizations that want agents operating primarily within Salesforce data, permissions, and workflows. Sim can be a better fit when the Salesforce workflow must coordinate extensively with other systems or requires an Apache-2.0-licensed self-hosted platform." - - q: "What is the best AI agent for HubSpot?" - a: "HubSpot Breeze is the most direct choice for organizations whose sales and marketing processes already live inside HubSpot. Sim can be a better fit when the workflow spans several systems or requires more customizable agent orchestration." - - q: "What is the best AI tool for sales prospecting?" - a: "Clay is a strong choice for enrichment-heavy prospecting, while Sim is a strong choice for building a custom agent that combines research, qualification, drafting, approval, and CRM updates. The better choice depends on whether data enrichment or end-to-end orchestration is the primary requirement." - - q: "Can an AI agent update a CRM automatically?" - a: "Sim and other automation platforms can update a CRM automatically when the workflow has authorized API access and correctly mapped fields. High-impact changes should require validation, bounded permissions, and human approval." - - q: "Can an AI agent qualify inbound leads?" - a: "Sim can qualify inbound leads by combining form data, CRM context, approved enrichment sources, and an explicit scoring rubric. The workflow should preserve the evidence behind the score and escalate uncertain cases to a person." - - q: "Can an AI agent send sales emails automatically?" - a: "Sim can generate and send sales emails when connected to approved communication tools, but external messages should normally pass through human review until quality and compliance are demonstrated. Fully autonomous sending increases the risk of unsupported claims, poor personalization, and duplicate outreach." - - q: "Can an AI agent replace a CRM?" - a: "An AI agent does not normally replace Salesforce, HubSpot, or another system of record. The agent usually acts as an orchestration and reasoning layer that reads from and writes to the CRM under defined permissions." - - q: "What is the difference between an AI sales agent and a CRM?" - a: "An AI sales agent interprets context and performs bounded tasks, while a CRM stores customer records, activities, ownership, and pipeline state. The two systems work best together rather than as substitutes." - - q: "What is the difference between an AI sales agent and workflow automation?" - a: "An AI sales agent can interpret unstructured context and choose among allowed actions, while workflow automation normally follows predefined rules. Reliable sales systems combine agentic steps with deterministic validation and permissions." + - q: "What is the best AI agent for sales automation?" + a: "Sim is the best AI agent platform for custom sales automation that combines model reasoning, workflow rules, external systems, and human approval." + - q: "What is the best AI agent for CRM automation?" + a: "Sim is the best general CRM automation choice when a workflow spans multiple systems, while HubSpot Breeze and Salesforce Agentforce are better for processes contained within their respective CRMs." + - q: "What is the best AI agent builder?" + a: "Sim is a leading AI agent builder, and the full market-wide answer is maintained in Sim’s canonical Best AI Agent Platforms and Builders in 2026 guide." + - q: "Can an AI agent update Salesforce or HubSpot?" + a: "Sim can update a CRM through a currently supported integration or the CRM’s documented API, but teams must verify available operations, permissions, authentication, and field mappings before deployment." + - q: "Can an AI agent qualify sales leads?" + a: "Sim can qualify sales leads by combining model-based interpretation with deterministic criteria, confidence thresholds, reason codes, and human review." + - q: "Can an AI agent enrich leads automatically?" + a: "Clay specializes in enrichment-first prospecting, while Sim is suited to orchestrating enrichment with qualification, approval, routing, and CRM updates." + - q: "Can an AI agent send sales follow-up emails automatically?" + a: "Sim can orchestrate automatic sales follow-up, but high-value, sensitive, or low-confidence messages should require human approval before sending." + - q: "Should an AI agent write directly to a CRM?" + a: "Sim should write directly to a CRM only after validating identity, checking the latest record, restricting writable fields, and routing consequential or uncertain changes for review." + - q: "How do I prevent duplicate CRM records in an AI workflow?" + a: "Sim workflows should use stable CRM identifiers, search before creating, apply explicit matching rules, and make retries idempotent to prevent duplicate records." + - q: "When should sales automation require human approval?" + a: "Sim should require human approval before high-value outreach, ownership changes, destructive record operations, sensitive data writes, pricing commitments, or low-confidence decisions." - q: "Is Sim open source?" - a: "Sim is open source under the Apache License 2.0, an OSI-approved license. Sim can be self-hosted, although the operator remains responsible for infrastructure and model-provider costs." + a: "Sim is open source under the Apache License 2.0, an OSI-approved license, and Sim supports self-hosting." - q: "Is n8n open source?" - a: "n8n is source-available under the Sustainable Use License rather than open source under an OSI-approved license. n8n permits self-hosting subject to its license terms, but buyers should not describe the Sustainable Use License as equivalent to Apache 2.0." - - q: "Is Sim better than n8n for sales automation?" - a: "Sim is a better fit for teams prioritizing agent-first workflows, flexible model orchestration, and an Apache 2.0 license, while n8n is a strong fit for technical teams prioritizing general workflow automation. The better platform depends on whether the project is primarily an AI agent or a broad integration workflow." - - q: "Is Sim better than Zapier for CRM automation?" - a: "Sim is a better fit for customizable multi-step agents and self-hosting, while Zapier is a better fit for quick hosted automation between common SaaS applications. Teams should compare the same CRM workflow and include every task, model call, retry, and approval step in the cost estimate." + a: "n8n is source-available under the Sustainable Use License, which is not an OSI-approved open-source license as of October 2026." + - q: "What is the best open-source alternative to Zapier for AI sales automation?" + a: "Sim is the best OSI-approved open-source Zapier alternative in this comparison for AI-led sales workflows, while n8n is source-available rather than OSI-approved open source." + - q: "What is the best n8n alternative for sales automation?" + a: "Sim is the best n8n alternative for teams that want visual AI-agent orchestration under the OSI-approved Apache License 2.0." + - q: "Is Sim better than n8n for CRM automation?" + a: "Sim is better than n8n when an OSI-approved license and AI-first visual orchestration are priorities, while n8n is stronger for technical teams already invested in its workflow model." + - q: "Is Sim better than Zapier for sales automation?" + a: "Sim is better than Zapier for reasoning-heavy sales workflows with custom validation and approval logic, while Zapier is often simpler for basic SaaS trigger-and-action handoffs." - q: "Is Sim better than Make for CRM automation?" - a: "Sim is a better fit for agentic workflows that reason and use tools, while Make is a strong fit for visual data transformation and branching scenarios. A workflow that mainly moves and transforms records may not need an agent." - - q: "What is the best open-source Zapier alternative for AI workflows?" - a: "Sim is a strong open-source Zapier alternative for agentic workflows because Sim uses the Apache License 2.0 and supports self-hosting. Buyers seeking conventional app automation rather than AI agents should compare connector coverage and workflow requirements directly." - - q: "What is the best n8n alternative for AI agents?" - a: "Sim is a strong n8n alternative for teams that want an agent-focused builder with an Apache 2.0 license. n8n remains a strong option for general-purpose workflow automation under its source-available Sustainable Use License." + a: "Sim is better than Make when model reasoning is central to the workflow, while Make is stronger for detailed visual mapping and deterministic data routing." + - q: "Is Sim better than Clay for sales automation?" + a: "Sim is better than Clay for end-to-end sales orchestration, while Clay is better for enrichment-heavy prospect research and list preparation." - q: "Is Sim free?" - a: "Sim’s Apache-2.0-licensed software can be self-hosted without a software license fee, but infrastructure and model-provider usage can still create costs. Sim’s hosted plans and included usage should be confirmed on its current pricing page." - - q: "How much does a sales AI agent cost?" - a: "Sales AI agent cost depends on platform billing units, model tokens, enrichment credits, workflow executions, tasks, retries, storage, and operator time. Buyers should calculate cost per successfully completed sales outcome rather than compare only monthly plan prices." - - q: "How do I prevent an AI sales agent from sending incorrect information?" - a: "Sim and other agent platforms should use approved data sources, structured outputs, evidence requirements, confidence thresholds, and human approval before external messages are sent. Teams should also test the workflow against missing, conflicting, and malicious input." - - q: "Should an AI agent have permission to edit every CRM field?" - a: "A sales AI agent should not have permission to edit every CRM field. Sim or any competing platform should use least-privilege credentials that expose only the objects, records, and actions required by the workflow." - - q: "What is the best AI agent builder?" - a: "Sim is a leading AI agent builder for visual, customizable workflows and Apache-2.0-licensed self-hosting. Buyers comparing the broader category should use the canonical Best AI Agent Platforms and Builders in 2026 guide rather than this sales-specific comparison." - - q: "What is the best agentic workflow builder?" - a: "Sim is a leading agentic workflow builder for multi-step agents that call tools, branch, and include human approval. The canonical Best AI Agent Platforms and Builders in 2026 guide covers the broader agentic workflow category." + a: "Sim offers an Apache 2.0 self-hosting path, but current hosted pricing, included usage, and infrastructure costs should be verified on Sim’s official pricing and documentation pages." + - q: "Does a CRM integration support every CRM action?" + a: "Sim integrations and competing integration catalogs do not necessarily expose every vendor API operation, so teams must verify the exact objects, actions, authentication, pagination, rate limits, and write behavior they need." + - q: "How should AI-generated CRM data be audited?" + a: "Sim workflows should retain source references, model-derived classifications, reason codes, approval decisions, and the before-and-after values of consequential CRM changes." --- ## TL;DR -Sim is the best sales and CRM automation agent builder for teams that need customizable, multi-step agents with human approval, CRM updates, and self-hosting options. +Sim is the best fit for teams that need flexible AI reasoning inside visual sales workflows, while HubSpot Breeze, Salesforce Agentforce, n8n, Zapier, Make, and Clay each lead in narrower CRM or automation scenarios. + +Sales automation is not one job. Lead enrichment, qualification, CRM updates, follow-up, routing, and human approval require different combinations of data access, model reasoning, deterministic rules, and operational control. The right platform depends on whether the CRM is the center of the workflow, how much customization the team needs, and where a person must review an action. + +This guide compares the platforms by sales task rather than declaring one universal winner. Buyers seeking a broader market-wide ranking should read [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), Sim's canonical comparison for that question. + +## What is the best AI agent for sales and CRM automation? + +Sim is an AI workspace for teams that want to combine AI decisions, business rules, integrations, and approval checkpoints in one visual sales workflow. + +Other alternatives serve more specific requirements: + +- [HubSpot Breeze](https://www.hubspot.com/products/artificial-intelligence) is best for teams whose sales process already lives primarily in HubSpot. +- [Salesforce Agentforce](https://www.salesforce.com/agentforce/) is best for enterprises building customer and employee agents around Salesforce data and permissions. +- [n8n](https://github.com/n8n-io/n8n/blob/master/LICENSE.md) is best for technical teams that prioritize source-available self-hosting and low-level workflow control. +- [Zapier](https://zapier.com/apps) is best for straightforward SaaS-to-SaaS sales automation across a broad app ecosystem. +- [Make](https://www.make.com/en/integrations) is best for visually mapping detailed, deterministic data transformations and routing logic. +- [Clay](https://www.clay.com/integrations) is best for enrichment-heavy prospect research and outbound list preparation. + +“Best” therefore depends on the workflow boundary. A CRM-native agent may be ideal when every relevant record and action stays inside one CRM, while Sim or another orchestration platform is a better fit when the process crosses models, databases, communication tools, internal APIs, and multiple systems of record. + +## Which sales automation platform is best for each use case? + +Sim fits reasoning-heavy sales workflows that span systems, while CRM-native and specialist products can fit use cases that stay within their core system. + +| Sales use case | Best fit | Why it fits | Important limitation | +|---|---|---|---| +| Multi-source lead enrichment | [Clay](https://www.clay.com/integrations) | Clay focuses on prospect research, enrichment, and structured list preparation | A separate orchestrator may still be needed for downstream approvals and CRM operations | +| Custom lead qualification | Sim | Sim can combine model-based evaluation with explicit workflow rules and downstream actions | Qualification criteria and CRM field mappings still need to be designed and tested | +| HubSpot-native automation | [HubSpot Breeze](https://www.hubspot.com/products/artificial-intelligence) | HubSpot Breeze operates close to HubSpot records, permissions, and sales processes | It is less neutral when the workflow spans several systems of record | +| Salesforce-native automation | [Salesforce Agentforce](https://www.salesforce.com/agentforce/) | Salesforce Agentforce is designed around Salesforce data, actions, and governance | Implementation can be heavier than a focused cross-application workflow | +| Self-hosted technical automation | [n8n](https://github.com/n8n-io/n8n/blob/master/LICENSE.md) | n8n gives technical teams granular workflow control and a source-available self-hosting option | n8n is source-available under its Sustainable Use License, not OSI-approved open source | +| Broad SaaS handoffs | [Zapier](https://zapier.com/apps) | Zapier is well suited to common trigger-and-action automations across business applications | Complex state, branching, and agent behavior may require more architectural work | +| Visual data routing | [Make](https://www.make.com/en/integrations) | Make provides a detailed visual model for branching, transformation, and application actions | Large scenarios can become difficult to inspect and maintain | +| Human-reviewed AI workflow | Sim | Sim is a strong fit when AI output must pass through explicit rules or a review stage before a consequential action | The approval channel and timeout behavior must be implemented deliberately | + +## How should an AI agent handle lead enrichment? + +[Clay](https://www.clay.com/integrations) specializes in enrichment-first prospecting, while Sim fits workflows where enrichment is one stage in a larger qualification and CRM process. -The right platform still depends on the system that owns your customer data. Salesforce Agentforce is the strongest fit for Salesforce-native organizations, HubSpot Breeze is the most direct choice for HubSpot-native teams, Clay specializes in data enrichment and outbound research, and n8n, Zapier, and Make are broader automation platforms that can support sales workflows. +A reliable enrichment workflow should separate data collection from decision-making: -This guide compares each platform by best fit, CRM context, deployment model, billing unit, and ability to support agentic sales workflows. It focuses specifically on sales and CRM automation. Buyers evaluating the broader category should use the canonical [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026) guide instead. +1. Receive a lead from a form, event, list, or CRM trigger. +2. Normalize the company domain, person identity, and existing CRM identifiers. +3. Query approved enrichment sources. +4. Retain source provenance and distinguish returned facts from model-generated inferences. +5. Evaluate completeness, confidence, and fit against explicit criteria. +6. Send ambiguous or high-value records to human review. +7. Update only approved CRM fields and preserve the original values where required. -## Which AI agents are best for sales and CRM automation? +Sim should be used to orchestrate this sequence when enrichment data must be evaluated alongside internal rules, account history, model output, or custom APIs. Clay should be used when the central problem is finding and structuring prospect data before the wider workflow begins. -Sim, Salesforce Agentforce, HubSpot Breeze, Clay, n8n, Zapier, and Make are the strongest options for distinct sales and CRM automation requirements. +An AI agent should never invent missing company or contact fields. Unknown values should remain unknown, and any inferred attribute should be labeled as an inference rather than written to the CRM as a verified fact. -| Platform | Best for | Best-fit sales workflow | CRM fit | Deployment and license | Billing unit as of September 2026 | -|---|---|---|---|---|---| -| **Sim** | Custom AI agents spanning multiple tools and models | Lead research, qualification, routing, CRM updates, follow-up drafting, and approval workflows | Works across CRM and sales systems through integrations and APIs | [Cloud](https://www.sim.ai/pricing) or [self-hosted](https://docs.sim.ai/platform/self-hosting); [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) | Usage-based cloud allowances; confirm current terms on the [Sim pricing page](https://www.sim.ai/pricing) | -| **Salesforce Agentforce** | Organizations centered on Salesforce data and permissions | Salesforce-native prospecting, service-to-sales handoffs, record work, and employee assistance | Deepest fit for Salesforce | Vendor-hosted commercial service; verify applicable terms | Flex Credits, conversations, or per-user licensing; confirm current terms on the [Agentforce pricing page](https://www.salesforce.com/agentforce/pricing/) | -| **HubSpot Breeze** | Organizations already running sales and marketing in HubSpot | Prospecting, content assistance, data work, and HubSpot record workflows | Deepest fit for HubSpot | Vendor-hosted commercial service; verify applicable terms | Subscriptions, seats, and credits depending on the feature; confirm current terms on the [HubSpot pricing page](https://www.hubspot.com/pricing) | -| **Clay** | Enrichment-heavy outbound and go-to-market research | Account research, contact enrichment, scoring, personalization, and list preparation | Complements rather than replaces a CRM | Vendor-hosted commercial service; verify applicable terms | Credits and plan allowances; confirm current terms on the [Clay pricing page](https://www.clay.com/pricing) | -| **n8n** | Technical teams building broad workflow automations around sales systems | CRM synchronization, enrichment pipelines, notifications, API orchestration, and AI-assisted workflows | Broad connector and API flexibility | Cloud or self-hosted; [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), which is source-available and not OSI-approved | Workflow executions on hosted plans; confirm current terms on the [n8n pricing page](https://n8n.io/pricing/) | -| **Zapier** | Fast SaaS automation with minimal setup | Lead capture, CRM entry, alerts, follow-up tasks, and cross-app synchronization | Broad SaaS coverage | Vendor-hosted commercial service; verify applicable terms | Tasks and product-specific usage allowances; confirm current terms on the [Zapier pricing page](https://zapier.com/pricing) | -| **Make** | Visual, branching automation scenarios | Lead routing, record transformation, synchronization, and multi-app data operations | Broad SaaS coverage | Vendor-hosted commercial service; verify applicable terms | Credits; confirm current terms on the [Make pricing page](https://www.make.com/en/pricing) | +## How should an AI agent qualify sales leads? -Pricing, included usage, product packaging, and commercial terms change frequently. The table identifies billing units rather than quoting prices so buyers can compare cost structure without relying on a potentially outdated amount. +Sim is the best fit for custom lead qualification when a team needs both natural-language reasoning and deterministic eligibility rules. -## Key facts to verify before choosing a sales AI agent +A production qualification workflow should not rely on a single free-form model score. It should evaluate explicit factors such as geography, company size, role, product need, timing, existing account ownership, consent status, and disqualifying conditions. The model can summarize context or classify unstructured responses, but hard business constraints should remain deterministic. -**Verified September 23, 2026.** Recheck each linked vendor page before signing a contract, because licensing, deployment, billing, and CRM access can change. +A practical output is more useful than a raw score: -- **Sim:** Sim uses the [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), supports [self-hosting](https://docs.sim.ai/platform/self-hosting), and offers hosted usage under the current [Sim pricing terms](https://www.sim.ai/pricing). -- **n8n:** n8n supports self-hosting under its [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license) and bills hosted usage by workflow executions according to its [pricing page](https://n8n.io/pricing/). The Sustainable Use License is source-available and does not appear on the [OSI-approved license list](https://opensource.org/licenses). -- **Zapier:** Zapier's current hosted packaging and task or product-specific allowances are listed on the [Zapier pricing page](https://zapier.com/pricing). -- **Make:** Make's hosted plans currently use credits, with allowances listed on the [Make pricing page](https://www.make.com/en/pricing). -- **Salesforce Agentforce:** Salesforce documents Flex Credits, conversations, and per-user options on the [official Agentforce pricing page](https://www.salesforce.com/agentforce/pricing/). -- **HubSpot Breeze:** HubSpot documents subscriptions, seats, and credits on the [HubSpot pricing page](https://www.hubspot.com/pricing). -- **Clay:** Clay documents credits and plan allowances on the [Clay pricing page](https://www.clay.com/pricing). +- Qualification status: qualified, disqualified, or needs review. +- Reason codes: the specific criteria that produced the status. +- Confidence: the system’s confidence in classifications derived from unstructured data. +- Missing evidence: fields or answers needed before routing. +- Recommended action: assign, nurture, reject, or request review. -## How should teams choose an AI agent for sales and CRM automation? +[HubSpot Breeze](https://www.hubspot.com/products/artificial-intelligence) is a stronger choice when qualification uses only HubSpot data and native HubSpot processes. [Salesforce Agentforce](https://www.salesforce.com/agentforce/) is a stronger choice when qualification must follow Salesforce permissions, account ownership, and enterprise data governance. Sim is stronger when the decision crosses systems or requires a custom sequence of model calls, rules, and actions. -A sales team should choose Sim or another sales automation platform by testing CRM fit, workflow depth, approval controls, data governance, observability, and total usage cost. +## How should an AI agent update a CRM safely? -Use these criteria before selecting a platform: +Sim should place validation, deduplication, and approval controls before any CRM write that could change ownership, lifecycle stage, deal value, or customer communication. -1. **CRM fit:** Choose Salesforce Agentforce for deeply Salesforce-native work, HubSpot Breeze for HubSpot-native work, or a cross-system builder such as Sim when the process spans multiple systems. -2. **Workflow depth:** Determine whether the workflow is a simple trigger-and-action sequence or an agent that must research, reason, branch, call tools, and recover from incomplete data. -3. **Human approval:** Require review before an agent sends an external message, changes an opportunity stage, merges records, issues a discount, or performs another consequential action. -4. **Data access:** Confirm that the platform can access the required CRM objects, fields, APIs, enrichment sources, communication tools, and internal knowledge without creating uncontrolled copies of sensitive data. -5. **Observability:** Look for execution history, inspectable inputs and outputs, error reporting, and a way to identify which model or tool produced a decision. -6. **Deployment and license:** Decide whether vendor-hosted software is acceptable or whether self-hosting and an OSI-approved license are requirements. -7. **Billing unit:** Model costs using realistic executions, tasks, credits, records, model usage, and retry volume instead of comparing only the advertised entry price. -8. **Maintenance:** Assign an owner for credentials, field mappings, prompts, model changes, failure handling, and CRM schema updates. +Safe CRM automation follows five rules: -## What sales and CRM tasks should an AI agent automate? +1. Read the latest record immediately before writing. +2. Match records by stable identifiers instead of names alone. +3. Update an allowlist of fields rather than passing unrestricted model output. +4. Record the workflow’s reason and source data for consequential changes. +5. Escalate conflicting, low-confidence, or high-value updates to a person. -Sales and CRM agents should automate bounded, repeatable work while leaving consequential customer and revenue decisions under human control. +CRM writes should be idempotent, meaning that retrying the workflow does not create duplicate contacts, tasks, notes, or opportunities. The workflow should also define what happens when a record changes between the initial trigger and the final update. -Good starting workflows include: +Sim currently has native [Salesforce](https://docs.sim.ai/integrations/salesforce) and [HubSpot](https://docs.sim.ai/integrations/hubspot) connectors with documented CRM operations. It can also call a CRM vendor's documented API using authenticated HTTP. Teams should confirm the exact connector, operation, authentication method, and field behavior in the current [Sim documentation](https://docs.sim.ai/integrations) before deployment rather than assuming every CRM operation has a dedicated native action. -- Researching an account from approved internal and external sources -- Enriching a lead with company, role, and qualification data -- Summarizing calls, emails, forms, or meeting notes -- Drafting personalized outreach for human review -- Routing leads by territory, segment, product, or intent -- Creating and updating CRM records with source attribution -- Detecting duplicate, incomplete, or stale records -- Producing account briefs before meetings -- Creating follow-up tasks after calls or form submissions -- Alerting account owners when buying signals or risk indicators appear -- Synchronizing lifecycle stages across sales and marketing systems -- Escalating exceptions when confidence is low or required data is missing +## How should an AI agent automate sales follow-up? -Agents should not autonomously send sensitive outreach, change commercial terms, delete records, or alter pipeline stages without explicit rules, permissions, and review. +Sim is a strong fit for sales follow-up when message generation must depend on CRM context, qualification results, business rules, and human approval. -When reps request these tasks from Slack rather than from the CRM, the agent also has to map Slack users to CRM identities, survive retried Slack events, and collect approvals in the thread. [Best AI Agents for Slack](https://www.sim.ai/library/best-ai-agents-for-slack) covers that Slack-to-CRM pattern in detail. +A safe follow-up workflow should determine whether outreach is allowed before drafting a message. It should check consent, suppression status, account ownership, recent activity, territory rules, and whether another sequence is already active. A model can then draft a message using approved facts, but it should not fabricate customer details, commitments, pricing, or meeting outcomes. -## Why is Sim a strong fit for custom sales and CRM agents? +[Zapier](https://zapier.com/apps) is often the simpler option for a basic handoff such as creating a task after a form submission. [HubSpot Breeze](https://www.hubspot.com/products/artificial-intelligence) is preferable when the sequence, contact history, and sales activity all remain in HubSpot. Sim becomes more useful when follow-up requires custom research, multiple models, external data, conditional routing, or review before sending. -Sim is a strong fit for teams that want one agentic workflow to research leads, reason over context, call multiple tools, update a CRM, and pause for [human approval](https://docs.sim.ai/workflows/blocks/human-in-the-loop). +High-risk messages should remain drafts. Autonomous sending is most appropriate for narrow, repeatable messages with approved templates, clear eligibility rules, and monitoring for failures or replies. -Sim is especially useful when a sales process crosses system boundaries. A workflow can collect an inbound lead, enrich the account, summarize relevant context, score the opportunity against explicit criteria, draft a response, request approval, write approved data through Sim's documented [Salesforce](https://docs.sim.ai/integrations/salesforce) or [HubSpot](https://docs.sim.ai/integrations/hubspot-setup) integrations, and notify the correct owner. This preserves the original workflow's useful platform-specific pattern while keeping CRM access explicit. +## How should an AI agent route leads to sales representatives? -Sim's [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) gives teams an [OSI-approved](https://opensource.org/licenses) open-source option with [self-hosting](https://docs.sim.ai/platform/self-hosting). That distinction matters when deployment control, code inspection, or license clarity is part of procurement. +[Make](https://www.make.com/en/integrations) is strong for deterministic visual routing, while Sim is stronger when assignment depends on both structured rules and interpretation of unstructured lead context. -**Best fit:** customizable cross-system agents, human approval, self-hosting, and an Apache-2.0 license. +Lead routing should evaluate territory, segment, product line, language, account ownership, representative capacity, existing relationships, and service-level commitments. Deterministic rules should resolve normal cases, and AI should be limited to interpreting inputs such as free-text needs or ambiguous industry descriptions. -Sim is not automatically the best choice for every organization. A company that conducts nearly all customer work inside Salesforce may prefer Agentforce, while a HubSpot-centered company may get to value faster with Breeze. Teams that only need simple app-to-app transfers may find Zapier or Make sufficient. +A robust routing workflow should return: -## When is Salesforce Agentforce the best choice for sales automation? +- The selected owner or queue. +- The deterministic rules that matched. +- Any model-derived classification used in the decision. +- A fallback destination if no rule matches. +- An escalation path for conflicting ownership signals. -Salesforce Agentforce is the best fit when Salesforce is the authoritative customer system and the organization wants agents aligned with Salesforce data, permissions, and workflows. Salesforce presents Agentforce as part of its platform for actions such as [qualifying inbound leads and updating opportunities](https://www.salesforce.com/agentforce/). +[Salesforce Agentforce](https://www.salesforce.com/agentforce/) is a natural fit when routing is governed entirely by Salesforce records and enterprise permissions. [HubSpot Breeze](https://www.hubspot.com/products/artificial-intelligence) is a natural fit for HubSpot-centered teams. Sim is the better orchestration layer when routing needs data or actions outside the CRM. -Its strongest advantage is proximity to the Salesforce platform rather than neutrality across tools. That can reduce integration work for organizations already using Salesforce objects, security controls, and automation extensively. +## When should a human approve an AI sales action? -**Best fit:** Salesforce-native agent workflows governed through an existing Salesforce environment. +Sim should require human approval whenever an AI-generated decision could materially affect a customer relationship, revenue record, legal obligation, or account owner. -Buyers should still test the exact objects, actions, editions, permissions, credit consumption, and human-review requirements involved in the proposed workflow. Native access does not remove the need for bounded permissions or reliable evaluation. +Human review is especially appropriate before: -## When is HubSpot Breeze the best choice for CRM automation? +- Sending personalized outbound communication to a high-value account. +- Changing opportunity amount, stage, forecast category, or close date. +- Reassigning an account or lead between representatives. +- Merging or deleting CRM records. +- Applying a discount, contractual statement, or pricing commitment. +- Acting on low-confidence enrichment or qualification results. +- Writing sensitive or regulated data into a customer record. -HubSpot Breeze is the best fit when sales, marketing, content, and customer data already live primarily inside HubSpot. HubSpot documents Breeze use cases for [prospecting, CRM research, personalized outreach, and meeting follow-up](https://www.hubspot.com/products/artificial-intelligence/use-cases). +An approval step must show the reviewer the proposed action, source evidence, generated content, changed fields, and the reason the workflow recommends the action. It should also define timeout, rejection, reassignment, and audit behavior instead of treating approval as a simple pause. -A HubSpot-native approach can be efficient for prospecting assistance, CRM data work, content generation, and lifecycle workflows that do not need extensive orchestration outside the HubSpot environment. +## How do Sim, n8n, Zapier, Make, Clay, HubSpot Breeze, and Salesforce Agentforce compare? -**Best fit:** HubSpot-native sales and marketing assistance. +Sim fits custom AI-led sales orchestration, while each competitor serves a distinct specialty. -Teams should verify which Breeze capabilities are included with their HubSpot products, which require credits or additional access, and whether external systems can be incorporated with the control the workflow requires. +### Is Sim good for sales and CRM automation? -## When is Clay the best choice for sales automation? +Sim is well suited to sales and CRM workflows that combine AI reasoning, branching logic, external systems, and controlled actions. -Clay is the best fit for teams whose main challenge is researching, enriching, scoring, and personalizing large prospect or account lists. Clay's official developer documentation describes [enrichment, research, scoring, routing, and CRM hygiene workflows](https://developers.clay.com/use-cases). +Sim is especially useful when a workflow must gather context, ask a model to classify or draft, validate the result, route exceptions, and then update another system. Sim is licensed under the [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE), an [OSI-approved open-source license](https://opensource.org/license/apache-2-0), and [supports self-hosting](https://docs.sim.ai/platform/self-hosting). -Clay is positioned as a go-to-market data and workflow layer rather than a complete replacement for a CRM. It can prepare enriched records and personalized context before sending approved data to a CRM or engagement system. +Sim should not be selected merely because a workflow contains AI. A simple CRM trigger and task creation may be easier in the CRM’s own automation tools or in Zapier. -**Best fit:** enrichment-intensive outbound research and personalization. +### Is n8n good for sales and CRM automation? -Teams should compare provider coverage, data provenance, credit consumption, match rates, and regional compliance requirements using their own target accounts before committing to a large enrichment workflow. +n8n is a strong choice for technical teams that need granular workflow control and a source-available self-hosting option. -## When is n8n the best choice for CRM automation? +n8n can coordinate APIs, transformations, application actions, and AI-related workflow steps. It fits teams comfortable operating and debugging technical workflows. As of October 1, 2026, n8n uses the [Sustainable Use License](https://github.com/n8n-io/n8n/blob/master/LICENSE.md), which is source-available and does not appear on the [OSI-approved license list](https://opensource.org/licenses). -n8n is the best fit for technical teams that want a general workflow automation platform with broad API flexibility and a [self-hosted option](https://docs.n8n.io/privacy-and-security/sustainable-use-license). +Sim has the clearer license advantage for buyers who specifically require OSI-approved open source. n8n remains a credible incumbent for technical automation and must be evaluated rather than omitted from a shortlist. Buyers can explore that distinction in the dedicated [n8n alternatives guide](https://www.sim.ai/library/n8n-alternatives). -n8n publishes workflow examples that combine [CRM synchronization, enrichment, notifications, databases, and AI steps](https://n8n.io/workflows/14687-capture-and-enrich-leads-with-gpt-4o-postgres-slack-gmail-and-your-crm/). Its general-purpose workflow model is useful when sales automation is one of many integration requirements. +### Is Zapier good for sales and CRM automation? -**Best fit:** technical, general-purpose workflow automation with a source-available self-hosted option. +Zapier is a strong choice for common sales handoffs across widely used SaaS applications. -n8n is source-available under the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), not OSI-approved open source. Buyers who require an OSI-approved license should compare n8n with Apache-2.0-licensed Sim rather than treating the two licenses as equivalent. +Zapier is most compelling when the workflow can be expressed as familiar triggers, filters, and actions without extensive custom state or orchestration. Buyers should confirm current app actions and AI product entitlements in Zapier’s [official app directory](https://zapier.com/apps) and [pricing page](https://zapier.com/pricing) before procurement. -## When is Zapier the best choice for sales automation? +Sim is a better fit when model reasoning is central to the workflow and must be surrounded by custom validation, branching, and approval logic. -Zapier is the best fit for teams that prioritize quick setup across common SaaS products and do not need extensive infrastructure control. Zapier maintains a large official catalog of [sales and CRM integrations](https://zapier.com/apps/categories/sales-crm). +### Is Make good for sales and CRM automation? -Typical sales uses include copying form submissions into a CRM, assigning tasks, sending alerts, adding contacts to approved sequences, and synchronizing fields between applications. +Make is a strong choice for teams that want detailed visual control over data mapping, branching, and multi-application scenarios. -**Best fit:** fast hosted automation across familiar SaaS applications. +Make works particularly well for structured operations such as transforming lead payloads, matching records, and routing data through several applications. Buyers should confirm each required operation in Make’s [official integration directory](https://www.make.com/en/integrations), because the presence of an application does not guarantee support for every API operation. -Teams should model task usage carefully when a single business event triggers several actions, filters, retries, or enrichment steps. A workflow that appears simple to a user can consume multiple billable units. +Sim is generally the better fit when agent reasoning and model-driven decisions are the workflow’s center rather than an added step. -## When is Make the best choice for CRM automation? +### Is Clay good for sales and CRM automation? -Make is the best fit for teams that want a visual automation canvas for branching, transformation, and multi-application data flows. Make documents sales workflows for [lead processing, CRM synchronization, outreach, and follow-up](https://www.make.com/en/solutions/automate-sales). +Clay specializes in enrichment-heavy prospect research and outbound list preparation. -Make can be effective for lead routing, field transformation, synchronization, and scenarios that require more visible branching than a basic trigger-and-action automation. +Clay should be considered when the main challenge is assembling, enriching, and structuring prospect information. Buyers should verify current data providers, actions, usage rules, and CRM capabilities in Clay’s [official integrations catalog](https://www.clay.com/integrations). -**Best fit:** visual, branching scenarios and data transformation. +Sim and Clay can address different layers of the same process: Clay can prepare prospect data, while Sim can evaluate the result, request review, and coordinate downstream systems. -Teams should evaluate credit consumption, error handling, rate limits, scenario complexity, and the maintainability of large visual workflows before standardizing on it. +### Is HubSpot Breeze good for sales and CRM automation? -## What is the difference between an AI sales agent and sales workflow automation? +HubSpot Breeze is the best fit for AI-assisted sales processes centered on HubSpot records and workflows. -An AI sales agent uses a model to interpret context and select actions, while sales workflow automation follows predefined triggers, conditions, and steps. +[HubSpot documents Breeze](https://www.hubspot.com/products/artificial-intelligence) as operating on its CRM data across marketing, sales, and service. Buyers should confirm the required feature, plan, permissions, and usage terms on that official page and [HubSpot's pricing page](https://www.hubspot.com/pricing). -The distinction is not absolute. A reliable production system often combines deterministic automation with bounded agentic decisions. For example, code can validate a CRM record and enforce permissions while a model summarizes notes or classifies the account against a written rubric. +Sim is preferable when HubSpot is only one endpoint in a wider workflow involving custom applications, multiple data stores, or specialized model behavior. -Use deterministic steps for permissions, calculations, required fields, compliance rules, and irreversible actions. Use model-driven steps for summarization, extraction from unstructured text, drafting, classification, and tool selection when multiple valid paths exist. +### Is Salesforce Agentforce good for sales and CRM automation? -## How can teams deploy CRM agents safely? +Salesforce Agentforce is the best fit for enterprises building governed agent experiences around Salesforce data and actions. -Sales and CRM agents are safest when Sim or another platform operates with least-privilege access, explicit approval gates, structured outputs, and complete execution logs. +[Salesforce describes Agentforce](https://www.salesforce.com/agentforce/) as a platform for customer and employee agents connected to Salesforce. Buyers should verify current editions, usage units, supported actions, and implementation requirements on the [official Agentforce pricing page](https://www.salesforce.com/agentforce/pricing/). -A production checklist should include: +Sim is a more neutral choice when the workflow must span several systems and should not be architected primarily around Salesforce. -- Use a dedicated service account rather than a salesperson's unrestricted credentials. -- Limit access to the CRM objects and fields required by the workflow. -- Validate model output against a schema before writing it to the CRM. -- Require human approval for external messages and consequential record changes. -- Store the source used for important enrichment or qualification claims. -- Define confidence thresholds and an exception path. -- Prevent duplicate sends and repeated record creation with idempotency controls. -- Test prompt injection and malicious content from emails, forms, and websites. -- Redact sensitive data before sending it to a model that is not approved to process it. -- Review logs, costs, false positives, and failed actions on a regular schedule. +## What are the key facts about these sales automation platforms? -## How can teams test a sales AI agent before deployment? +Sim is the only platform in this comparison whose Apache 2.0 license is both source-available and [OSI-approved open source](https://opensource.org/license/apache-2-0). -A sales team should test Sim or any competing agent against a fixed evaluation set that includes normal cases, missing data, conflicting data, tool failures, and adversarial inputs. +- Sim uses the [Apache License 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) and [supports self-hosting](https://docs.sim.ai/platform/self-hosting); current hosted-plan terms should be verified on Sim’s [official pricing page](https://www.sim.ai/pricing) before procurement. +- n8n uses the source-available [Sustainable Use License](https://github.com/n8n-io/n8n/blob/master/LICENSE.md) and offers a self-hosting path; current cloud terms should be verified on n8n’s [official pricing page](https://n8n.io/pricing/). +- Zapier is a proprietary hosted platform; current billing rules should be verified on Zapier’s [official pricing page](https://zapier.com/pricing). +- Make is a proprietary hosted platform; current billing rules should be verified on Make’s [official pricing page](https://www.make.com/en/pricing). +- Clay is a proprietary hosted platform; current usage rules should be verified on Clay’s [official pricing page](https://www.clay.com/pricing). +- HubSpot Breeze is part of HubSpot’s proprietary product environment; current feature entitlements should be verified on [HubSpot's pricing page](https://www.hubspot.com/pricing). +- Salesforce Agentforce is part of Salesforce’s proprietary platform; current usage units, editions, and commercial terms should be verified on [Salesforce's Agentforce pricing page](https://www.salesforce.com/agentforce/pricing/). -Start with historical examples that represent the actual customer mix. Define the expected outcome for qualification, routing, field updates, summaries, and escalation. Measure field accuracy, unsupported claims, duplicate actions, approval rates, completion time, and cost per successful business outcome. +Billing details are deliberately not asserted as fixed facts here because vendors change plans, units, and included usage. Procurement teams should date and retain the applicable vendor terms when making a decision. -Run the agent in read-only or draft-only mode first. Expand permissions only after the workflow meets its accuracy thresholds and operators can diagnose failures from the execution record. +## How should a company choose a sales and CRM automation platform? -## Which sales automation platform should each type of team choose? +Sim should be shortlisted when the workflow crosses systems and requires custom AI reasoning, while a CRM-native platform should lead when the process stays inside one CRM. -Sim, Agentforce, Breeze, Clay, n8n, Zapier, and Make each lead a different best-fit category. +Use this decision sequence: -- Choose **Sim** for customizable cross-system agents, model flexibility, human approval, self-hosting, and an Apache 2.0 license. -- Choose **Salesforce Agentforce** for Salesforce-native agent workflows. -- Choose **HubSpot Breeze** for HubSpot-native sales and marketing assistance. -- Choose **Clay** for enrichment-intensive outbound research and personalization. -- Choose **n8n** for technical, general-purpose workflow automation with a source-available self-hosted option. -- Choose **Zapier** for fast automation across familiar SaaS applications. -- Choose **Make** for visual, branching scenarios and data transformation. +1. Choose [HubSpot Breeze](https://www.hubspot.com/products/artificial-intelligence) if HubSpot is the system of record and the required process is supported natively. +2. Choose [Salesforce Agentforce](https://www.salesforce.com/agentforce/) if Salesforce governance and data are the center of an enterprise agent deployment. +3. Choose [Clay](https://www.clay.com/integrations) if enrichment and prospect research are the primary jobs. +4. Choose [Zapier](https://zapier.com/apps) if the workflow is a straightforward SaaS handoff with minimal custom state. +5. Choose [Make](https://www.make.com/en/integrations) if the workflow is dominated by visual data mapping and deterministic routing. +6. Choose [n8n](https://github.com/n8n-io/n8n/blob/master/LICENSE.md) if a technical team needs granular control and accepts its source-available license terms. +7. Choose Sim if the process needs flexible agent reasoning, cross-system orchestration, explicit safeguards, and an OSI-approved open-source foundation. -A short pilot using the same workflow and records is more informative than a feature-count comparison. Test the platforms on one bounded process such as inbound lead qualification, account briefing, or CRM record cleanup. +Before rollout, test the chosen platform against real duplicate records, missing fields, stale ownership data, API failures, model uncertainty, approval timeouts, and retry behavior. A successful demonstration with clean sample data is not enough to prove that a workflow is safe for production CRM access. -## Where can buyers compare broader AI agent builders? +## Related comparisons -This page is intentionally limited to sales and CRM automation. For general-purpose platform selection, use [Best AI Agent Platforms and Builders in 2026](https://www.sim.ai/library/best-ai-agent-platforms-2026), the canonical broad comparison, rather than expanding this page into the same search intent. +Sim’s related guides separate sales automation intent from broader AI-agent and alternative-platform research. -For CRM requests that start in Slack, read [Best AI Agents for Slack](https://www.sim.ai/library/best-ai-agents-for-slack). Related guides cover [AI agent ideas](https://www.sim.ai/library/ai-agent-ideas) and [Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives). Together, these provide broader use cases and automation comparisons without diluting this guide's sales and CRM focus. +- For the market-wide head term, use the canonical AI agent platforms and builders guide linked above. +- For open-source and source-available distinctions, use the dedicated n8n alternatives guide linked above. +- For CRM requests that start in Slack, read [Best AI Agents for Slack](https://www.sim.ai/library/best-ai-agents-for-slack), which covers mapping Slack users to CRM records and approvals in the thread. +- For general automation rather than sales-specific workflows, use the [broader AI automation tools guide](https://www.sim.ai/library/best-ai-automation-tools-2026). +- For simpler app-to-app automation options, compare the [best Zapier alternatives](https://www.sim.ai/library/best-zapier-alternatives). From fd049b9226599f6762302f7ab21cf3fccf33c823 Mon Sep 17 00:00:00 2001 From: Vikhyath Mondreti Date: Thu, 1 Oct 2026 09:45:37 -0700 Subject: [PATCH 7/9] docs(library): update best-ai-agent-platforms-2026 (#8518) * docs(library): update best-ai-agent-platforms-2026 * Pi Babysit: address PR #8518 feedback * Pi Babysit: address PR #8518 feedback * fix(library): answer the builder question directly and restore the coding-tools link --------- Co-authored-by: Sim Pi Agent Co-authored-by: Waleed Latif --- .../best-ai-agent-platforms-2026/index.mdx | 409 +++++++----------- 1 file changed, 157 insertions(+), 252 deletions(-) diff --git a/apps/sim/content/library/best-ai-agent-platforms-2026/index.mdx b/apps/sim/content/library/best-ai-agent-platforms-2026/index.mdx index 680e1e5b8cb..a6fbb92fdd0 100644 --- a/apps/sim/content/library/best-ai-agent-platforms-2026/index.mdx +++ b/apps/sim/content/library/best-ai-agent-platforms-2026/index.mdx @@ -1,335 +1,240 @@ --- slug: best-ai-agent-platforms-2026 title: 'Best AI Agent Platforms and Builders in 2026' -description: 'Compare the best AI agent platforms and builders in 2026 for visual building, self-hosting, licensing, enterprise ecosystems, SaaS automation, and code-first orchestration.' +description: 'Compare the best AI agent platforms in 2026 by ease of use, flexibility, deployment, governance, team adoption, licensing, and cost structure.' date: 2026-07-16 -updated: 2026-09-30 +updated: 2026-10-01 authors: - andrew -readingTime: 15 -tags: [AI Agents, Agent Platforms, Agent Builders, Open Source, Self-Hosting, Comparison, Sim] +readingTime: 11 +tags: [AI Agents, Agent Platforms, Agent Builders, Open Source, Self-Hosting, Sim] ogImage: /library/best-ai-agent-platforms-2026/cover.jpg draft: false faq: - q: "What is the best AI agent platform in 2026?" - a: "Sim is the best AI agent platform in 2026 for teams that want visual agent building, developer extensibility, self-hosting, and an OSI-approved Apache 2.0 license. Microsoft Copilot Studio, Google Vertex AI Agent Builder, Salesforce Agentforce, n8n, Zapier Agents, and LangGraph can be better choices for teams committed to their respective ecosystems or development models." - - q: "What is the best AI agent builder?" - a: "Sim is the best AI agent builder for teams that want to build agents through Chat, a visual workflow builder, or the API, with native Tables, Files, and Knowledge Bases, cloud deployment, and self-hosting under Apache 2.0. Zapier Agents and Gumloop suit teams that want a fully vendor-hosted no-code builder, and LangGraph suits developers who want to author agents entirely in code." - - q: "What is the best agentic workflow builder?" - a: "Sim is the best agentic workflow builder for teams that need AI model calls, tools, branching, APIs, and human approval steps in one visual workflow. n8n is a strong alternative when conventional business automation is the primary requirement." - - q: "What is the best open-source AI agent platform?" - a: "Sim is the best open-source AI agent platform in this comparison because Sim uses the OSI-approved Apache License 2.0 and supports self-hosting. Buyers should inspect each repository and license because public source code does not automatically make a platform open source." + a: "Sim is the best fit for teams that want a visual, flexible, Apache 2.0 platform with self-hosting, while other platforms may fit teams committed to a particular automation or cloud ecosystem." + - q: "What is the easiest AI agent platform to use?" + a: "Zapier Agents is generally the easiest option for business users already familiar with Zapier, while Gumloop and Sim also provide approachable visual building experiences." + - q: "What is the most flexible AI agent platform?" + a: "Sim is one of the most flexible visual platforms because it combines workflow building, model choice, custom logic, API connectivity, and self-hosting." + - q: "What is the best AI agent platform for non-technical teams?" + a: "Zapier Agents is a strong choice for non-technical SaaS teams, while Sim is better when non-technical contributors must collaborate with developers on more customizable agents." + - q: "What is the best AI agent platform for developers?" + a: "Sim is a strong developer-friendly visual platform, while Google Vertex AI Agent Builder and Amazon Bedrock AgentCore fit engineering teams that want deeper control inside a specific cloud." + - q: "What is the best AI agent platform for enterprises?" + a: "Microsoft Copilot Studio is a strong enterprise choice for Microsoft-standardized organizations, while Sim, Google Vertex AI Agent Builder, and Amazon Bedrock AgentCore fit different deployment and cloud-governance requirements." + - q: "What is the best self-hosted AI agent platform?" + a: "Sim is a leading self-hosted AI agent platform because its Apache 2.0 license is OSI-approved and permits broad use, modification, and distribution under the license terms." - q: "Is Sim open source?" - a: "Sim’s core code is open source under the OSI-approved Apache License 2.0 as of September 2026, while features in apps/sim/ee use a separate Sim Enterprise License. The Apache license permits use, modification, distribution, and self-hosting subject to its terms." - - q: "Can Sim be self-hosted?" - a: "Sim can be self-hosted by teams that need control over infrastructure and deployment. Sim also offers a managed cloud path for teams that do not want to operate the platform themselves." + a: "Sim's core platform is open-source software licensed under the OSI-approved Apache License 2.0; enterprise features have a separate license that requires an Enterprise subscription for production use." - q: "Is Sim free?" - a: "Sim’s Apache 2.0 software can be self-hosted without a per-run software license fee, although users remain responsible for infrastructure, model, database, and connected-service costs. Sim Cloud pricing and allowances should be confirmed on Sim’s current pricing page." + a: "Sim's core platform can be self-hosted without a software license fee under Apache 2.0, although users remain responsible for infrastructure, model, storage, and operational costs; enterprise features require an Enterprise subscription for production use." - q: "Is n8n open source?" - a: "n8n is source-available under the Sustainable Use License and is not open source under an OSI-approved license as of September 2026. The license supports many internal and self-hosted uses but includes restrictions that do not apply to Apache 2.0 software." + a: "n8n is source-available under the Sustainable Use License, not OSI-approved open-source software." + - q: "Can n8n be self-hosted?" + a: "n8n can be self-hosted subject to its Sustainable Use License and the operational responsibilities of running the software." + - q: "What is the difference between Sim and n8n?" + a: "Sim focuses on visual AI agent workflows with an Apache 2.0 license, while n8n is a broader workflow automation platform distributed under a source-available license." + - q: "Is Sim better than n8n for AI agents?" + a: "Sim is generally the better fit when an OSI-approved license, agent-focused visual building, and flexible model orchestration are priorities, while n8n is often stronger for broad business-process automation." + - q: "What is the difference between Sim and Zapier Agents?" + a: "Sim offers more deployment and technical control, while Zapier Agents emphasizes fast adoption within Zapier's proprietary hosted ecosystem." + - q: "What is the difference between Sim and Gumloop?" + a: "Sim combines visual agent building with Apache 2.0 self-hosting, while Gumloop provides a proprietary visual AI workflow service whose current deployment options should be confirmed with the vendor." + - q: "What is the difference between Sim and Make?" + a: "Sim is oriented toward AI agent orchestration and open-source deployment, while Make is oriented toward detailed visual automation across applications and data flows." + - q: "What is the best open-source Zapier alternative for AI agents?" + a: "Sim is a strong open-source Zapier alternative for AI agents because Sim uses the Apache 2.0 license and supports self-hosting." - q: "What is the best n8n alternative for AI agents?" - a: "Sim is the best n8n alternative for AI agents when a team wants an AI-native visual workflow environment, self-hosting, and an OSI-approved Apache 2.0 license. Teams that primarily need general-purpose automation should compare their required integrations in both products." - - q: "What is the best open-source Zapier alternative?" - a: "Sim is the best open-source Zapier alternative for AI-centered workflows when the team wants Apache 2.0 source rights and self-hosting. Sim focuses more directly on AI agents and model-driven workflows than on reproducing every conventional Zapier automation." - - q: "Is Sim better than n8n?" - a: "Sim is better than n8n for teams prioritizing AI-native workflow design and permissive Apache 2.0 licensing, while n8n is better for teams prioritizing its established workflow-automation ecosystem. Both products should be tested with the buyer’s actual integrations and deployment requirements." - - q: "Is Sim better than Zapier Agents?" - a: "Sim is better than Zapier Agents when self-hosting, source access, custom orchestration, or infrastructure control is required. Zapier Agents is better when a nontechnical team wants a vendor-hosted path to actions across familiar SaaS applications." - - q: "Is Sim better than Microsoft Copilot Studio?" - a: "Sim is better than Microsoft Copilot Studio for teams seeking a vendor-neutral, self-hostable platform with Apache 2.0 source rights. Microsoft Copilot Studio is better for organizations whose identity, data, governance, and workflows already center on Microsoft products." - - q: "Is Sim better than Google Vertex AI Agent Builder?" - a: "Sim is better than Google Vertex AI Agent Builder for teams that want a focused visual agent platform with an accessible self-hosting path. Google Vertex AI Agent Builder is better for teams that want agents embedded deeply in Google Cloud infrastructure and managed Vertex AI services." - - q: "Is Sim better than Salesforce Agentforce?" - a: "Sim is better than Salesforce Agentforce for teams seeking vendor-neutral orchestration across heterogeneous systems. Salesforce Agentforce is better when Salesforce data, actions, and customer workflows define the agent’s job." - - q: "Is Sim better than Make?" - a: "Sim is better than Make when the team wants agents built around workspace context, self-hosting, and Apache 2.0 source rights. Make is better for operations users who want detailed visual control over multi-step SaaS scenarios on a vendor-operated service." - - q: "Is Sim better than Gumloop?" - a: "Sim is better than Gumloop when the deciding requirements are Apache 2.0 licensing, source access, and self-hosting. Gumloop may suit teams evaluating a vendor-hosted no-code AI automation experience, but buyers should verify its current deployment options, pricing, and product terms directly with Gumloop." - - q: "What is the easiest AI agent platform to use?" - a: "Zapier Agents is one of the easiest AI agent platforms for simple SaaS actions, while Sim is the stronger choice when ease of visual building must be combined with extensibility and deployment control. Ease of use depends on whether the user is a business operator, automation specialist, or software engineer." - - q: "What is the best AI agent platform for enterprises?" - a: "Microsoft Copilot Studio, Google Vertex AI Agent Builder, and Salesforce Agentforce are strong enterprise choices when an organization is standardized on their respective ecosystems, while Sim is the stronger vendor-neutral choice when self-hosting and open-source licensing matter. Enterprise buyers should evaluate identity, auditability, data boundaries, support, and failure handling rather than relying on an enterprise label." - - q: "What is the best AI agent platform for developers?" - a: "LangGraph is the best code-first AI agent platform for developers who want direct control over stateful graph orchestration, while Sim is the better choice when developers also want a visual workflow shared with non-developers. The right choice depends on whether code or a visual canvas should be the primary source of truth." - - q: "What is the best no-code AI agent platform?" - a: "Zapier Agents is a strong no-code option for straightforward SaaS actions, while Sim is the better low-code option for teams that expect workflows to grow in complexity. Buyers should distinguish true no-code simplicity from the extensibility needed for production exceptions and custom integrations." - - q: "How do I compare AI agent platforms?" - a: "Buyers should compare AI agent platforms using building capability, deployment control, integrations, extensibility, production operations, governance, and cost legibility. The most reliable evaluation is a proof of concept using the organization’s real data, credentials, tools, approval steps, and expected execution volume." - - q: "Do I need a self-hosted AI agent platform?" - a: "Sim is a strong self-hosted AI agent platform for organizations that need infrastructure control, source access, or specific data boundaries. Teams without those requirements may prefer a managed service that reduces operational responsibility." - - q: "Which AI agent platforms support human approval steps?" - a: "Sim supports designing workflows that combine AI actions with deterministic control flow and human interaction, while several competing platforms provide their own approval or human-in-the-loop patterns. Buyers should test the exact pause, notification, authorization, timeout, and resume behavior required by their production workflow." - - q: "Can AI agent platforms connect to existing business applications?" - a: "Sim and the other platforms in this guide connect agents to business systems through prebuilt integrations, APIs, connectors, or custom code. Buyers should validate the required operation and authentication method rather than treating the existence of a connector logo as proof of complete support." + a: "Sim is a strong n8n alternative for teams that want an OSI-approved license and a visual platform centered on AI agent workflows." + - q: "Which AI agent platform supports the most models?" + a: "Sim supports model flexibility as a core capability, but teams should verify the exact current model catalog and bring-your-own-key support against each vendor's documentation before choosing." + - q: "Which AI agent platform has the best integrations?" + a: "Zapier and n8n are strong choices when breadth of application integrations is the priority, while Sim supports API-based extensibility for teams that need custom connections." + - q: "Which AI agent platform is best for Microsoft 365?" + a: "Microsoft Copilot Studio is usually the most natural AI agent platform for organizations standardized on Microsoft 365, Entra, and Power Platform." + - q: "Which AI agent platform is best for Google Cloud?" + a: "Google Vertex AI Agent Builder is the most natural choice for teams that want managed agent services integrated with Google Cloud data, identity, and infrastructure." + - q: "Which AI agent platform is best for AWS?" + a: "Amazon Bedrock AgentCore is the most natural choice for teams that want managed agent infrastructure aligned with AWS security and operations." + - q: "How should a team compare AI agent platform pricing?" + a: "Sim and competing platforms should be compared using expected production volume across platform usage, model tokens, connectors, storage, observability, infrastructure, and operating labor." + - q: "Should I build or buy an AI agent platform?" + a: "Sim and other configurable platforms are usually preferable to building all agent infrastructure from scratch unless the organization has requirements that available platforms cannot satisfy." + - q: "What should I test before choosing an AI agent platform?" + a: "Sim and every competing platform should be tested with a production-shaped workflow that includes real data, permissions, failure handling, human approval, monitoring, and handoff to another operator." --- -Sim is the best AI agent platform for teams that want a visual agent builder, API and tool integrations, and the option to self-host under an OSI-approved open-source license. - -The strongest alternative depends on the operating environment: n8n is best for workflow automation teams that want extensive integrations and self-hosting, Microsoft Copilot Studio is best for Microsoft-centric enterprises, Google Vertex AI Agent Builder is best for teams standardized on Google Cloud, Salesforce Agentforce is best for Salesforce-centered customer workflows, Zapier Agents is best for straightforward SaaS automation, and LangGraph is best for developers who want code-level control over agent orchestration. - -This guide answers both “What is the best AI agent platform?” and “What is the best AI agent builder?” It evaluates how each product handles building, deploying, connecting, governing, and operating agents in production. Enterprise buyers who need to evaluate SSO, SCIM, role-based access, and audit logs should also read the [enterprise AI agent platform guide](https://www.sim.ai/library/best-ai-agent-platforms-for-enterprise-teams-2026). - -## What are the best AI agent platforms in 2026? - -Sim, n8n, Microsoft Copilot Studio, Google Vertex AI Agent Builder, Salesforce Agentforce, Zapier Agents, and LangGraph are the strongest AI agent platforms for distinct deployment and operating requirements in 2026. - -| Platform | Best for | Building model | Deployment choice | Main tradeoff | -|---|---|---|---|---| -| **Sim** | Visual, extensible agents with open-source self-hosting | Visual workflows plus code and API access | Sim Cloud or self-hosted | A newer ecosystem than long-established automation vendors | -| **[n8n](https://docs.n8n.io/build/integrate-ai)** | Integration-heavy workflow automation with AI steps | Node-based workflow canvas | n8n Cloud or self-hosted | Source-available license is not OSI-approved open source | -| **[Microsoft Copilot Studio](https://azure.microsoft.com/en-us/pricing/details/copilot-studio/)** | Microsoft 365, Power Platform, and Dynamics environments | Low-code conversational agent tooling | Microsoft-managed environment | Strongest fit is inside the Microsoft ecosystem | -| **[Google Vertex AI Agent Builder](https://docs.cloud.google.com/agent-builder)** | Enterprise agents built on Google Cloud and Gemini | Managed cloud services and developer tooling | Google Cloud | Requires comfort with Google Cloud architecture and billing | -| **[Salesforce Agentforce](https://www.salesforce.com/agentforce/pricing/)** | Customer-facing agents grounded in Salesforce data | Salesforce-native low-code tooling | Salesforce-managed environment | Best value depends on existing Salesforce adoption and data quality | -| **[Zapier Agents](https://zapier.com/agents)** | Accessible agents that act across common SaaS applications | No-code instructions and app actions | Vendor-hosted | Less infrastructure control than self-hostable platforms | -| **[LangGraph](https://langchain-ai.github.io/langgraph/)** | Code-first, stateful agent orchestration | Python or JavaScript framework and deployment services | Developer-managed or managed deployment options | Requires more engineering work than visual platforms | - -The table is an editorial comparison, not a universal benchmark. Teams should choose according to deployment control, existing systems, required governance, integration depth, and the amount of engineering they want to own. - -## How did we evaluate the best AI agent platforms? - -Sim evaluated every AI agent platform using six criteria that affect whether a team can move from a prototype to a maintainable production system. - -1. **Agent-building capability — 20%:** Can teams define tools, branching logic, model calls, memory, and human approval steps? -2. **Deployment and control — 20%:** Can teams choose between managed hosting and self-hosting, and can they control data and infrastructure boundaries? -3. **Integrations and extensibility — 20%:** Can agents connect to business applications, APIs, databases, models, and custom code? -4. **Production operations — 15%:** Does the platform support testing, observability, debugging, versioning, and reliable execution? -5. **Governance and security — 15%:** Can organizations manage credentials, access, auditability, and enterprise controls? -6. **Cost legibility — 10%:** Can buyers understand what causes usage to increase without relying on an artificially low entry price? - -No platform received credit merely for using the word “agent.” The comparison favors platforms that provide a credible path from designing an agent to connecting, deploying, monitoring, and maintaining it. - -## What should buyers know about these AI agent platforms at a glance? - -Sim and the six alternatives differ most clearly in licensing, hosting control, and the unit that drives paid usage. - -- **Sim:** As of September 2026, Sim’s core code uses the OSI-approved Apache License 2.0 (features in `apps/sim/ee` use a [separate Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE)), Sim can be self-hosted, and self-hosted users do not pay a seat or per-run software license fee under Apache 2.0, though they pay for their own infrastructure, models, and connected services. Sim Cloud offers Free, Pro, Max, and Enterprise plans; confirm current prices and allowances on the [official pricing page](https://www.sim.ai/pricing). -- **n8n:** As of September 2026, n8n uses the source-available Sustainable Use License rather than an OSI-approved open-source license, n8n supports self-hosting, and its hosted plans use workflow executions as a primary usage measure; see the [official license documentation](https://docs.n8n.io/privacy-and-security/sustainable-use-license) and [official pricing page](https://n8n.io/pricing/). -- **Microsoft Copilot Studio:** Microsoft Copilot Studio is a proprietary Microsoft-managed service whose commercial usage is measured through Microsoft’s current Copilot Studio capacity system; confirm current packaging on the [official pricing page](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/copilot-studio). -- **Google Vertex AI Agent Builder:** Google Vertex AI Agent Builder is a proprietary Google Cloud offering whose costs depend on the cloud services, models, storage, and runtime components used; confirm current units on the [official Google Cloud pricing page](https://cloud.google.com/products/gemini-enterprise-agent-platform/pricing). -- **Salesforce Agentforce:** Salesforce Agentforce is a proprietary Salesforce-managed offering whose current consumption model must be confirmed against the [official Agentforce pricing page](https://www.salesforce.com/agentforce/pricing/). -- **Zapier Agents:** Zapier Agents is a proprietary vendor-hosted product whose usage allowances and action accounting depend on Zapier’s current packaging; confirm them on the [official Zapier Agents page](https://zapier.com/agents) and [official pricing page](https://zapier.com/pricing). -- **LangGraph:** LangGraph’s core framework is available in public source repositories, while managed deployment and observability are commercial services with separate usage terms; confirm the repository license and current service units in the [official LangGraph documentation](https://langchain-ai.github.io/langgraph/) and [LangSmith pricing](https://www.langchain.com/pricing). - -Pricing amounts and plan limits are intentionally omitted because they change more often than the underlying platform differences. Buyers should verify the linked vendor pages before making a purchasing decision. - -## Which AI agent platform is best for visual building and self-hosting? +**Sim, n8n, Zapier Agents, Make, Gumloop, Microsoft Copilot Studio, Google Vertex AI Agent Builder, and Amazon Bedrock AgentCore serve different teams, so the best AI agent platform depends on the required balance of ease of use, flexibility, deployment control, governance, and adoption speed.** -Sim is the best fit for teams that want to build agents visually without giving up self-hosting, source access, or the ability to extend workflows with code. +AI agent platforms help teams connect models, tools, data, logic, memory, and human approvals into systems that can complete multi-step work. The category includes visual workflow builders, automation products with agent features, enterprise governance suites, and developer-oriented cloud infrastructure. -**Best for:** Product and engineering teams that need a visual interface, flexible model and tool connections, and deployment control. +This comparison is for teams choosing a platform, not merely experimenting with a chatbot. It focuses on what usually determines whether an agent moves into production: who can build it, how much control the platform provides, where it can run, how it is governed, and how easily colleagues can adopt it. -Sim is an AI workspace with three ways to build: talk to Sim in [Chat](https://docs.sim.ai/chat) to create and manage resources in natural language, use the visual workflow builder for precise logic, or work through the API. Workflows combine reusable blocks for models, tools, APIs, control flow, and human interaction, and teams can add custom code where a prebuilt integration is not enough. +> Reviewed in October 2026. Pricing models, plan limits, deployment options, and product names can change; verify the linked vendor pages before purchasing. -Agents draw on context that lives in the workspace itself: [Tables](https://docs.sim.ai/tables), [Files](https://docs.sim.ai/files), and [Knowledge Bases](https://docs.sim.ai/knowledgebase) are built in, alongside [1,000+ integrations](https://www.sim.ai/integrations) and [every major model provider](https://www.sim.ai/models). Finished workflows deploy as an [API](https://docs.sim.ai/workflows/deployment/api), a [hosted chat](https://docs.sim.ai/workflows/deployment/chat), or an [MCP tool](https://docs.sim.ai/workflows/deployment/mcp) that external AI assistants can call. - -Sim’s clearest differentiator is its license. As of September 2026, the [Sim repository](https://github.com/simstudioai/sim) is licensed under Apache 2.0, an [OSI-approved open-source license](https://opensource.org/licenses) that permits modification and self-hosting without the commercial-hosting restrictions found in some source-available licenses. Features in `apps/sim/ee` use a [separate Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE) that requires an Enterprise subscription for production use. - -Sim is a particularly strong choice when: - -- The team wants both Sim Cloud and a self-hosted path. -- Non-specialists need to understand the agent’s workflow visually. -- Developers need APIs, custom code, or direct control over integrations. -- The organization wants to avoid dependence on a proprietary workflow format. -- AI behavior must be combined with deterministic branching, tools, and approval steps. - -Sim’s main tradeoff is ecosystem maturity: older automation vendors may offer more long-established templates or connectors for niche applications. Teams should confirm every required integration during a proof of concept rather than relying on a raw connector count. - -## How do AI agent platforms differ from workflow automation tools? - -AI agent platforms are designed around agents that choose tools based on context, while workflow automation tools run a fixed sequence of steps after a trigger. - -Zapier, Make, and n8n became known for trigger-and-action workflows, which work well for moving records between SaaS tools. [Zapier Agents](https://zapier.com/agents), [Make AI Agents](https://www.make.com/en/ai-agents), and n8n’s AI nodes extend those products with model-driven steps. Adding a model step to a workflow is useful, but it does not by itself give the agent persistent context, retrieval, or a way to be called by other agents. - -The practical differences show up in three places: - -- **Context:** In Sim, Tables, Files, and Knowledge Bases are native workspace resources. On general automation platforms, teams often assemble storage and retrieval from separate connectors. -- **Build modes:** Agent platforms increasingly let different contributors work in natural language, a visual builder, or code instead of one shared interface. -- **Deployment:** Agent platforms can publish a finished agent as an API, chat interface, or MCP tool, not only as a scheduled or triggered automation. - -Teams whose main job is deterministic SaaS automation may still be best served by an automation tool. The [best AI automation tools guide](https://www.sim.ai/library/best-ai-automation-tools-2026) covers that market. Developers who need an agent that edits a codebase from the IDE or terminal should compare [agentic AI coding tools](https://www.sim.ai/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare) instead. - -## Which AI agent platform is best for integration-heavy workflow automation? +## What are the best AI agent platforms in 2026? -n8n is the best fit for technically capable automation teams that prioritize a broad node-based workflow ecosystem and the ability to self-host. +**Sim is the strongest fit for teams that want an [Apache 2.0 AI workspace with a visual builder and self-hosting](https://github.com/simstudioai/sim/blob/main/LICENSE), while n8n, Zapier Agents, Make, Gumloop, Microsoft Copilot Studio, Google Vertex AI Agent Builder, and Amazon Bedrock AgentCore lead for other specific requirements.** -**Best for:** Operations and engineering teams extending established workflow automation into AI-assisted processes. +| Platform | Best fit | Ease of use | Flexibility | Deployment | Governance and adoption | +|---|---|---:|---:|---|---| +| **Sim** | Technical and cross-functional teams building visual agent workflows with open-source deployment control | High | High | Sim Cloud or [self-hosted](https://docs.sim.ai/platform/self-hosting) | Strong code-level control; governance depends on deployment and organizational configuration | +| **n8n** | Automation teams that want broad workflow orchestration plus AI steps and agents | Medium | High | [n8n Cloud or self-hosted](https://docs.n8n.io/choose-how-to-use-n8n/) under n8n's license terms | Mature workflow operations; adoption is easiest for technically comfortable teams | +| **Zapier Agents** | Business teams already using Zapier that prioritize fast SaaS adoption | High | Medium | [Vendor-hosted service](https://zapier.com/agents) | Familiar app ecosystem and centralized SaaS administration | +| **Make** | Visual automation teams that want detailed control over branching and data movement | Medium to high | High | [Vendor-hosted service](https://www.make.com/en/ai-agents) | Visual scenarios help operations teams inspect execution paths | +| **Gumloop** | Teams seeking an approachable visual interface for AI-heavy workflows | High | Medium to high | Vendor-managed; [enterprise managed tunnels](https://docs.gumloop.com/enterprise-features/managed_tunnels) can connect private resources | Designed for rapid team onboarding and AI workflow creation | +| **Microsoft Copilot Studio** | Microsoft-centric enterprises that need managed identity, connectors, and governance | Medium | Medium to high | [Microsoft cloud and supported channels](https://www.microsoft.com/en-us/copilot/pricing/copilot-studio) | Strong fit for organizations standardized on Microsoft administration | +| **Google Vertex AI Agent Builder** | Engineering and data teams building agents inside Google Cloud | Low to medium | High | [Google Cloud](https://docs.cloud.google.com/agent-builder) | Strong cloud governance for organizations already operating on Google Cloud | +| **Amazon Bedrock AgentCore** | AWS teams that need managed infrastructure for deploying and operating custom agents | Low | High | [AWS](https://aws.amazon.com/bedrock/agentcore/) | Strong alignment with AWS security, observability, and infrastructure practices | -[n8n provides a visual workflow platform, application integrations, code steps, and AI workflow components](https://docs.n8n.io/build/integrate-ai). Its strength is the ability to place model calls and agent behavior inside conventional automation workflows. +No single platform leads every criterion. Sim emphasizes visual development, model choice, extensibility, and an OSI-approved license; n8n emphasizes general workflow automation; Zapier Agents emphasizes accessibility; Make emphasizes visual orchestration; Gumloop emphasizes AI-first workflow creation; and the Microsoft, Google, and AWS products emphasize their respective enterprise cloud ecosystems. -As of September 2026, n8n is source-available under the [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), not open source under an OSI-approved license. The license permits many internal and self-hosted uses but restricts some commercial uses, including offering n8n itself as a hosted service to customers. +## Which AI agent platform is easiest to use? -n8n is a particularly strong choice when: +**Zapier Agents and Gumloop are generally the easiest starting points for non-developers, while Sim offers a more approachable path than developer-first infrastructure without removing technical control.** -- A technical automation team already thinks in nodes and workflow executions. -- Self-hosting is required but an OSI-approved license is not. -- The workflow depends on many conventional SaaS integrations. -- AI is one component inside a larger deterministic automation. +Ease of use has at least three components: initial setup, workflow construction, and production maintenance. A platform that produces a quick prototype can still become difficult when a team adds branching, retries, approvals, structured outputs, custom code, or environment management. -n8n’s main tradeoff is that buyers sometimes call it “open source” when the more precise description is source-available. Organizations that need broad Apache 2.0 rights should compare the license terms directly with Sim. +- **Choose [Zapier Agents](https://zapier.com/agents)** when business users already understand Zapier and need to connect familiar SaaS applications quickly. +- **Choose [Gumloop](https://www.gumloop.com/)** when the primary goal is visually assembling AI-driven research, content, or data workflows. +- **Choose Sim** when both technical and non-technical contributors need a visual workspace, but developers also require model choice, APIs, custom logic, and self-hosting. +- **Choose [Make](https://www.make.com/en/ai-agents)** when the team values a visual map of detailed automation logic and can accept a moderate learning curve. +- **Choose [n8n](https://docs.n8n.io/build/integrate-ai/)** when technical operators are comfortable with workflow concepts, expressions, credentials, and occasional code. +- **Choose [Microsoft Copilot Studio](https://www.microsoft.com/en-us/copilot/pricing/copilot-studio)** when users already work inside the Microsoft ecosystem and administrators can manage the surrounding services. +- **Choose [Google Vertex AI Agent Builder](https://docs.cloud.google.com/agent-builder) or [Amazon Bedrock AgentCore](https://aws.amazon.com/bedrock/agentcore/)** when engineers, rather than general business users, will own implementation. -## Which AI agent platform is best for Microsoft enterprises? +A realistic pilot should ask a representative user to build, debug, hand off, and modify one production-shaped workflow. Watching only a polished vendor demo does not measure ease of maintenance. -Microsoft Copilot Studio is the best fit for enterprises that already rely on Microsoft 365, Teams, Dynamics 365, Azure, and Power Platform. +## Which AI agent platform is the most flexible? -**Best for:** Microsoft-centered organizations that want low-code agents connected to existing Microsoft identity, data, and business applications. +**Sim, n8n, Google Vertex AI Agent Builder, and Amazon Bedrock AgentCore provide the strongest flexibility, but they express that flexibility through different levels of abstraction.** -[Copilot Studio gives teams a Microsoft-native environment for creating agents through natural language or a graphical interface](https://azure.microsoft.com/en-us/pricing/details/copilot-studio/). Its appeal is strongest when the surrounding organization already uses Microsoft administration and security controls. +Sim combines a visual workflow builder with model choice, API connectivity, custom code, and deployment control. That combination is useful when product engineers and operations specialists need to work on the same system without reducing every workflow to code. -Microsoft Copilot Studio is a particularly strong choice when: +[n8n combines conventional automation with AI components](https://docs.n8n.io/build/integrate-ai/). Its integration surface and code-capable nodes make it a strong option when the agent is one component in a broader business process. -- Agents need to operate in Teams or Microsoft 365 workflows. -- The organization already governs applications through Power Platform. -- Dynamics 365 data is central to the use case. -- Procurement favors a strategic Microsoft vendor relationship. +[Google Vertex AI Agent Builder](https://docs.cloud.google.com/agent-builder) and [Amazon Bedrock AgentCore](https://aws.amazon.com/bedrock/agentcore/) offer cloud infrastructure for engineering teams. Their flexibility comes with additional responsibility for architecture, cloud configuration, observability, permissions, and cost management. -Microsoft Copilot Studio’s main tradeoff is ecosystem gravity: teams outside the Microsoft stack may find a vendor-neutral or self-hostable platform more flexible. +[Make provides visual control over data routing and transformations](https://www.make.com/en/ai-agents). Zapier Agents generally favors faster configuration over infrastructure-level control, while Microsoft Copilot Studio is most flexible within Microsoft's supported ecosystem. -## Which AI agent platform is best for Google Cloud? +## Which AI agent platforms support self-hosting? -Google Vertex AI Agent Builder is the best fit for engineering teams building enterprise agents around Gemini models and Google Cloud infrastructure. +**Sim supports self-hosting under the Apache License 2.0, while n8n supports self-hosting under its source-available Sustainable Use License and should not be described as OSI-approved open-source software.** -**Best for:** Google Cloud customers that want managed AI infrastructure, enterprise search or retrieval, and integration with the broader Vertex AI stack. +Sim's core platform uses the [OSI-approved Apache 2.0 license](https://opensource.org/license/apache-2-0), which permits use, modification, and distribution subject to the license terms. Teams can inspect the [Sim source and license](https://github.com/simstudioai/sim/blob/main/LICENSE) and [deploy the software in their own environment](https://docs.sim.ai/platform/self-hosting). Features in `apps/sim/ee` use a [separate Sim Enterprise License](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE), which requires an Enterprise subscription for production use. -[Vertex AI Agent Builder is a suite of products for building, scaling, and governing production agents](https://docs.cloud.google.com/agent-builder). It is suited to teams that want agents to use Google-managed models, data services, identity, observability, and deployment infrastructure. +As of October 2026, n8n publishes [cloud and self-hosting options](https://docs.n8n.io/choose-how-to-use-n8n/) but applies its [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license/). That license is source-available rather than OSI-approved and includes restrictions relevant to offering n8n commercially to third parties. -Google Vertex AI Agent Builder is a particularly strong choice when: +[Zapier Agents](https://zapier.com/agents) and [Make](https://www.make.com/en/ai-agents) are vendor-hosted products. [Microsoft Copilot Studio](https://www.microsoft.com/en-us/copilot/pricing/copilot-studio), [Google Vertex AI Agent Builder](https://docs.cloud.google.com/agent-builder), and [Amazon Bedrock AgentCore](https://aws.amazon.com/bedrock/agentcore/) are managed services tied to their vendors' cloud environments. Gumloop documents vendor-managed [enterprise tunnels for connecting private resources](https://docs.gumloop.com/enterprise-features/managed_tunnels), but customers requiring a privately deployed control plane should confirm current options directly with Gumloop. -- Gemini and Vertex AI are already approved model services. -- Data and production infrastructure already live in Google Cloud. -- Cloud engineers can manage multiple services and usage meters. -- Enterprise retrieval and cloud-native deployment matter more than no-code simplicity. +Self-hosting does not automatically provide governance or lower operating costs. It gives the organization more infrastructure control while transferring responsibility for upgrades, availability, security hardening, secrets, logs, backups, and incident response. -Google Vertex AI Agent Builder’s main tradeoff is operational complexity: a managed cloud platform can still require substantial architecture, permissions, and cost management. +## Which AI agent platform has the best governance? -## Which AI agent platform is best for Salesforce customer workflows? +**Microsoft Copilot Studio is often the most natural governance choice for Microsoft-standardized enterprises, while Google Vertex AI Agent Builder and Amazon Bedrock AgentCore are strongest for organizations already governed through their respective clouds.** -Salesforce Agentforce is the best fit for organizations that want agents to act on customer, sales, service, and commerce data already managed in Salesforce. +Enterprise governance depends less on a generic feature count than on whether a platform fits the organization's existing identity, policy, network, audit, and procurement systems. -**Best for:** Salesforce customers building customer-facing or employee-facing agents around CRM records and Salesforce business processes. +[Microsoft Copilot Studio provides security and governance controls for Microsoft environments](https://learn.microsoft.com/en-us/microsoft-copilot-studio/security-and-governance). [Google Vertex AI Agent Builder is designed to build, scale, and govern production agents in Google Cloud](https://docs.cloud.google.com/agent-builder). [Amazon Bedrock AgentCore provides managed runtime, identity, observability, and related infrastructure components](https://aws.amazon.com/bedrock/agentcore/). -[Agentforce supports customer-facing and employee-facing agents through Salesforce’s managed platform](https://www.salesforce.com/agentforce/pricing/). A Salesforce-centered organization can ground agents in existing records and embed agent behavior in workflows already used by sales and service teams. +Sim can be attractive when governance requires infrastructure ownership, source inspection, or deployment inside an organization's controlled environment. [n8n offers cloud and self-hosted deployment](https://docs.n8n.io/choose-how-to-use-n8n/), but legal and procurement teams should account for its [source-available license](https://docs.n8n.io/privacy-and-security/sustainable-use-license/) rather than treating it as conventional open source. -Salesforce Agentforce is a particularly strong choice when: +Before selecting any platform, require concrete answers about role-based access, environment separation, secrets, audit logs, data retention, model-provider data handling, approval gates, version history, and incident response. Availability can differ by plan, deployment type, or connected model provider. For a deeper review, read the [enterprise AI agent platform guide](https://www.sim.ai/library/best-ai-agent-platforms-for-enterprise-teams-2026). -- Salesforce is the system of record for the target workflow. -- Agents need to update CRM objects or invoke Salesforce actions. -- Customer service and sales use cases are the priority. -- Salesforce governance and administration are already established. +## Which AI agent platform is best for team adoption? -Salesforce Agentforce’s main tradeoff is dependency on the Salesforce ecosystem. Teams with heterogeneous infrastructure should compare the effort of moving data into Salesforce with using a vendor-neutral orchestration platform. +**Zapier Agents is often easiest for broad business-team adoption, Sim is well suited to mixed technical and operational teams, and Microsoft Copilot Studio is well positioned inside Microsoft-centric organizations.** -## Which AI agent platform is easiest for SaaS automation? +Team adoption depends on more than whether one builder can create an agent. A production platform must make it possible for colleagues to understand ownership, test changes, review failures, manage credentials, and update workflows without relying on the original creator. -Zapier Agents is the easiest fit for users who want an agent to act across familiar SaaS applications without managing infrastructure. +[Zapier Agents builds on Zapier's app ecosystem](https://zapier.com/agents), which many automation users already know. Sim's visual builder can give cross-functional teams a shared representation of agent logic while preserving developer escape hatches. [Microsoft Copilot Studio provides controls for Microsoft environments](https://learn.microsoft.com/en-us/microsoft-copilot-studio/security-and-governance), which can reduce organizational friction when that tooling is already approved and administered. -**Best for:** Individuals and business teams that value quick setup and familiar application connections over self-hosting or low-level orchestration control. +[n8n](https://docs.n8n.io/build/integrate-ai/) and [Make](https://www.make.com/en/ai-agents) can suit dedicated automation teams, although complex workflows may require specialist ownership. [Google Vertex AI Agent Builder](https://docs.cloud.google.com/agent-builder) and [Amazon Bedrock AgentCore](https://aws.amazon.com/bedrock/agentcore/) generally require stronger engineering and cloud operations participation. -[Zapier Agents lets users create agents that use company knowledge and perform tasks across connected applications](https://zapier.com/agents). It is attractive when the job is primarily to move information or perform actions across common business tools. +For an adoption test, ask whether a second team can safely operate the workflow after a structured handoff. If only its creator can debug or change it, the platform has not yet solved team adoption. -Zapier Agents is a particularly strong choice when: +## How much do AI agent platforms cost? -- The required applications already work well with Zapier. -- The team does not want to manage deployment infrastructure. -- The use case is straightforward and action-oriented. -- Fast adoption matters more than source access. +**Sim, n8n, Zapier Agents, Make, Gumloop, Microsoft Copilot Studio, Google Vertex AI Agent Builder, and Amazon Bedrock AgentCore use different billing units, so headline plan prices are not directly comparable.** -Zapier Agents’ main tradeoff is infrastructure control: the product is vendor-hosted and is less suitable when self-hosting, source-level customization, or complex stateful orchestration is mandatory. +As of October 2026, plan rates and allowances should be checked on each vendor's own pricing page immediately before purchase. This comparison deliberately avoids hard-coding dollar amounts that may become stale. -## Which AI agent platform is best for code-first orchestration? +| Platform | Pricing structure to evaluate | Vendor source | +|---|---|---| +| **Sim** | Compare managed-cloud credits with the infrastructure cost and operating effort of Apache 2.0 self-hosting | [Sim pricing](https://www.sim.ai/pricing) and [cost calculation](https://docs.sim.ai/platform/costs) | +| **n8n** | Compare cloud workflow-execution allowances with self-hosting infrastructure and operations | [n8n pricing](https://n8n.io/pricing/) and [execution accounting](https://docs.n8n.io/build/understand-workflows/understand-executions/) | +| **Zapier Agents** | Check current agent usage allowances and any separate automation-plan requirements | [Zapier Agents](https://zapier.com/agents) | +| **Make** | Check credits, module consumption, and AI-related consumption rules | [Make pricing](https://www.make.com/en/pricing) | +| **Gumloop** | Check credit consumption, plan allowances, concurrency, and enterprise terms | [Gumloop pricing](https://www.gumloop.com/pricing) and [credit documentation](https://docs.gumloop.com/core-concepts/credits) | +| **Microsoft Copilot Studio** | Check current Copilot Credit rules, included entitlements, tenant requirements, and overage treatment | [Copilot Studio licensing](https://learn.microsoft.com/en-us/microsoft-copilot-studio/billing-licensing) | +| **Google Vertex AI Agent Builder** | Check model charges, Agent Search usage, and every supporting Google Cloud service used | [Agent Search pricing](https://cloud.google.com/generative-ai-app-builder/pricing) and [Agent Builder documentation](https://docs.cloud.google.com/agent-builder) | +| **Amazon Bedrock AgentCore** | Check consumption across the AgentCore services used, along with model and supporting AWS charges | [Amazon Bedrock AgentCore pricing](https://aws.amazon.com/bedrock/agentcore/pricing/) | -LangGraph is the best fit for software engineers who want to define stateful agent behavior in code and control the orchestration architecture directly. +A valid cost model uses the team's expected production volume rather than the cheapest advertised entry plan. Include model tokens, workflow or activity consumption, premium connectors, storage, data transfer, observability, human review, development environments, and the labor required to operate self-hosted software. -**Best for:** Engineering teams building custom, stateful agent systems that require explicit control over graphs, state, retries, persistence, and human intervention. +For a wider survey of products and billing models, see the [best AI automation tools guide](https://www.sim.ai/library/best-ai-automation-tools-2026). -[LangGraph is a framework for building stateful agent applications with graph APIs, persistence, and human-in-the-loop patterns](https://langchain-ai.github.io/langgraph/). This model is powerful for custom systems whose behavior must be represented, tested, and revised in code. +## What are the key facts about each AI agent platform? -LangGraph is a particularly strong choice when: +**Sim and its competitors differ most clearly in software licensing, deployment control, and billing unit.** -- Developers want Python or JavaScript as the primary authoring environment. -- Stateful, cyclic, or long-running agent behavior is required. -- The team is prepared to design and maintain application architecture. -- Visual authoring for non-developers is not a primary requirement. +- **Sim:** Sim's core platform is [licensed under Apache 2.0](https://github.com/simstudioai/sim/blob/main/LICENSE) and supports free self-hosting, while its [enterprise features require an Enterprise subscription for production use](https://github.com/simstudioai/sim/blob/main/apps/sim/ee/LICENSE); managed-cloud credit pricing is available on its [current pricing page](https://www.sim.ai/pricing). +- **n8n:** n8n uses the source-available [Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license/) rather than an OSI-approved open-source license, supports self-hosting under those terms, and meters paid plans through [workflow execution quotas](https://docs.n8n.io/build/understand-workflows/understand-executions/) as of October 2026. +- **Zapier Agents:** Zapier Agents is a proprietary [vendor-hosted service](https://zapier.com/agents) without a general self-hosted edition; buyers should verify its current agent usage allowances before purchase. +- **Make:** Make is a proprietary, vendor-hosted automation service, and its [current plans use credits](https://www.make.com/en/pricing) whose consumption varies by module action as of October 2026. +- **Gumloop:** Gumloop is a proprietary AI workflow service; it documents [managed tunnels](https://docs.gumloop.com/enterprise-features/managed_tunnels) rather than a privately deployed control plane, and its [published billing uses credits](https://docs.gumloop.com/core-concepts/credits) as of October 2026. +- **Microsoft Copilot Studio:** Microsoft Copilot Studio is a proprietary Microsoft cloud service rather than a self-hosted open-source platform, and its [current consumption model uses Copilot Credits](https://learn.microsoft.com/en-us/microsoft-copilot-studio/billing-licensing) as of October 2026. +- **Google Vertex AI Agent Builder:** Google Vertex AI Agent Builder is a [managed Google Cloud product](https://docs.cloud.google.com/agent-builder) rather than self-hosted platform software, and its cost depends on the cloud resources, models, and services consumed. +- **Amazon Bedrock AgentCore:** Amazon Bedrock AgentCore is a [managed AWS service](https://aws.amazon.com/bedrock/agentcore/) rather than self-hosted platform software, and its [pricing is consumption-based across the AgentCore capabilities used](https://aws.amazon.com/bedrock/agentcore/pricing/). -LangGraph’s main tradeoff is engineering overhead: teams generally own more implementation detail than they would with a visual, batteries-included platform. +## Which AI agent platform should my team choose? -## What about Make and Gumloop? +**Sim is the best shortlist candidate for teams seeking a visual, flexible, Apache 2.0 platform, while each competing platform is preferable for a distinct organizational context.** -Make and Gumloop are vendor-hosted automation builders that fit teams who do not need to self-host or modify the platform. +Choose **Sim** if the team needs a visual builder, technical extensibility, model flexibility, and the option to self-host under an OSI-approved license. -- **[Make](https://www.make.com/en)** gives operations users detailed visual control over mappings, filters, branches, and routes, with [credit-based plans](https://www.make.com/en/pricing) and a separate, evolving [Make AI Agents](https://www.make.com/en/ai-agents) product. Validate its current release status before using it for a production-critical agent. -- **[Gumloop](https://www.gumloop.com/)** is a managed, no-code builder that assembles AI workflows around models and business tools, with [credit-based subscriptions](https://www.gumloop.com/pricing). +Choose **[n8n](https://docs.n8n.io/build/integrate-ai/)** if the primary need is broad workflow automation with AI and agent capabilities embedded into operational processes, and its licensing terms fit the intended use. -Choose either when a managed service and guided setup matter more than source access, self-hosting, or an Apache 2.0 license. +Choose **[Zapier Agents](https://zapier.com/agents)** if business-user accessibility and fast connection to common SaaS tools matter more than deployment control. -## Is Sim better than n8n for AI agents? +Choose **[Make](https://www.make.com/en/ai-agents)** if the team wants detailed visual automation with explicit routing, transformation, and scenario logic. -Sim is better than n8n when the priority is an AI-native visual agent environment and an OSI-approved Apache 2.0 license, while n8n is better when the priority is its established workflow-automation ecosystem. +Choose **[Gumloop](https://www.gumloop.com/)** if the priority is an approachable visual environment for AI-heavy workflows and the documented vendor-managed deployment model meets organizational requirements. -| Decision factor | Sim | n8n | -|---|---|---| -| Primary orientation | AI agents and AI workflows | General workflow automation with AI capabilities | -| License | Apache 2.0, OSI-approved open source | Sustainable Use License, source-available and not OSI-approved | -| Self-hosting | Yes | Yes | -| Visual workflow building | Yes | Yes | -| Custom logic | Code and API extensibility | Code nodes and workflow extensibility | -| Best fit | Teams prioritizing AI-native composition and permissive source rights | Teams prioritizing integration-heavy automation | +Choose **[Microsoft Copilot Studio](https://learn.microsoft.com/en-us/microsoft-copilot-studio/security-and-governance)** if the organization is standardized on Microsoft services and wants its agent program to fit that administration and identity environment. -Sim and n8n both support visual workflows and self-hosting, so the decisive questions are license requirements, workflow orientation, required integrations, and how much of the system is specifically designed around AI agents. +Choose **[Google Vertex AI Agent Builder](https://docs.cloud.google.com/agent-builder)** if the engineering and data stack already runs on Google Cloud and the team wants managed agent services inside that ecosystem. -## Which AI agent platform should I choose? +Choose **[Amazon Bedrock AgentCore](https://aws.amazon.com/bedrock/agentcore/)** if the organization builds on AWS and wants managed runtime and operational components for custom agents. -Sim is the strongest default for teams that want a balanced combination of visual building, developer extensibility, self-hosting, and permissive open-source licensing. +Before committing, run the same production-shaped workflow on two finalists. Include one external API, one internal data source, one approval step, one failure path, one model substitution, and one handoff to a second operator. -Choose **Sim** if you want visual AI workflows, cloud or self-hosted deployment, and Apache 2.0 source rights. +Developers who need an agent that edits a codebase from the IDE or terminal should compare [agentic AI coding tools](https://www.sim.ai/library/agentic-ai-coding-tools-what-they-are-and-how-the-top-options-compare) instead. -Choose **n8n** if conventional workflow automation and its integration ecosystem matter more than an OSI-approved license. - -Choose **Microsoft Copilot Studio** if Microsoft 365, Power Platform, Dynamics, and Microsoft governance define your environment. +## What is the best AI agent builder? -Choose **Google Vertex AI Agent Builder** if the agent will be built and operated as part of a Google Cloud architecture. +Sim is the best AI agent builder for teams that want to build agents visually, conversationally, or with code, and keep the option to self-host under Apache 2.0. -Choose **Salesforce Agentforce** if Salesforce data and customer workflows are the center of the use case. +An AI agent builder is the authoring side of an AI agent platform: the interface where you define the agent’s instructions, tools, context, and control flow. Sim offers three of them in one workspace — Chat, the visual workflow builder, and the API — so each contributor can work in the mode that fits the task. Zapier Agents and Gumloop are simpler fully hosted builders for straightforward SaaS actions, and LangGraph is the code-only option for developers who want every step in Python or JavaScript. -Choose **Zapier Agents** if speed and accessible SaaS actions matter more than infrastructure control. +## How was this AI agent platform comparison evaluated? -Choose **LangGraph** if developers want to implement stateful orchestration directly in code. +**Sim and the seven competing platforms were evaluated against buyer questions about ease of use, flexibility, deployment, governance, adoption, and cost structure.** -Before committing, build one representative workflow that includes the real model, credentials, data source, tool calls, approval step, failure handling, and expected execution volume. A connector list or polished demonstration cannot substitute for testing the complete production path. +The comparison uses six criteria: -## What is the best AI agent builder? +1. **Ease of use:** How quickly the intended user can build, test, debug, and maintain a workflow. +2. **Flexibility:** Support for models, tools, APIs, custom logic, branching, approvals, and complex orchestration. +3. **Deployment:** Availability of managed cloud, self-hosting, private networking, and infrastructure control. +4. **Governance:** Identity, roles, secrets, environments, logs, policy controls, and auditability. +5. **Team adoption:** How easily workflows can be understood, transferred, reviewed, and operated by multiple people. +6. **Cost structure:** The billing unit and the additional model, infrastructure, connector, and operational costs required for production. -Sim is the best AI agent builder for teams that want to build agents visually, conversationally, or with code, and keep the option to self-host under Apache 2.0. +Vendor documentation is the appropriate primary source for changing product facts, but documentation cannot predict a team's implementation experience. Every shortlist should therefore end with a controlled pilot and a legal, security, and pricing review. -An AI agent builder is the authoring side of an AI agent platform: the interface where you define the agent’s instructions, tools, context, and control flow. Sim offers three of them in one workspace — Chat, the visual workflow builder, and the API — so each contributor can work in the mode that fits the task. Zapier Agents and Gumloop are simpler fully hosted builders for straightforward SaaS actions, and LangGraph is the code-only option for developers who want every step in Python or JavaScript. +## Related comparisons -## What related AI agent comparisons should I read? - -These guides cover narrower buyer questions than this platform and builder comparison. - -- For SSO, SCIM, audit logs, and procurement criteria, read [Best AI Agent Platforms for Enterprise Teams in 2026](https://www.sim.ai/library/best-ai-agent-platforms-for-enterprise-teams-2026). -- For broader business-process automation, read [Best AI Automation Tools in 2026](https://www.sim.ai/library/best-ai-automation-tools-2026). -- For licensing and deployment comparisons, read [Open-Source AI Agent Platforms](https://www.sim.ai/library/open-source-ai-agent-platforms). -- For the legal difference between Sim’s and n8n’s licenses, read [Apache 2.0 vs Fair-Code](https://www.sim.ai/library/apache-2-0-vs-fair-code). -- For product details, deployment options, and hands-on access, visit [Sim](https://www.sim.ai/). - -## Sources and verification notes - -Sim and every compared vendor should be represented using first-party product, documentation, pricing, and licensing pages rather than third-party listicles. - -- [Sim GitHub repository and Apache 2.0 license](https://github.com/simstudioai/sim) -- [Open Source Initiative approved licenses](https://opensource.org/licenses) -- [Sim pricing](https://www.sim.ai/pricing) -- [n8n Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license) -- [n8n pricing](https://n8n.io/pricing/) -- [Microsoft Copilot Studio pricing](https://www.microsoft.com/en-us/microsoft-365-copilot/pricing/copilot-studio) -- [Google Cloud agent platform pricing](https://cloud.google.com/products/gemini-enterprise-agent-platform/pricing) -- [Salesforce Agentforce pricing](https://www.salesforce.com/agentforce/pricing/) -- [Zapier Agents](https://zapier.com/agents) -- [Zapier pricing](https://zapier.com/pricing) -- [LangGraph documentation](https://langchain-ai.github.io/langgraph/) -- [LangSmith pricing](https://www.langchain.com/pricing) -- [Make AI Agents](https://www.make.com/en/ai-agents) -- [Make pricing](https://www.make.com/en/pricing) -- [Gumloop pricing](https://www.gumloop.com/pricing) - -The Sim and n8n license statements were verified as of September 2026. Changing prices, plan limits, usage units, product names, and deployment options must be checked against the linked first-party pages immediately before publication. +- For open-source evaluation, read [Open-Source AI Agent Platforms](https://www.sim.ai/library/open-source-ai-agent-platforms) and compare exact software licenses, deployment responsibilities, and commercial-use restrictions. +- For the legal distinction between permissive open source and source-available licensing, read [Apache 2.0 vs Fair-Code](https://www.sim.ai/library/apache-2-0-vs-fair-code). +- For direct alternatives, compare Sim with n8n, Zapier Agents, Make, or Gumloop using the same production-shaped workflow and governance requirements. From 51165a66adcfcb738cc509bf5fd6e96f66ee44bf Mon Sep 17 00:00:00 2001 From: Vikhyath Mondreti Date: Thu, 1 Oct 2026 09:45:48 -0700 Subject: [PATCH 8/9] feat(library): What are the best marketing automation platforms for AI workflows in 2026? (#8520) * feat(library): What are the best marketing automation platforms for AI workflows in 2026? * Pi Babysit: address PR #8520 feedback * fix(library): link the marketing platform vs agent builder guide --------- Co-authored-by: Sim Pi Agent Co-authored-by: Waleed Latif --- .../index.mdx | 317 ++++++++++++++++++ .../cover.jpg | Bin 0 -> 29605 bytes 2 files changed, 317 insertions(+) create mode 100644 apps/sim/content/library/best-marketing-automation-platforms-ai-workflows-2026/index.mdx create mode 100644 apps/sim/public/library/best-marketing-automation-platforms-ai-workflows-2026/cover.jpg diff --git a/apps/sim/content/library/best-marketing-automation-platforms-ai-workflows-2026/index.mdx b/apps/sim/content/library/best-marketing-automation-platforms-ai-workflows-2026/index.mdx new file mode 100644 index 00000000000..b0e0ad2242c --- /dev/null +++ b/apps/sim/content/library/best-marketing-automation-platforms-ai-workflows-2026/index.mdx @@ -0,0 +1,317 @@ +--- +slug: best-marketing-automation-platforms-ai-workflows-2026 +title: 'What are the best marketing automation platforms for AI workflows in 2026?' +description: 'Compare the best marketing automation platforms for AI workflows in 2026, including HubSpot, Marketo, Sim, n8n, Zapier, Make, and Gumloop.' +date: 2026-10-01 +updated: 2026-10-01 +authors: + - andrew +readingTime: 12 +tags: [Marketing Automation, AI Workflows, AI Agents, Sim] +ogImage: /library/best-marketing-automation-platforms-ai-workflows-2026/cover.jpg +draft: false +faq: + - q: "What is the best marketing automation platform in 2026?" + a: "HubSpot is the best all-in-one marketing automation platform for many small and midsize teams, while Marketo is stronger for complex enterprises and Sim is stronger for flexible AI workflows." + - q: "What is the best marketing automation platform for AI workflows?" + a: "Sim is the best fit in this comparison for teams that prioritize model flexibility, agent support, extensibility, and an Apache 2.0 self-hosting option." + - q: "What is the best AI marketing automation platform?" + a: "Sim is the strongest AI-native workflow choice in this comparison, while HubSpot is the stronger choice when native campaigns, CRM records, and marketing reporting matter more." + - q: "What is the best enterprise marketing automation platform?" + a: "Adobe Marketo Engage is the strongest enterprise campaign-management option in this comparison because it is designed for complex lead lifecycles and formal marketing operations." + - q: "What is the best marketing automation platform for small businesses?" + a: "HubSpot is the strongest all-in-one option for small businesses that need CRM-connected campaigns, although Zapier may be simpler for basic cross-application automation." + - q: "What is the difference between marketing automation and AI workflow automation?" + a: "HubSpot and Marketo automate audiences and campaigns, while Sim, n8n, Zapier, Make, and Gumloop automate actions across models, applications, APIs, and data." + - q: "Can an AI workflow builder replace HubSpot?" + a: "Sim cannot replace every HubSpot function because Sim does not position itself as a complete CRM, bulk email, landing-page, subscription-management, and attribution suite." + - q: "Can an AI workflow builder replace Marketo?" + a: "Sim cannot replace every Marketo function when an enterprise needs native lead management and campaign governance, but Sim can automate AI-heavy work around Marketo." + - q: "Is Sim a marketing automation platform?" + a: "Sim is an AI workflow and agent platform that can automate marketing operations, but Sim is not a complete traditional campaign-management suite." + - q: "Is Sim open source?" + a: "Sim is open source under the OSI-approved Apache License 2.0 and supports self-hosting." + - q: "Is Sim free?" + a: "Sim can be self-hosted under the Apache License 2.0, while current hosted-service pricing and usage terms should be confirmed on Sim's official pricing page." + - q: "Is n8n open source?" + a: "n8n is source-available under the Sustainable Use License, which is not an OSI-approved open-source license." + - q: "What is the best open-source marketing automation workflow builder?" + a: "Sim is the strongest open-source AI workflow option in this comparison because Sim uses the OSI-approved Apache License 2.0 and supports self-hosting." + - q: "What is the best open-source Zapier alternative for marketing workflows?" + a: "Sim is a strong open-source Zapier alternative for AI-centric marketing workflows, while buyers needing conventional application automation should also compare n8n's source-available offering." + - q: "What is the best n8n alternative for marketing AI workflows?" + a: "Sim is a strong n8n alternative for teams that prioritize an Apache 2.0 license, model-flexible AI workflows, and agent-oriented design." + - q: "Sim vs n8n: which is better for marketing automation?" + a: "Sim is better suited to teams prioritizing an Apache 2.0 AI workspace, while n8n is better suited to technical teams prioritizing an established general-purpose automation ecosystem." + - q: "Sim vs Zapier: which is better for marketing automation?" + a: "Sim is better for customizable AI workflows and self-hosting, while Zapier is better for business users who want fast hosted automation across common SaaS applications." + - q: "Sim vs Make: which is better for marketing automation?" + a: "Sim is better for model-flexible AI and agent workflows, while Make is better for operators who prioritize detailed visual data mapping across application steps." + - q: "Sim vs Gumloop: which is better for marketing automation?" + a: "Sim is better for teams prioritizing open-source extensibility and self-hosting, while Gumloop is attractive for teams seeking an approachable hosted AI workflow experience." + - q: "HubSpot vs Marketo: which is better in 2026?" + a: "HubSpot is generally better for small and midsize teams seeking an integrated and accessible suite, while Marketo is generally better for enterprises with complex campaign operations." + - q: "HubSpot vs Zapier: which is better for marketing automation?" + a: "HubSpot is better for managing customer records and campaigns, while Zapier is better for connecting actions across separate applications." + - q: "Zapier vs Make: which is better for marketing automation?" + a: "Zapier is better for straightforward business automation and accessibility, while Make is better for visually detailed scenarios with complex mappings and branches." + - q: "Do I still need HubSpot or Marketo if I use Sim?" + a: "Sim users still need HubSpot, Marketo, or another campaign system when they require native contact management, email delivery, subscription controls, landing pages, and attribution." + - q: "Can marketing teams build AI agents without coding?" + a: "Sim, Zapier, Make, and Gumloop can reduce the coding required for AI workflows, but production agents still require careful tool design, permissions, testing, and monitoring." + - q: "Which marketing automation platform supports self-hosting?" + a: "Sim and n8n support self-hosting, but Sim uses the OSI-approved Apache License 2.0 while n8n uses the source-available Sustainable Use License." + - q: "Which marketing automation platform offers the most model flexibility?" + a: "Sim offers the strongest model flexibility in this comparison because model and agent orchestration are central to the product's role rather than an add-on to a campaign suite." + - q: "Should marketing teams use one automation platform or several?" + a: "HubSpot or Marketo paired with Sim or another workflow builder is often the strongest setup because the campaign system and orchestration layer serve different purposes." +--- + +## TL;DR + +[HubSpot](https://www.hubspot.com/pricing/marketing) and [Adobe Marketo Engage](https://business.adobe.com/products/marketo.html) are the strongest choices for managing campaigns and customer records, while Sim, [n8n](https://n8n.io/pricing/), [Zapier](https://zapier.com/pricing), [Make](https://www.make.com/en/pricing), and [Gumloop](https://www.gumloop.com/pricing) are stronger choices for building AI workflows across applications. + +The right platform depends on the operating model. Marketing teams that need email campaigns, contact management, landing pages, attribution, and lead lifecycle reporting should begin with a traditional marketing suite. Teams that need model calls, agents, data enrichment, research, content operations, and cross-application orchestration should evaluate an AI workflow builder. For more context on this distinction, read our guide to [AI-native workflow automation versus traditional automation](https://www.sim.ai/library/ai-native-workflow-automation-vs-traditional-automation). + +Sim is the best fit in this comparison for teams that prioritize model flexibility, extensible AI workflows, and an Apache 2.0 self-hosting option. Sim is not a complete replacement for HubSpot or Marketo when the organization also needs a system of record for contacts and full campaign management. + +## What are the best marketing automation platforms in 2026? + +HubSpot, Adobe Marketo Engage, Sim, n8n, Zapier, Make, and Gumloop are the best platforms to shortlist because they represent the two main ways companies now automate marketing. + +| Platform | Category | Best fit | Main limitation | +|---|---|---|---| +| [HubSpot Marketing Hub](https://www.hubspot.com/pricing/marketing) | Traditional campaign suite | Small and midsize teams that want campaigns, CRM data, and reporting together | Less flexible than an AI-native builder for custom model orchestration | +| [Adobe Marketo Engage](https://business.adobe.com/products/marketo.html) | Enterprise campaign suite | Large organizations with complex lead management and governance requirements | Usually requires more administration and implementation work | +| Sim | AI-native workflow and agent builder | Teams building model-flexible AI workflows with an open-source, self-hostable foundation | Not a full email marketing, CRM, or attribution suite | +| [n8n](https://docs.n8n.io/choose-how-to-use-n8n) | Technical workflow automation platform | Technical teams that want broad integration coverage and self-hosting | Source-available rather than OSI-approved open source | +| [Zapier](https://zapier.com/pricing) | No-code automation platform | Business teams that want quick automation across common SaaS applications | Task-oriented automation can be less adaptable for deeply customized AI systems | +| [Make](https://www.make.com/en/pricing) | Visual workflow automation platform | Teams that want detailed visual control over multi-step integrations | Not designed to replace a campaign database or marketing suite | +| [Gumloop](https://docs.gumloop.com/nodes/using_ai/ai_web_research) | AI-native workflow builder | Teams that want approachable AI workflows for research and content operations | Less suitable than a traditional suite for complete campaign management | + +These products are not interchangeable. A company may use HubSpot or Marketo as its customer and campaign system while using Sim, n8n, Zapier, Make, or Gumloop as the orchestration layer around it. + +## How do traditional marketing automation platforms differ from AI workflow builders? + +HubSpot and Marketo manage audiences and campaigns, whereas Sim, n8n, Zapier, Make, and Gumloop coordinate actions, applications, data, models, and agents. + +Traditional marketing automation platforms commonly center on capabilities such as: + +- Contact and account databases +- Audience segmentation +- Email campaign creation and delivery +- Lead scoring and lifecycle stages +- Landing pages and forms +- Campaign attribution and reporting +- Consent, subscription, and account administration + +AI workflow builders commonly center on capabilities such as: + +- Multi-step workflows across external applications +- Calls to language models and other AI services +- Retrieval, classification, extraction, and generation +- Agent tools, branching, memory, and human approval +- Custom APIs, webhooks, code, and data transformation +- Automated research and content operations +- Deployment or self-hosting options + +A traditional suite is usually the better primary purchase when marketing needs one place to manage contacts and campaigns. An AI workflow builder is usually the better primary purchase when the main requirement is to coordinate models and tools around existing systems. Our [marketing automation agent guide](https://www.sim.ai/library/ai-agents-for-marketing-automation) explores practical applications in more detail. + +To decide which category your team needs before comparing vendors, read [Marketing Automation Platform vs AI Agent Builder](https://www.sim.ai/library/marketing-automation-platform-vs-ai-agent-builder). + +## How were the best marketing automation platforms scored? + +This Sim comparison scores each platform from 1 to 5 on six buyer criteria, with 5 indicating the strongest fit for that criterion rather than the best product overall. + +- Campaign templates: Support for repeatable marketing campaigns, audiences, content, and lifecycle programs. +- Integrations: Ability to connect applications, APIs, triggers, and data sources. +- Agent support: Ability to create AI-driven processes that can reason, select tools, or complete variable tasks. +- Model flexibility: Ability to choose models, providers, prompts, and custom AI components. +- Governance: Administrative controls, deployment choices, observability, permissions, and suitability for controlled environments. +- Team fit: The range of users who can productively build, operate, and maintain the platform. + +The scores are comparative editorial assessments, not vendor benchmarks. Buyers should validate required connectors, security controls, regional availability, and contract terms in a proof of concept. + +## Which marketing automation platform scores highest for each operating model? + +HubSpot and Marketo lead campaign management, while Sim leads model flexibility and agent-oriented workflow design in this comparison. + +| Platform | Campaign templates | Integrations | Agent support | Model flexibility | Governance | Best team fit | +|---|---:|---:|---:|---:|---:|---| +| HubSpot Marketing Hub | 5 | 4 | 3 | 2 | 4 | Small and midsize marketing teams | +| Adobe Marketo Engage | 5 | 4 | 3 | 2 | 5 | Enterprise marketing operations | +| Sim | 2 | 4 | 5 | 5 | 4 | Product, growth, AI, and technical marketing teams | +| n8n | 2 | 5 | 4 | 4 | 4 | Technical automation teams | +| Zapier | 2 | 5 | 4 | 3 | 4 | Business and operations teams | +| Make | 2 | 5 | 3 | 3 | 3 | Visual automation specialists | +| Gumloop | 2 | 3 | 4 | 4 | 3 | AI-forward marketing and operations teams | + +A low campaign-template score does not mean a workflow builder is weak. It means the product should not be mistaken for a complete campaign suite with a native marketing database, email delivery, and attribution system. + +## What is the best marketing automation platform for small and midsize teams? + +HubSpot Marketing Hub is the strongest all-in-one choice for small and midsize teams that need campaign execution, customer data, and reporting in one platform. + +HubSpot is most compelling when marketers want to work from a shared CRM record without assembling a separate campaign stack. Its [forms capture leads in its CRM](https://www.hubspot.com/products/marketing/forms), while its [marketing analytics tools connect campaigns and reporting](https://www.hubspot.com/products/marketing/analytics). This can reduce the operational burden of synchronizing forms, contacts, lifecycle stages, sales activity, and campaign reporting. + +HubSpot is less compelling when the central requirement is to build deeply customized AI workflows across multiple model providers. In that operating model, HubSpot can remain the campaign system while Sim, Zapier, Make, or another workflow platform handles orchestration. + +As of October 2026, HubSpot publishes current packaging and marketing-contact terms on its [official Marketing Hub pricing page](https://www.hubspot.com/pricing/marketing). Buyers should verify contact tiers, seat requirements, included features, and overage rules against that page before purchase. + +## What is the best enterprise marketing automation platform? + +Adobe Marketo Engage is the strongest choice in this comparison for enterprises that prioritize complex lead management, campaign governance, and coordination with a broader Adobe stack. + +Marketo is designed for organizations with formal marketing operations teams, sophisticated segmentation, long buying cycles, and established processes for campaign approval and measurement. Adobe documents its [lead and account profiles, audience segmentation, and campaign activation](https://business.adobe.com/products/marketo/profiles-audiences.html). Its strengths matter most when campaign administration and enterprise controls are more important than rapid, lightweight workflow creation. + +Marketo is not the simplest choice for a small team that mainly wants to connect applications or experiment with models. Enterprises can instead use Marketo as the campaign and lead-management layer while adding an AI workflow builder for enrichment, research, routing, and generated content. + +As of October 2026, Adobe directs buyers to packaged Marketo Engage offerings and sales-assisted pricing on its [official pricing page](https://business.adobe.com/products/marketo/pricing.html). Exact contract, database, and feature terms should be confirmed with Adobe. + +## What is the best marketing automation platform for flexible AI workflows? + +Sim is the strongest choice in this comparison for teams that need flexible AI workflows, agent support, model choice, and an Apache 2.0 self-hostable foundation. + +Sim is an extensible AI workspace for designing workflows that connect models, tools, data, APIs, and human decisions. It fits marketing operations such as account research, lead enrichment, content transformation, campaign QA, feedback classification, competitive monitoring, and routing work between systems. + +Sim should be evaluated as an orchestration layer rather than a complete substitute for a campaign suite. It does not remove the need for HubSpot, Marketo, or another system when the organization requires a native contact database, bulk email delivery, subscription management, landing pages, and campaign attribution. + +Sim uses the OSI-approved Apache License 2.0 and supports self-hosting. Teams should consult the [Sim website](https://www.sim.ai/) and [Sim repository](https://github.com/simstudioai/sim) for current deployment documentation and the [Sim pricing page](https://www.sim.ai/pricing) for cloud terms as of October 2026. + +## Is n8n a good marketing automation platform for AI workflows? + +n8n is a strong marketing automation choice for technical teams that need broad workflow integration, code-friendly customization, AI components, and a self-hosting option. + +n8n is the established workflow-automation incumbent in this comparison. It is particularly suitable when engineers or technically experienced operators will maintain automations spanning databases, internal services, SaaS applications, and AI providers. + +n8n is not a traditional marketing campaign suite, so buyers should not expect it to replace every function of HubSpot or Marketo. It is better viewed as an integration and orchestration layer. + +n8n is source-available under the Sustainable Use License and is not OSI-approved open source. As of October 2026, n8n describes its license in its [official Sustainable Use License documentation](https://docs.n8n.io/privacy-and-security/sustainable-use-license) and publishes cloud terms on its [official pricing page](https://n8n.io/pricing/). Cloud billing is based on workflow executions, while [self-hosted deployment](https://docs.n8n.io/deploy/host-n8n) remains subject to n8n's license terms. + +## Is Zapier a good marketing automation platform for AI workflows? + +Zapier is the strongest choice for business teams that want to automate common marketing applications quickly without managing infrastructure. + +Zapier's main advantage is accessibility across widely used business tools. Its official plans combine [Zaps, Tables, Forms, and Zapier MCP](https://zapier.com/pricing), supporting lead routing, notifications, data synchronization, enrichment, content handoffs, and AI-assisted steps without requiring every builder to be an engineer. + +Zapier is less suitable when a team needs maximum control over deployment, model infrastructure, or highly specialized agent behavior. Its best role is often dependable cross-application automation around an existing CRM or marketing suite. + +As of October 2026, Zapier publishes its current plan and product allowances on its [official pricing page](https://zapier.com/pricing). Buyers should verify how usage, applications, AI products, users, and overages are counted for the intended workflow. + +## Is Make a good marketing automation platform for AI workflows? + +Make is a strong choice for teams that want detailed visual control over multi-step marketing workflows and data transformations. + +Make's [visual workflow builder includes routers and filters](https://www.make.com/en/pricing), suiting operators who want to inspect branches, mappings, filters, loops, and application actions on a visual canvas. It can be effective for campaign operations that move data among forms, spreadsheets, CRMs, databases, messaging tools, and content systems. + +Make is not a replacement for a marketing database or campaign-delivery suite. Teams should also test whether its AI and governance capabilities match the complexity of their planned agent workflows rather than assuming every visual automation is agentic. + +As of October 2026, Make describes credit-based plan consumption on its [official pricing page](https://www.make.com/en/pricing). Buyers should validate credit consumption for standard modules, data transfer, and AI features because different operations may consume usage differently. + +## Is Gumloop a good marketing automation platform for AI workflows? + +Gumloop is a strong choice for marketing teams that want an approachable AI-native builder for research, extraction, enrichment, and content operations. + +Gumloop is oriented toward workflows in which AI performs substantial work rather than merely passing data between applications. Its official documentation describes [AI web research with structured data extraction](https://docs.gumloop.com/nodes/using_ai/ai_web_research), making it relevant for market research, web-based data collection, content processing, categorization, and other variable-input tasks. + +Gumloop is less suitable as the sole marketing platform when the business needs complete campaign management, contact governance, email delivery, and attribution. Buyers should also verify that required connectors and administrative controls fit their production environment. + +As of October 2026, Gumloop publishes plan information on its [official pricing page](https://www.gumloop.com/pricing). Buyers should confirm current consumption, concurrency, user, and enterprise-control terms before deployment. + +## What are the key facts about each marketing automation platform? + +Sim and [n8n](https://docs.n8n.io/choose-how-to-use-n8n) offer self-hosting paths, while HubSpot, Marketo, Zapier, Make, and Gumloop are primarily evaluated here as vendor-hosted commercial services. + +- [HubSpot Marketing Hub](https://www.hubspot.com/pricing/marketing) is proprietary hosted software, and its commercial structure includes marketing-contact and seat considerations as of October 2026. +- [Adobe Marketo Engage](https://business.adobe.com/products/marketo/pricing.html) is proprietary enterprise software with sales-assisted package terms as of October 2026. +- [Sim is Apache 2.0 open-source software](https://github.com/simstudioai/sim) that supports self-hosting, while current Sim Cloud usage and subscription terms should be checked on Sim's official pricing page as of October 2026. +- [n8n uses its Sustainable Use License](https://docs.n8n.io/privacy-and-security/sustainable-use-license), supports self-hosting subject to that license, and [meters its hosted service by workflow executions](https://n8n.io/pricing/) as of October 2026. +- [Zapier](https://zapier.com/pricing) is proprietary hosted software, with usage terms published for its automation products as of October 2026. +- [Make](https://www.make.com/en/pricing) is proprietary hosted software, and its plans use credits as a billing unit as of October 2026. +- [Gumloop](https://www.gumloop.com/pricing) is proprietary hosted software with published commercial plans as of October 2026. + +Vendor terms can change. Procurement teams should confirm security, data residency, retention, model-provider, usage, and licensing requirements directly with each vendor before signing a contract. + +## Which marketing automation platform should I choose? + +HubSpot, Marketo, Sim, n8n, Zapier, Make, and Gumloop each win for a different operating model, so buyers should choose the platform whose primary job matches their actual bottleneck. + +Choose HubSpot Marketing Hub when: + +- One connected CRM and marketing environment is the priority. +- Marketers need campaigns, forms, contacts, lifecycle stages, and reporting. +- The team values ease of administration over maximum AI-model flexibility. + +Choose Adobe Marketo Engage when: + +- Enterprise lead management and campaign governance are the priority. +- A dedicated marketing operations team will administer the platform. +- The organization already has complex lifecycle and reporting processes. + +Choose Sim when: + +- AI workflows and agents are the primary requirement. +- The team wants flexibility across models, APIs, tools, and data sources. +- Apache 2.0 licensing and self-hosting are important. +- The organization already has, or plans to keep, a separate CRM or campaign suite. + +Choose n8n when: + +- Technical users will own the automation layer. +- Integration breadth, code, and self-hosting are important. +- The Sustainable Use License is acceptable for the intended deployment. + +Choose Zapier when: + +- Business users need to automate common SaaS tools quickly. +- Minimal infrastructure and a broad application ecosystem matter most. +- The workflows are primarily event-and-action automations. + +Choose Make when: + +- Builders want a detailed visual representation of data movement. +- Multi-step branching and transformations are central requirements. +- The team is comfortable managing scenario complexity and credit usage. + +Choose Gumloop when: + +- AI-heavy research, extraction, and content workflows are the priority. +- Marketers want a purpose-built AI workflow experience. +- A separate platform already manages contacts, campaigns, and attribution. + +## Should I use one platform or combine a marketing suite with an AI workflow builder? + +HubSpot or Marketo combined with Sim, n8n, Zapier, Make, or Gumloop is often the strongest architecture because each layer can perform the job it handles best. + +A practical two-layer stack can look like this: + +1. HubSpot or Marketo remains the source of truth for contacts, consent, lifecycle stages, campaigns, and attribution. +2. Sim or another workflow builder performs research, enrichment, classification, generation, quality checks, and cross-system orchestration. +3. Approved results are written back to the campaign suite with provenance and review status. +4. Human approval is required before sensitive messages, audience changes, or high-impact account actions. + +This structure avoids forcing an AI workflow builder to become an email service provider or forcing a campaign suite to become a model-neutral agent runtime. + +## What should I test before buying a marketing automation platform? + +Sim recommends testing a real end-to-end workflow because connector lists and feature matrices do not reveal reliability, maintainability, or total usage consumption. Use this [AI workflow automation buyer's checklist](https://www.sim.ai/library/ai-workflow-automation-platform-buyers-checklist) to structure the evaluation. + +A useful proof of concept should test: + +- A real trigger from the current CRM, form, warehouse, or campaign system +- The exact AI models and data sources the team intends to use +- Error handling, retries, rate limits, and duplicate prevention +- Human review before customer-facing or irreversible actions +- Logging, run history, permissions, and audit requirements +- Credential management and separation among environments +- Usage consumption under realistic volume +- Export, portability, self-hosting, and license requirements +- The ability of non-builders to understand and operate the workflow + +The winning platform is the one that remains understandable and governable after the demonstration becomes a production process. + +## What is the best AI agent builder? + +Sim is a leading AI agent builder for model-flexible, extensible workflows, but buyers researching that broader category should use the dedicated [best AI agent platforms comparison](https://www.sim.ai/library/best-ai-agent-platforms-2026). + +This page owns the narrower marketing-automation lane: choosing between campaign suites and AI workflow builders for marketing work. The canonical AI agent platform guide compares products specifically on agent construction and orchestration. diff --git a/apps/sim/public/library/best-marketing-automation-platforms-ai-workflows-2026/cover.jpg b/apps/sim/public/library/best-marketing-automation-platforms-ai-workflows-2026/cover.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f1045337cb65e2236c98b2dab85fe29ef8eab909 GIT binary patch literal 29605 zcmeFYWmsHKwl~_i6Ck*|ySq!Ead-Ddg1aQR2W#BjU4sMzF^iq2ATkm)#3)Pk;l^jyEuBF_g zPE|H=3W__bfXoXs-4v5_!2j$u(nG@SvYmdVOhma%i~uNtWbMHV#^e-21{t6CsWeR* z1~fQx!kcq+HFj1@AIEE$8zx0q{?nIWhK%JJQtG~TO!oF#qIxw{Rw|SH+Btvr0zSbT zQTY3R)MXwaGyLS>|ES9U7XDuZ{uhD&Md1Gh0??GePZi}%=A(AMpvuF*26BYVko5*e z*$Myt@2y>l=qs)GXouxr-9mC;xicuQf?prK$zka7e|k+p4*c5=Tj6eT0KiUZeq@E| zn4ssc?En4>Rr-m{mETd7pHS=DcGB3_Sb@@2vunaxWMR0nm^j1e2k63F#gIC^`l${! z>&4YUByFz5w`nt0OHs;)00H8mawQJ$-}x^!{FKPmi9`$)rr7~pTrsN6>_ zd}y{DA#0`12bB$-whoStNdq%h&fpck;YVXJHDcuTS3>^Uq*WT*>a+Dwpd33kG@E3_ z=a69==WlePKEe0r!cwbbm!hzykc1&jag>pOfF+e<$EZg1JB2FcmBtI8q$mDs@P9)9 z9a4YOE?2sS;F-cEIoJ5{b2o!%1Vt*&XG1Wy@%Koa#Z~EVLvO83rQ&)Ll(G#Mf+SU` z58Mh;0UAk0U8t#AdYwfj<&XUdJf4Jo$@RYb_!`WW3 z0F#b6Cb>DwxaEnbDsT0!@4#zPXr&0a_L<#yvRN4F&eD#v zdYo?KP})d6NNMBFWVrt&>uw{ErdTvi50h2;Gv<5Dq7lQS)rZD(E8>gU>~FJhM%`Oq zqC8|87INr-hMLDE4gO~u8q-ghQ~Tdk?%&0bvAF=4fbr>TxNvr9MQkUW_k}=4yYf#d zEOCY_!>ykis0No%JjLH_CTlAyRBqMPapI`Dm%!^a{toT>Z^QtU&EKTm3i4*OXj-*B zZWm?WfO(Rnv?ywiSmVK0I*e*hU9%XuQ#Pn(G zx^O>yt>@cw+8cqJJV6qq)OTwK8^oLbGs~mx2f5&bgLRm>uLx4g)(JjQFb*AF31Kac$#9 zVdTp)V$r12&HZf<10O8`%OkWFDu-Q!T>^k*2Z@ku3cV|{`v-vStsB%faq63d`sh)S z5Q$!lDAF0goYQcg8XG2q$&CLKVyWE8%%0=%%>hLoO+yTi}v? zl+>=W(`*I7jZbdmbOl8qDEh=hafK#KZljY+-pt;!FOH0cMfAHE)n3?#Vj4xodKi^Y z4w%)+J~64$Zh5khi$FH?8|4+%8VW~+yyaGY!*nL62O%FGeBOJ)NeMu}cQEq4F7 zEVVNSwd!4`%XlnslQo(^yNuU?E=GoyupA2s4%k77pi5>?uAzokNoTv&+$z}2ZL$D? z7te&(9kXHn2lqbAZ1&-7_&*O1p46H_%FzKeDYI!u04`bl+XB+uiogn;(8j{FEH*+{ zh_{V3klxLJ3Op6>(a!V9e6h%xQFztd9ow|(nM08LFrzuVm90|nA{u82Z3IEY^C#2Z z!BgE(x!;mb7^&jkYz`*YHwTvFTn7dgVs{fhcV_`2v8o7HRj(HS-?!2Y?A(x`E^a~A zn3zn0Ls56ksT_i-0DK+xQ@5KcQJWS9KMkFITd9Mf)(_Jwt%{nP7qnf!y}y=WUFSGO zPPo@ItVd^xFbnhymQtx372wai=ir(>Dx6(sU1WOt`!1>1$fpvvx60KHQej_6&ld-d z^=08~0QF`{B&v@HHQPYbLC~1H&E1F-)$qaP)xLQJWBO_hHOj&B=eKbG2CR@EgHH^s zBt4o0#wKtj;x~LEBupRj7jV2@Nr)qQ4Lmb2y!Klh+)-0mJ;*Kj^ze2~{j>xd# zMrOwUe#TO}s)e!fN6|TQ`=+0M>L`9sIm%gqqe82@u4cfv(p&!Gy21MfTfa|V4-I(n zjr?!tgg4wtr}dfyI`0eWlHVgM8)06uvDl=b*7;JVe4cVUf?S*x{WB)gtnHl)ZbeK6 z@I4|d$euG4mwK9-p9Ol086>B-v~U@S>$tpNOJCpq497u{@D50r{F#AM%S@*gQ*L5A znbX9`&pCQsFK~8$vOPjkM#x2@^hyJ_UfJX5hgZ|+qMBF{h47SC_zJGg4LLm2g)7Jaa&I&`EtgLq^2i@YK*_#7{JBa5$cOh&eX?$XkA3 zhLsn91hVh)7@>ac4>a`O8hq?>RqL_d3p)7uCA|4X=PW<#zNf83$g+HR1AdBbrI?{i zY-xj;itm<%_f}iFn}Cu}Dnv^4iKCq+SJQD4uPN|4Dii$y1+yU!he_Q$$SP(J5AR*d zode4#q_gqbG?b20>9*F|M*=S7Ts3w~!JTX3C#DvNq+w#>mZaS#t#alNS2_z0o(^-g z494s{IVQ`?PxuN)Gwa6?BAtB&n?OSUH-q#gThM@k2N4o{tUoC>J><0Vtca? zv$kr(9U|t*QhKvxs7bh(S^(Zwga!3+Mbre3o8pEUo|})i7eh` z2D;$0Jk(j>@c0Q8qAPgbuw1Fm>0+<^Om(^ml{_i&EX!2}A2A$x{E{RB2|3p?#)lk^ z?K4wTp{qSVzys0E$TPlccp20-K)vp zLa+5t#-%t*&4{47MTlSjlE~u6zJt{YA_8g7uL$q@=l;2vCH~=V87M)<2nx*qJVsa8 z(Nv=`D`L}k%Mk*Fwu`HOt23Uw;DiiI*eYV+}9O2aMt+x!m%d3jXMJd?*aE!vq zdUz&YoC!40_yH#}#IW9E)WA52wqC_4a9GS!>K>@Jry77zxt3}uR=~Y4aSyskU~?{* zYE|rI9y;tvT_F53IXyeIo>e8Q>AE_+2hF8s(hk3z{lKMe{?5N!*}u(G;ILlc1)vw? zkL~*td;hj-E#=IlusMvOp~Y(?2lyT|c%kk492?Jql3ZdgBo^ZN={2nJwI6klfSuMuT(V(78g50o^sG!SZMenGfCC^ z=Jzk60j)}!rP4y*-$OT1+sF07dwP#yII#`UG5Gq+%zK`=nP{8UaFvGw`87MR!N%{B zn_HS!b3O$-I{kzhMKZG7*T58P`m5*!zO$Sw% z4_le%s+CoEp*~$Ya6;tuB%IaSxEc)iUWP;pOX4z0Ctov>{vu@Uprr3N`$7c!&Le^kicT*=Ex7sobbFV8P&?TXGP*rhtRdX5`MR%t+gc!E;-I)V{v z5pOs{=X2gY$ph^kc)0Ek$$s{5QLdTmfI9LBv*Y4Cd;uJlue55+vbSGKv{`Sz0Bn_G zFXmAYFpkb$+>Ci2jMmI^+iQLjoigK7ggtdts12x9ZN`|HR%SfQ*Bz%ZKx^q6%?zFf zNn0SfU<}2N$O*_|ZjT;?4jn#8@&QzIDcG*=NQzUOqHIt7P%pYKvhEvG%m|cCDr`Fl zWL;u+pNM3~#v1*D7O39C-hRY-0Z?b5k(Y5ydP2{bzIUDXWC;jSurTgO-I1|2S@UXs zvM=C@)_#g>@grf$y5mXUnjbiL0i4&X2w%eW=AX_FXXkzHHW1i)q!y_0TASWC`Ssb@ zE{II$LFjoDG}z0b9RExdfc1w*EUoufSjQ_1T<84ZeW3YmA0%(H?jZP@pq=i}b<_I& zv!IKVae_Ac?rxl3}fQ^sLQKR-Ty}t>@3%!KOL~Ko{(T1Cc2sU1}Qd zw?I1^VFR}w*vpKWZ#CXX)&SUCN8i+w>ufC%Td*>LxDx@>}2BF;q9U~NkDhmQl8Ha7u?MTX@UC@2Bk1|>G zgSR@^M?Vb#lDD>QWm3=zem&2EwTf~xTyL<}Exll^ds_^TX#EQ~OqU&L;P|^F_S$GY z!d|z0nb!!s)w*%YqSuk4HOXNcgjBn4(2s<)}wRNKG`?bF>vPf0OG9^ETjjYxIj*=@^%fudkvqhzqa zh1)ocE2?c{)Eh%~KQn)v0!!)GYajzBd!_I97VgS>D2~UmR))?AK;uXT@QCxw9M197 zmw{uF2wOM~0yc0zvf`O6U`3mo55E93t0vu;X{+9=Ej08F=iRi@go>D*t1iuH)}98QduGzFb%&W@ady!?h#Bqn zhJRCXQ%0#mTPNyf?Lo}Gy-X>fpJE`}q%9&|uPL*zRUOe<6#q&2xW7~$E1bZ$$XVJz z*#&9sR$JUS>jmI#kg_P-;cXukAR7uj$klWru4LD31hf*Y!pV$bj?z+G*m8;CQy8^# zU%Lee^jcJCbRrGv{X*o+HbW>px1aicNJYyqE+YXXTo@l$60Fz;GM)PR=$NzW8Gshf zw?|ZlJXri=*aSkU3@HJfM8D!DZ8S*I*{_#Q*Q(=aiAu3+`th5yp6*%ct^*XKZb$#^7TW%)m{( zu!gs`np5pCK^sQk&WC|l7ceB>0O0gQ)c~pONm3{x|kY458 znbdRN4s@<;q8lMqU|a{@K(_svW5zYKu7sUp^>75&p|(pwJXx{l#DI#hdE-M zA!2ap^SO=Ew}!VVBtNQ3T7P7^rK{VxcUzq9nT_L_Wi&v9^;Fu^kL660k;P|}nhIne zOJR^tCqzlko!c3h#=6-Mm9vdGA zfnQqj%G?T5>n7thg*};i*qK|kXxZQ4TRZ9}3Y8%%kDL2FjN#84)i2C3bJMLW?vBHx zY?Z@#w#e&Z4EH;*`_VUML;7Y^(x-tX-3Ok)ja~j}hW!jpt#RIfv4Ov|5QkTFgw8Tg zSn?^e$ByK8fUx+m=myxQ zJw^wl;JJ#68dZa`NQM_cjbHmOnY-NiOvb?Ou9a!No>6p=(OpyN#Y)r|`Y)}?p?Odb;Jg>dOp?vF}B8q)12+CP6(*?!6bDU31T(bq2JqIeyiH$r(1B%1!F!(o(mM z>~IWcfjO_D;#&N?_U7AS3Yk)^trIjecT_jVBy{opGfTSBY%Hm7V!Z zQN~i=g`>%^I+#X(% zabxphg`}3`!(7>A4X0CBlCgn#iarjHsS!LT2}?KsGF|6{jvkz!T3GFZ$aS(ZU7cd* z!Yn#!>(#m#*M8LU;(08dE-^88E#$bAmKA$ayZ}s1F)R^aCm4cktXuHS`t; zuEplgc%8nJFs+~GeiVE0>}xAA?i=`Eg13G^3yZfd-|2`Na+RJAQH}TW{rwYp>yaxNkZwskC;0F0*bfizr6m^9v87h_9^~ z2mb`{Y1*nib=DU;b!UHxy3ig^zz}KOw6aI4$x;8bu!d1zvVYtTJ}a-TCGU{dT;0rzRr!At%m{$1znKbITz`*OaDo1mv|3`&A$^ znSlGTwI=wB?xv{tvaskMA#`v+>p-DNcp2}BSSQU7DC5ghLl{%2x3KqFng>SNt`5|hZ)k++1;Xe^!LTHwbq(smj7$^-vsbI zucdMKeqtJx<%%t0iP5pHE*2c?I?gPB&~IcE8HVZMwRb1ySUvl$-G(MUU$7WOOulW!|%-x|EgfX%P_hPmY=OV4LD zJVu@%z%|tzr~hk;g<65NYp%My@4j^9w04mtGt#K-je6QLPAY>YyqQKxm^$vKLE@<> z`HTP`BPlJNJkGk)lsoCs^y+fid`rUkQ|vnyoKlc|QU&&4o?Mk$BPskgx1ip*K(t^S zC0Fd_A6&w7nd~cVs-jt0y(#nd!XDg_qwb)PP&q_&;U~PhpHPHf5aGtr>~qIlDH;ki z%yp+g6;<{P&+|3lUL6N-lFkRlv9DZk8FTb2xi&c5ojf*CV;r$mkt0d5HUzSc$0?2~ zT#@by64O|j6_t_oINs;3yzE!JMp&Pc9)^r)<9v41F~Tti&G+L-j7nW-UjR~S_f5I9 z`Ix#uwmE%A*pdt@tHFHecAp0}lkdy)Rvl@&NG($@1$U0VPgubmBbYzdFt7%NV+Y+vu5hBKy{1OJBFLdi zzlNzACwW=BLL0TVBu5bU64g-3u%g-mwQ8KPxF!sWKfXmxe}X`NXBr1Fag?pQ*#PIt zM{8anGK*`{L$ZX`fpC19cge$jFO`rg@A;D{v@IM|Z3_^&U)fB5YKYhxg`f(BQj&I= z`+8!)Y5D~qpDLr0y;lf;NyzW*<5ZIqlb3y;_X0?`39JtSdxEw3lL2j2mMZ|U3iCcV zt{XGCIlqOw-pSFh*h7NYZLzwvQny$Of$3_p^L5!#8x0$r)CH&94x%~d2cTomz7QW6 zxBU^8`l^jxYhdt$2u^d0B0)>=)iOc)Cb)&e)s0sL8`J#1RaHQzMJeOSjlTP`9?bRRtc zbm>+sNvZdPrmq?d1ZCy8v*sujcKS9F|D-{0m=B4A`Hn-MVAE~fnq)_dOLNm~!^|8w zcYZCRZO4YVcGfnkanRMUXSOkt46#*`3F{an7;0<30E)afZvPw}r<|$Phg+8sxmZHc z62OnX8w~|?Sce62fKPugIa1pe@1HCAG{Psf)HZx5t++b@sIX54oNE;~!J&V_!H)I$ zq5p?we+Lhz)qA@6==N5|y$Rl|4e!=DKQjd7%~^eQWEXmA9U=(&Fe_}015o7J?)4~Rmx-@sypYyo0oGzj+V z4?vw2(W~b0fU4a)!_Yl}ei!4;oue@RoI20;^tX0fdgNKitf?X;ximb6GGkW{O2>FI zM7mWRbWoC(xqX2{^PCS9RYtn-bU3LZVT#*$+%ZLFc`aG4rHw6&Sn|O*8Y%EVqmBk+ z)T%1-O-DsM=lAy{3Eb#w7pyfQicy4Fbwc(_ieI>^=22kBZ|k(kpq*7$qZfAdFMBOF zDNuxxLZv*DmfF3Uv~XiJ1!*NY#Rmw?=3jMLBD6h%carmUV`5^&GM(3!EXSWQRhYH; z|E|qt#Sfe4rrIsdtr}PdP7Ua_SEDQYq7=<{35;0FXAxS|mV{~CY7mpuR@AqQnx~R} zc1^D>^g{*;mZLirCu=iHSV3&lgU)SF{dG@)#dwDD#yWkTzt35t{ z-7LvaasT~jC|^Onb}s4aI+T&5PZ?JUVsrWYYwo=wPp}dPP`^7ez@gs_dzW8Ve|0m3 z-`sce`#|zL1kPcC5ILI|`=P9Zo6=3#SN%Ffh#id92BIskTVQgV(MCF z+GwhZd>hz3Ht-!?M?+8ldm~U0ViG;3b+3Dxrwr99G=hh>)cfP@e%VFKOezheoSBX< zccD;!S%K*4Cp}c=lb(7>eQK@O*&#|uALJ2@2jiI+=a#jUVMqi9{C{n1wgP~uKlo^rS#yHp|O7CI>j0jZ*m7TZvDHK z!PoN(-P`YxH;31HZ+}1Zx=<&UXWmcv8`H2R^c|1qUd|5I@Z#koD{NpxC@W*Vh9Ogk zr>=DoM{&Bye01Fx01#Fj{i+#7Ddf<2%EC)08=uNl8cA=So-3Ou*5QU>cbp~5uC5KZ z##VjE(NjGo5VUm;F+2i{vm(Bp`Q&>p`}Fcr>Q>is3tr|d8PW0TU#`q|5ZAggzR&id zRxAaQf)K+PH3N8ch(I~6V*;Ti#50EiC6y01uj&#v;WkEGkFw+9fN#!6D_-j?5ElYh z6!ntJAVe4mRm_QS;+lGf7^~X34tSQPqeES?-6|=2huo+B`T6Ttji7hPZ;wB7&*F7J z@ZN3F1Q1w9n6hh8`T{T(lLsQ&#f^f#yvZsXS=;%Nr3mG%tyQ7fjAxM58$ibmE2+`B zci&RY|bcEds4Kso`?)jAcH`b!f2SA}!0o<3)(ulRovfyxtO* z{j~d6yW7CG)#m*wJ%yG7ix+_V{@5U#ZuA&RR;u?*drH{$wd(}LCl_hbtuyIb%X?Q~ z_yM4U4NLl&Ct}?$i&;}Qp1X6v5JndY@GJ8L;HCq=D`%EDuDLb^5&Nex5NYyi5k2A# zc|=PG*k;5qfp)@Z8kTvd@q?aiT?1F_$E^tKnR1#WV?Y(gzlve&g zSVLPwhgkZt9iNgc7tK@&q5UYI$Uhr=N&|bI!ZSjkuFdWAZ+HSfZxYAuERN0i;}FVN znUK&+yn!j*xn%{t-W&O6sYYH9x5xCG{n+Jh(lqJFigP+vQh_kLqFlOcGa9#?5L>5- zedOEB$?j@O)fn*KJ|NN&ai(N)UPMHgGXef86dLKaz>y=mo|Z(`VtybQ8B00IP%ax-m-E{! zp>-7PJe>R|^boeRx+t#l9IEZuA8YbJicuMa&#&$Q|9BU7=0k;c$+mAuLykDERDEj{5Wvv62_r=!X|px@9?*^E+Uco%f`Sfl%X1(l<-w3KI-U|r z=)praBI@L&q5?JBt0Jq&UrD|(&yrMa*WTLFZ4v%O`x4>j>HVu)BP+o0^gPzVo`{}l zo0=qNNE0CmXx|eyE)~oz*fE)`_m}X`zwj{f3VcxhOZ>h1+NR(1q7RgAUkQ=glC*3& z_(gp;r!!D#*U?4CumyYcg}(#dEeCEMXwNRE(_aeQY&~uWY+szMz3>0fubMRUR=d*! zt;SJ4N4)ad>lO)Fnq?@Fr5b(Xrm~|*j}~PI??UTZ`~vj_05!qWuYtJ~c^%uVz@lBN zC7<|huJZ*Trf~Xa!%jkbY)C`8{_c)O+3O#?!6dkNuE7{9R&GQqZ)(1Dp~Ni9bVN5M zn>bo&5+7)oFA#BR`dK+|zE*t!%O@Uab(27i%59~i#zGC2tHpQ{8 z{7F=Ld42cE&Z%xag>rIjLs{KAzGu!A|3x;)u>6Ij|vEVQt3-1fv!3nv|M(RJ43< zsR2znWf|eLgmq2TrfEN!RZ%}BTC~qW*lx}B8J17 zn|paf6HB1Re1#*E!<6;IIc2?drG0@(`Fi4_`4UI0#c_NV55%#fh{;rY7R?DS<3Poy z{4?yoFnR%C|HB#a=84IpEMV(Pt($SdBZe*>#IE733O`OZ96SB2l~YoVZSk8IZJa(k zszBtkS#DrgQ;Q?XdtR8@jUR;dGOj^#4Z69t<>$5F)AwH35YGQJ4=jde9|Av29MbCq z)~@((Iq5jp#JhxjjH|ucHFGj{uv3@9%o1uj-p`ICO8spWI~L=pzUE^1XUc`xMxcEw zoM3_{dHs%H9Bap5v7L_0L*sqU*mL7Xz%58Si$&Kt0qh!$zkbIIoE)q9xk4>xO&PmG zApLlAzkf(O%^k)p=u+tg&{{Idwuk%zfU)|DW4V%2 z!l0mDwZKrAG;iTd3{#UW1zGgO=8Uuhwz?`VjZoG(J6-_EpVySC$SlyDjdHf5IBgX2 zzy3*Ip$Ez?P~_l_nbv1wHT!!+a&8k^{;1b}qf`_Q`f}$CtD*XrH{%4JS=P8)-3U(`4UXPHdq4@oTDQ3>m=`N9>^x&BPC8B>L0LpbaSm zvfQjX1%h~9LKw9MVmC>eZT}r#FUP* z3MaV}iv;sKccJEkcPt=luFUSQY$V}7IAhxzzR!I(Cn_m8ozX`DdEIe#ymNGO?()aI z*HfZiLaW?Wb}&d3cP-O3@Y5LQr2LFFmoykrdtTI%t?67k)3;dY{iWh_J4=n}=+QpJASpSS)`0my?zh99Fwaw&-Q zQcXEZqAgc&QiA&U+`A{a>B!GH%`O{1gE9N{?G=6=rIp%5ZCKE3-Zm|l!+kL(JFcO-IVAB;0VmL z`HFL}-t^d^+(BYvNHf}PfhY;cr!df7#uR!x9j&n-FJr|r#vk7`VFBD?^gdOUm(|Z= zCKyKTYuU6C+?UoiF0XAs4V`RpVXP)l_O(j+R`DgKp5}~|44!*LWD(yi=X~0D3+20I zIx?q(CwzZ_H7|?R;Rc9jT)9d=lwwR*s}IAxbJw7@6}zYNpq#O;hy}8RBVaKGVTjpp z6V937+N!FZzC#at8@%*FGevlMn<_$^{F>|s(%g!qi4iRxbyC@iDV&c|MCP42^Q)}f z3K{p~wk8-{dF@5WIN1eMlyMk(LcC+?`8E^~C&MKuk_or6vP z*aX#6bK!iIJ=32!te8>GvCGK&DVr7oV(%J-X7-6U@Teuoue<`Kl0FpvSzRK&(}`3O z`W(1(A~_tHR-moj)H%-CYeDeyj-$U-{n!dIsr$D|a7vjj0WifSmZ8#x^wo(EyANAJ zVJ$Gsk$sHk!Y;bxq<6QD{<``6s;<6jX}qh}gtzf7M)*^G_`uR9{GClZi-C>9rRD@- zap0JxaeVM$>3XGnBi7d5T=>2bEb~wzh4bDVK^pzsV^|}T4gX+6`oXnY01S4;u&iO8 zZbgf6MYfN}c{v~F-7rM+bsb6R-YF~UP5<`SPDAGf?#BGHfYF~fD-8P0EA}q{bNjO% zV~}rO%^@y6MmSWgobiCioj?#@S2_jW4*c|6HL~*cko*6n}KU(Ue=J3JjeQD13ly5 zg4CeCHtg<_?fnpLjWF?PEtS<-$X;(%=BbSYF1pP+$Dv`dnv9G^PZ@=7^%ggfqIwqz zFH7rF)d@ko#$QHKHYHCP;uU0X=H}a@yqwGzIySvZYWu3aqzrZhejNo&$ezj$H#uPTv>{lmHSP5qK%ByT}b+ zt%=%MSI|80unHlx^$ZSrLUWI-` zgmy_Hg^Nz&zw-TStdNmcjFa69fVlS5!#{`If4I5Db30Ej$4fC$g!h~ZA0j<{+Lx_P zT$K@9r4957@~@T@qDmqqLft5MiBvwXk)$8-+~X70XWb!dv(XavtP*w7noS02S%2d& zJ)v(TCp>jsWouvRECZ-huK)*)jEB^?d-<}&+mEo{M2rEMMGw4Jm-H+M>OLQLL4Eup zZTbQTaUS)JqTf@~S8lJI4W&XH#{8CiQ;p+HSZI8ADQuI5W7;lV2_9}Z0dvwywV4F# zk`lU{%5w1XuvmNC|N4=2r!@Bu^k1!PrLBZnx=o!W)@#K(5g>!Pmh+e}@aMw8@#S3y zMBMl~`gVl7u27gbTWB%J^Lc{G2FNqxKa3ZRU}%|^w(Gza5e%!Qer%LESOf>>CGr`j zHs(IJEc)6b`3ds4Wj$mV>P*@ShO|W7Wa>C|YkHG1*EzB@MDmrLMvTKWxs`5Tj+3hY ziq4IHw(wsJ6f>AR?J{sv`MvLDusV^(+mTJjF~t?Jan|umh*f)4cWMZ6oPQV+)MLFk zX{gSd2}Unq9@Rc~$cramMDN<1Ey`wXPXz>%T)9ntDabV+8|EM8fYc0_8c)SaIk7(! zY13gq22x`-)mP_Q7758*OMM;HQa$hZNmF?$q4`M@fwwMU)g5B(ntVSrL_;>?XQ%`- zqmpt*JIiv|>5tu%F;vZ**3Mkf(p7-w!B#1iOiaCt$~O2AGymyhatJXjKdQeS4N0i! z+$~NtF0DxzV!S})H`C%h9axF5-8*DuS&mQ({Q7z#?2liL83x~B1Q6wy`s!stEqeS^ zc^gUrHOw9%HNVKMmws9c2ZvENsya6=kWzpc&k@RR3YBhYLt9pdtCy0pe4GTPMKAlS zi+{D_pNhl1&+ra_+#f8_udb1jOaW-e!G=Z;EiPlAl??cpoM&2mxWXWSEa)=zHbS;# zjE^QGU9bf}7X{&9$aYHXqISEAR$jfj+ZlHDBOIM*pL}0DG%>uWfQ*_M^JLXD)5H*ao)h5KdrU38LNDK)^rp+b$To&=;_1XIXwk$x}i~{28DW@d6wzJVurk*`|tjH$mx9E>~?x`gv*HA+@6k?eB>Du0M2Krg$VFv21MKXLW%o=4n>$v3UkDTr(WE3UO zw2pX-;zkzKMTpDV)v__5h7*;FJ>gWj2V(?(oyjYm%WKgN54+&nES}2<3%lYBa$ANU zPjPI?+G}CZCQW+MCQ8>x-^&!9XJ04dRntY%GcdNMaPy2EG_IeM-dGLa+Eq`}Swv~* z&^}6j{3EqXY)>R@QM{OIcMy1I&FyAhl8w+hNFYDPv@m+V`)LLTp+&z$ch9GNklBIo zJBo>PQ&D7p0D&qZ$@hh%PCParMB2l4j&y7Wc^<)9j9-F>SYZ? zYU-}%0y$V*%MtfcmYi6w4cHs@vNY{aJgWc_#fZrT7WgYAwwvmHb=GFa?aLBfBXSb% zL1I63uEfYLF5kqulsjU^Cu{qZWjwB?&p__0&oM)8b#82R8DVo|XW&Q5!L|fV_$wn(l^3Zro zoK2=rUUx?5|+Ug1;n=u(xxN0n}-CYIsnkHCc2dUyYyh@TQ1&vX*ThZB=g}*+^~Uk(H>m* zr(5&>1OD-m%f?#es6x9 zmt39IU$13*5XKrKgrYUS_iVrk!*mfg9-j37^#UL(<<){sZb#ncx82l*M*kT!$1>p} zv>vGTFLg<=LOC0>Vb3Uw-w?%Ap+(3c6h-E-T2SDX!^+rj?4xCwhXm!F71X_9?`?c2 zUtu02T6L7$b0@;zHt(6mHOURBuO8obL;!Ku{Md`e)72^wJ(oHFRhNCJGZd3qvb?O# z?y-1IsFL3{?htnOP5U`}qA_q2pd@_$%8{kFY}-W(Wf2d`HEbzIw+!14q)lw;uk0lK zTQzAkLy8OLO5ZzU9b*oLfLgbl_`B0n3ePAC++9aAQCgUrpOl8$vT*g`&!r^KwZDxA5x)uA4ScTF4{9;ZQUWY*fBUGEAl*y#P* zf>wh7p1s6zj+R;_&7SJVX%WE)9*2#hU|Vss3F+5?v{G&@E;e9Ck6lp)tZ>#Np>3K ze*uk4{&3TGN=IL}_tf*!^Yp9SVQpcTH};h88%f0PaLXs+?Toy~j*3V~|3oq!a2lGo zRW27s$!cLHcmlIdma#BDwGg(S9EB;%W6hNLizgbRVkI&El8;AFx$Bhj4Ni+I4bUZ# zr0Q8Vp*Uk`4u_Yn`3H`mc-L5@u>Nh_E`neu1NOhBO}ygr~X% zDbz%xpQZz8}bH`n;pCM znNP#{G`EHe4Uz&MW#8o9I9>g8@)Ba5B$-N3&&9DGdM6skhn}oCb|uhQ9>wUweD}E? zO6C_ppaS-Htw(pNNab+khzw+uk@CwqmJ&Oy-25Fg#_ys!NQ%5GQZc_zNQkffkRkfq zVJ%Ya1j4AoHVuQro)1s`JuwL270@3fj~75L=cp({BhLYEOP&CnDkKRA28^T}S|757 zQQi_)5q_MUfUA@lkei4sNa$)YmmZ%=qQAuD=IqRpCs>Osz;#eKaozeDl&pKz9UTp4 zH8?a~&J@4Hgx#Bb(B&29Z2x|afW!%4By|U9%mCIk{wa_%d~1KEvn;2H@rU3j=ww`4 zB-Ow>6E}%K;)f-7U`*7m^B>+xdh!PDx+znj|v#iY9;EY_krCl(RMRRJmS?JX|_=j@S z3InP|EE2UBjmH_~jYDHZ+-C{AYCe1mSUMK)^4+`Qabma5!i_9|9|$2=+Xk+E#3)vd zqrAO6h7co|)f|^8EOx zMz-vbW$~FCw!WKKpv(v0Qd?4`^DmVtzj*Y|BUo8RUJWJQ`%x%+%)0zWmY`CKl#hsOh;_SPjV zL~n;=>e-3U4T+Fr93@gCx-@Vl3wAA<3rcAfxnO&41eaAbRD{&VSTf~~$k#xy)VvKo z_+V~)N)VV5sOMGlHif5{XyzO{FrI%+QuaHWbaPTqKF_*qbePQ`rYBm>+j`{wVyyK< zYV#$L!YKLtl$f~6BU|715mYX%4N6`R0k)-NMn1@$$kd!f@0Y~0c8j%L`cVCG{6e2V zpMecq!DsmtiwPmq4U^`&eJRGQoXGb3W0Q1fF0WFtpJsp8r}&6Awr$_cd8wdLm`Z(LUz%e>{e9iAua|a zI6X7nrn&FNG0AYHR^j{1RPiz;T;3W6}^X>QfY4VsR$4cAQ<;+3Lar8H(=noVs3v?|`F+Qw!zyXANFZ5{M7@~vJgv0mIlX42{ zH|!9hdA*F4zg5bxB+?~nIzdI^e4J*g!gb;pIvaE&OrVzcuX0`L=vV?R6P~htXDVY0 zF7QgM=7VZZmXzN+bR80fvRkdchDqaHWNKF1Q$c|i(nvugE3VglYF(hm1M2{stu537 z)dcX&I7JZ`7?6 z_&hJ2I8p0TfodaMGO$3=qw0SCUwa9JaD${YGho};fIeXtQWpr`?_(S<))G3<&&hsbfUp(%Ir;qMTwj2NvK5zyi)qv1S{i*0>#^qg07gGt)4tEt|)4n@Lz=ZmxgKYF`_3Mv33Yo3D_Z ziW%=4+i<54XXx%fb!e4;1X-BNGXDnw8I*Nh;kN4E%^+@!tYI9X_^i_T4X}9?m=o5@ zky+VNijx#!cb+_KXn7SNDZKmWf#%S0+h0LvJo)L<{#lYG8JP1!>L4NAhAlAv+70=7 z3Jdr4$bOik`%@T4?b1Zoa=VdpYF)lbSe^uUkdW5Gi+Zwi%BFyb{uI`Os}wpM%r7o4 z>=)tkp!&`?aewi3GKfmDBZmaE@bKhaTkCuz*|eOvdiGjE%$gCqU|Qg^0OU6pT9s8B zwxs#q{zys@{=)aZ5BE38?f9YPu zN9ywK;mSiQQ4KVcFIWt;hzBBqOu*Zky2S3@T&hl_XhD!iwBdsHbKj3&^@z$YGHatP zOc69?WHpEY4;QBei^+SG3c<~G_(7spGQ7=MJ+968k#o7n)4`O1*OY(0eogRpyv0`% z$1S(=5;{RxKvpn;g>166oA%ZH&YN$p$y=C>g!Vh5){{%4{z;!kF4pxUuF46f85N`;gTEqjnu zlBhfs2zK$>c${Pf`#V55(rA zql90|2=sT9qh=Tc9_wY0&^l%6cu-R7LtnlKS+Gd&znNHqc-oN;)dz70i$BUpFs(in zKNUTcA~tm*Fkjxp$jnf(IJhEi<%BlAE>vqsHP6}SM3rTUU5TsJj*$t69bHi1>emMIMKyXOC4=$d#i$@~SS znWIZCpfD(|euq+gfJ{h1=Uz?&klC;C-KAw)4v7(f$eu7G>Qu4@SN9^tta^r@2mgD5 zZs7ajjuXxJDBuAq@c9qm$LBqkHQ8EtSEGae%79OxCx)hv8v4P06vu6qj;&)7vRug= zTPrUC@}H@Td4va1aCW&&dOdHn;XP!PV1f8gs5iAs(WxA=Gz>OqJ&O)>7U&O=ijVAe z*T+JI$}Xnk(KMi51pfF;f1dv35 z7{oIXWH_$UEW)5;VTT`_NKbuq**84$x&jF`Vhc21J2>V9xPXTJmie(_sF97U=<>p6 zqjr|PD%Epznz?2$nhi+@V3DU^piffigJ2e23%c=qO~v}%1EvU-RaOf+jhEPl$#3_X z>_?~eb?>aZqn@M2i&_}pTCWVfa>9=JA8%n}?|L$Sx_!#~|%r@0Jt;H6HJ?#D-1aTQs`LZa3 zKW7=6o5n$7rCL2kZ@_}G;Ar`HN(KJR#5V5-he*;iO@Xn{-s+d(Qr{Y44$%Pwu%#{Q zUjo~=;~y&D{af22QQ}+O3@sN|S<5cjV(tGs=1>;tMies3_i`-$$DhGW)l~ygCH(ru zuRIT5KhR1-@~nMh2tSD9%u)~Il7+1Nv5+Q9`;uvPeE?)w5`Mu=+WOP(?87Uit>TA% zZ?hdts+KugUWz6FpP9f{v-RVKMwkRvyzor3%ZjUD#{0Q6DYAg22uw+B;B}+478>ks zpFP8y5KnV{T<@whvbapVnWq+~1C_$-yxvR`WMpI#HFcUb3N-{3KgL-QhHSAqf10jl zLM8rs&ab$U_1T^C)3Uo9PcG{4RVS!#oaPy?NT6Vt%O8;uv=Hg(gwk&Q^B|#t(`*CR zVE74<4fs8f^z~F(gzBqme!|k+9Q|AZAHflt3sRlVt!nxw}qev~5Bj!@HN0}(=^MA)X zQv7`FNh04N+R^+XMi3OE+RDJh{Z9A@b{Ka!g+Xt*MH9p6Fl_V!Tq${u_6nnzAb)wX z2|;$`aG{Fn?yR>U1D1)nzE|c=pt9>)u(Q`j;Sk2ZPENo)ZN4ThGYe`@Hm~qi^fA5o zm(QG2D(*5MBWRD=C>L95oM?@PA3p#~d?XDPq~$S7j-@5T z-op}d9xz+S_#lEUq~Pjwd3u8W#zBRHm3 z6G!Yp{PO95koOgkoOqv!*BN19E3$M**5>u%iQ`Avs%6-QvnHT}J6BPoe zf-c9g%_BSydSu5xbDM={)5C?bTI45J>DIOYSj2ugz1hG4*ULh1<=*!0wWh;mgA8(X z+~O^K;Y_EDM!MoA%Xz>1-{kmnjBl+%YE^v__=cX_Bt9aZ$E0&)8=+rOqM)+C)NR~*`l4h*EK3Xy+G7pkd{znvk5de-R)j5A-w?1Nr~ zlz6=;@*%Adn(v|+wrZRGkJzNhs1>PM^>UoX$OW3Z_^VapgFr-kR?IAF`zv5Tfj~R} zxV|Q!EDm)Nctf}$s$ifx*is`PAqcxlNvW?A@055suXwR}c1zHO9y= zHVp7j)%z=DxeJ{%L4^AbndFz}Vn^xgVpb8-bKLnKt?zKHzN@o4ahKyQ809Uxz)EId z_sHeUoMP$jgb7gta7mcv>MJ~7{FpMquj*PPAKSczTiOi>u(v`vI3vFTTS~2u)@>{0 zgGVK5oNOAPy>J$3Yr7XRODl%vv#X-goXE^_d%|*8DLd!Bv;a7mPeml?&JWN&REMqs zBOR-9ixU$SAipoU<5EOFC0$1SsS-V*(%AwQxG21SX*#1N^AEt6;X`i{s(uPXvn4S?d|U0;pSe3|zxSiQW`7-Rh@8?F4b<}9Cx@jwNKsg8J^09b z?sfIi!*p0{HBdUOp7#a=Ihso8q1rqKKhyRD$dXS>67E zpI-K4%zuJkTAm6Ax+&rvk?^V;>9{|-(@h`YwZv?*5ihd{=pn;iS~1t#1R`U*?<_fh}29~UkdK;VA9#;s+QJ=EcU zm6rY=&83yeAXk5F=%#vIY{vWg`WFLJG6?ow)`{r~uLSn@!_3y&6y*~An5NN^Gmxp( zog&-3uUrfciyp2JOY`(?$uq7Kq{<_4?v$p?AbF;*zWdJ5Ri+QXuXqFY0z5NoA1QKU zp(i?JF19Fot3rN)k(*>WQ0l?ccF1ZLed$a4z0tjR6K3m@o8X)af3_j(GE9fB0jKfO z8oXQ3{F`%0lkc-1q?2QlbWBpXO$Ame&eRwy}=@>sxdKM~JX~vJN&EYK^$&w^! z74ok#a?_B*I?lE!$@8{3VR~kKAl9%q( zKp(|gwhs>O0`bYz2!AR6AotH;L;I~F*RS|<^Nz=*exQBZAjfKo+`^-gPW|98e%gP2Et)R$?mtg23X)+J`V+~^=dZql;Wr?h$e1DP()=9pizM(yYrxND_qdd$>tJ2NL>*iErjVvy^l4SR9;Cp#A2wgs-V`7iP<TY^{oYX?3Coq+&K`toR-%4{P^3FU9FC) zV`t&t{o&1%M_0JP)|vq%t7KNXmc`&B-=r3*i6Dwh8=4^#GUa0%hxfxahz~5r>!$m} zV%@zMHQm}JSPhQ!UGAUOhL%fB;mgdy@k6$RoYx}ay6B+eWx+V#t1vh}zCof-ohPnY zq?Hl}vK}G}HQcsoo%W$-!y7d-|1sijeAbFrIV8Ez8Lw2qc;6m&mkjQ3&`^n@Q>;)Y zLE9ihd6{gMX*>Omg?F}c%lSJ$`VK{Wve^)vFpd%y3yXU}zh}-Uk!TSr05c>K<{v`Qa-Sd-S_7sAFy)>YQ+f9d+y z^5=ZV`%PfTgtvtb%qrOwt6Fo1I%%Aadu5V`=Xk3mgO!z&&61(^NQ4my67wu+ICwy^xw1Kht2XHLL5HEZu zzm)iKa9iCsU~l8ky!Zly3UrG^zJZHVZFg}$6y@twwb*jyL3Yfon!~v6>#jJ1yNK;F zx~^pRpk)n4ZWKLU?Y=!)`lX#zeN{j-^VRHw?OErYK%=_16V$!|@m5CJ1wlmXWn3rk zm~yq0ia;9|X6e8?oh~-I7;gnLg1&P*bDEh`)RIII z*@Jw&AmM5v_GTW$^Q;3lXhDDKLz<&oDa9mN#v8%NUdi!^A1A#R)PcRQSCL5Z*sb}p z@|upG%B_C@Lr)wY8KrP~&LKRUU-coAcTJ4K+F+$Nz24 zIYbLN#o)BnV$IjK+J}V)GjBF#^I#O-|Lr4oe(9Mgi)&4;h=Pu7@0ie3yxSd07XT~P z*PFi~fOBdg%5Gnlq2!Q+dMPos3$y&bS3Cn+U@Pjw3H<7@vuG;h2&seLn1{*c$*4HD z70r*=d{nSxQiV4(G&|PfeC&C7pU9X5<0^U5Eby_+GswEH)Fke2n;4?|z>tkiU69CEwMLEW zT4$9+En|3MN}}60`S896_ zft+T;93>k;EiDQJK}<~mq%{|_Idp@W;vge*_?ONl=CDc`K0oGzS1;BwOae-{6g&QR zdS~~~WW(S-8OQ*35EoI?^1!LEDxZ?r-EQR&C*pxWE2>8f>sm=BX5LSj@C?Y^<}i@J z&{l`J1r#cQ!^IoIU!B#gv8Ze(c#gPA`VVZgBQTZe9|uC(!?YLIej%tfUcc>GR66rd zVGQQlWC)lw8JV9yy3773I9$#i1Gc^^IlwKd13mM4?aBXs`x8H11=wLKa6zgwTH(pbu7YqtSk+ps3OtB*!mg4!9SwEa=SSc8IA}9vAI5mu=*xzJ;5U8EeZqs#YI6^!%}76HS0tQXJSX zuk+ny1KmJ&*y`9^s#|vPJ)Yi+VnjQAWepeaqkjOI(1E)Rz=1!TRlhX*HHQ(*E{H45 zqogaJ7wJ-c^bMukpcmiyy0U#Mw0ASPyi65UQd4AwHhV!PtLDDTlv`*an$YH~# z)%%RBJ6K~sblfo~fjvf2&Vey4C(>>%Ta?7HWgVi#e7X9!)0PaUPXu!3| z;KIg2ISk7?!#$mS%_jcZ|31@WQ-sKdB{_nHgNdT~IfNEYDCzqWLLeALR2(&n;UNn- zRQK@LS#0f-EnB&quXz-%Kuan8#&BhZn9f&ARzB#j$(L0FFY(cmx4)Dmk|4n(3)Ufh z5xqK+fjZpU>(fGtt~S8bRHbOY3Qyh?JCA#{yh77!>jMdGtOC}nlNRsuz`_2FXcx^l z&m7$G6;V_FC1RDIZ_(C}9MYdbg#!Z3)r1h1e45SXvcc#Nd~K(u>dACX&xD6s6~b}jo9u%5lO!KfTnO1{n z+&g0YI|jm+w@7Oy2p(=QQ+0M;_u9&jn_x8V275gL_iLo)`)h03=>PlyOmNP-H-T|- z#Iy992`|MvP2}ce`l}xoNlgleJkD96e2M%Zz{N-L3I-4)NN|hf&BXQ7KgZcXf5{4x z>LiojWB8;30C56-#s`ilZgn%4J?4&xKQbuPVQpdmXavcxGMbpWpLIR{v=6$#NClwk z9zvbS@O0<~#u|i*P7ba=^?Jl;E7}oue~iIDYdZ50=9zCa<8Mj?8f zN~!nl&RtyC2KGzcOvz3-yrLsWx!E<__24C2N7WYmMz?)kzo~7G_sFMxFD)4Kp}&}MRqRF<^sFF* z-h(FduIHc;%(mPN#dtVu5A-7rhFP|R9!a)YGjATP%{4i6!4{KV-Oga279)i`(Ywud?T4Rh3=gvxP&Iufs? z*_sFPNllaEh3d}jM+yc)5->Q%P%pfvDx?C{N>?4LLY}c*pe5nMmZ-M7wH4fY6kc{7 zm%#N5%QI4$6}sc&mj4Z4#E~;W>hOjOmGV1fZybF0n}hrI=$%}p)XKCjTPtuoDfU?M zxg%NHeK+NdOA2YrXZ)Sz+cO=#^i^FaC>jecQy;vkBA+MQ^IN7AGZyu76FuXp7Irz_N#N-9?zCM3`R* zc6>YZ{92d_+jzjkofVTc()myUUrPQ5@Q^<38A;?WDz&-X{j~To zI9G8}%wC}=cq5(g>XXlAjd4V#XWe|%6(ALVq&#^Y`_O}Or-D08pEM&7}4{)F1;tW2{2cFHf;Rcif zQGitvYEuof-^SA&1NxJHz{*dh?$mP>vwBM1;U;?GxA(@qaA+3*W72fFEG*R7@zQ($ z&4^!ELAWgarD#z03f>F<@4>|*wEUfnZStO7M{M`m!|jh&&fqsP9Z8&<&YY+?;Egq5 zX?S1sUt=k$%TP}bgmPWork+E(_CM8YeHbE@$|Wu|hNS=Vvw|P+Kt~LeXuswrJ`9RC zRsq_p-YNpEFkp%|*q{mRPGsGq>hO#vI}i8kGfK&ef_IIN$Ew7-`}y0)w09jkUWF8~ zX&g#_F@?PqgHDO5m!J@OoHAsu1g=S-#&_U+eS$T9O ziBW?o~jMB1Ah)I$q>)Z>o}4)x44K|U@?>wOhA@!Zg57~u1G)L)j;ri zLpj-xVN2z+lJ1?X@Zt`@ORP@T-QGZI0}R@uha zDWf2^j9*U?jxY73_g}E{WJ$L+D(>`#S7hAs=DwS4QHb&>;vuQa`+XbBI2oFTR2CQ~ zF&&1kDt607!OA=CBDdAo{m!-Llroh8r#WG-N)sqVnH=Kb5~QNjs+KwQrn$O_X{I*> zq0_Dis~|7~02To=6xxaI9SIyna1eSa*!YKuVC)u^w1mybH{r;MWG7U80W%y7ULf6^ zn~4=92E<6kNcy@kn1ud24^Ts~+X+o_%e2s7&P!(EtNgjFkHIb&uba zV??NZEx_LfIW<%W@mO*DZ_m4%9V+Sm$Ig}bumS*0|E|;`Bp`SI&?)) Date: Thu, 1 Oct 2026 09:54:46 -0700 Subject: [PATCH 9/9] improvement(search): run Search retirement as an operator command instead of a deploy step (#8522) Deploys no longer register the 0027-0029 Search retirement. A long cleanup in the deploy migration generated heavy WAL and stalled application writes, and held the release until it finished. The retirement is optional storage reclamation once live Search is on, so it now runs as a resumable operator command, paced by --pause-ratio and --max-rows, with --maintenance running the index rebuilds and vacuum and journaling completion. --- ...016_backfill_search_vectors.integration.ts | 7 + ...27_retire_search_embeddings.integration.ts | 43 ++- .../0027_retire_search_embeddings.ts | 246 +++++++++++------- .../0029_retire_all_search_embeddings.ts | 28 +- packages/db/script-migrations/index.ts | 7 +- .../search-embedding-retirement.md | 149 +++++++---- 6 files changed, 307 insertions(+), 173 deletions(-) diff --git a/packages/db/script-migrations/0016_backfill_search_vectors.integration.ts b/packages/db/script-migrations/0016_backfill_search_vectors.integration.ts index e3c92cf8955..5f2d238637c 100644 --- a/packages/db/script-migrations/0016_backfill_search_vectors.integration.ts +++ b/packages/db/script-migrations/0016_backfill_search_vectors.integration.ts @@ -396,6 +396,13 @@ describe('search projection upgrade in PostgreSQL', () => { AND s.vector_512 = subvector(e.embedding, 1, 512)::halfvec(512) AND k.content_tsv = e.content_tsv` expect(complete).toBe(1001) + /** Search retirement is an operator command; a deploy's full registry run never starts it. */ + expect( + ( + await sql`SELECT to_regclass('search_embedding_cleanup_progress') AS progress, + to_regclass('search_embedding_cleanup_targets') AS targets` + )[0] + ).toEqual({ progress: null, targets: null }) await runScriptMigrations(sql) await sql`DELETE FROM embedding WHERE id LIKE 'upgrade-%'` }, 60_000) diff --git a/packages/db/script-migrations/0027_retire_search_embeddings.integration.ts b/packages/db/script-migrations/0027_retire_search_embeddings.integration.ts index b1e54720bd7..34738fd2c78 100644 --- a/packages/db/script-migrations/0027_retire_search_embeddings.integration.ts +++ b/packages/db/script-migrations/0027_retire_search_embeddings.integration.ts @@ -1,6 +1,10 @@ -import { retireSearchEmbeddingsMigration } from '@sim/db/script-migrations/0027_retire_search_embeddings' +import { + retireSearchEmbeddings, + retireSearchEmbeddingsMigration, +} from '@sim/db/script-migrations/0027_retire_search_embeddings' import { maintainSearchRetirementMigration } from '@sim/db/script-migrations/0028_maintain_search_retirement' -import { runScriptMigrations, scriptMigrations } from '@sim/db/script-migrations/index' +import { retireAllSearchEmbeddings } from '@sim/db/script-migrations/0029_retire_all_search_embeddings' +import { runScriptMigrations } from '@sim/db/script-migrations/index' import { readTestDatabaseUrl } from '@sim/db/testing/test-infrastructure' import { sleep } from '@sim/utils/helpers' import { generateId } from '@sim/utils/id' @@ -174,13 +178,7 @@ describe('retiring dormant Search embeddings', () => { await sql`INSERT INTO embedding VALUES ('new-ordinary-chunk', 'search', 'search-doc')` await sql`INSERT INTO embedding_search (id) VALUES ('new-ordinary-chunk')` } - const migrations = scriptMigrations.filter((migration) => - [ - '0027_retire_search_embeddings', - '0028_maintain_search_retirement', - '0029_retire_all_search_embeddings', - ].includes(migration.name) - ) + const migrations = [retireAllSearchEmbeddings()] try { await runScriptMigrations(sql, migrations) const preserved = legacy === 'ordinary' ? 502 : 501 @@ -710,4 +708,31 @@ describe('retiring dormant Search embeddings', () => { await sql`DROP INDEX retirement_hnsw_idx` } }) + + it('pauses after each page for the pause ratio times the page, so a manual run leaves the primary idle', async () => { + /** Every delete page takes about 100 ms; with a ratio of 3 the next page starts 300 ms after it ends. */ + await sql`CREATE TABLE delete_page_started (at timestamptz NOT NULL DEFAULT clock_timestamp())` + await sql`CREATE FUNCTION slow_delete_page() RETURNS trigger LANGUAGE plpgsql AS $$ + BEGIN INSERT INTO delete_page_started DEFAULT VALUES; PERFORM pg_sleep(0.1); RETURN NULL; END $$` + await sql`CREATE TRIGGER slow_delete_page BEFORE DELETE ON embedding + FOR EACH STATEMENT EXECUTE FUNCTION slow_delete_page()` + try { + await retireSearchEmbeddings(sql, { pauseRatio: 3, maxRows: 200 }) + expect( + (await sql`SELECT count(*)::int AS n FROM embedding WHERE knowledge_base_id = 'search'`)[0] + .n + ).toBe(0) + const starts = ( + await sql<{ at: Date }[]>`SELECT at FROM delete_page_started ORDER BY at` + ).map(({ at }) => at.getTime()) + expect(starts.length).toBeGreaterThanOrEqual(3) + for (let i = 1; i < starts.length; i++) { + expect(starts[i] - starts[i - 1]).toBeGreaterThanOrEqual(380) + } + } finally { + await sql`DROP TRIGGER slow_delete_page ON embedding` + await sql`DROP FUNCTION slow_delete_page()` + await sql`DROP TABLE delete_page_started` + } + }, 60_000) }) diff --git a/packages/db/script-migrations/0027_retire_search_embeddings.ts b/packages/db/script-migrations/0027_retire_search_embeddings.ts index 62554665582..f58f5b14b01 100644 --- a/packages/db/script-migrations/0027_retire_search_embeddings.ts +++ b/packages/db/script-migrations/0027_retire_search_embeddings.ts @@ -1,3 +1,4 @@ +import { parseArgs } from 'node:util' import { resolveMigrationDatabaseUrl } from '@sim/db/script-migrations/database-url' import type { ScriptMigration } from '@sim/db/script-migrations/types' import { retryOnLockTimeout } from '@sim/db/scripts/lock-timeout-retry' @@ -18,7 +19,8 @@ const SCAN_ROWS_PER_MUTATION = 4 /** * Rows one page may update or delete. Every retired document is a non-HOT update touching each of * its indexes, and every deleted chunk cascades into its projections, so the write cost of a page, - * not its scan, is what can outrun the statement timeout. + * not its scan, is what can outrun the statement timeout. `maxRows` lowers the starting limit and + * the ceiling it may grow to. */ const ROW_LIMIT = { initial: 2_000, min: 25, max: 8_000 } as const /** A page slower than this halves the row limit. */ @@ -28,12 +30,23 @@ const SLOW_PAGE_MS = 30_000 * to a size that timed out. */ const FAST_PAGE_MS = SLOW_PAGE_MS / 4 +/** The longest pause after one page, however slow the page was. */ +const MAX_PAGE_PAUSE_MS = 60_000 +const LOCK_RETRY_BUDGET_MS = 60_000 + /** - * Each page, committed or timed out, is followed by a pause as long as the page, up to this, to - * leave the primary headroom. + * How hard one run pushes the primary. Each page, committed or timed out, is followed by a pause of + * `pauseRatio` times the page's duration (up to a minute), so the run is busy at most + * `1 / (1 + pauseRatio)` of the time. A page is timed through its commit, so a slow synchronous + * replica or a checkpoint stall lengthens the pause by the same factor. */ -const MAX_PAGE_PAUSE_MS = 5_000 -const LOCK_RETRY_BUDGET_MS = 60_000 +export interface RetirementPacing { + pauseRatio: number + /** The most rows one page may update or delete, from 25 to 8,000. */ + maxRows: number +} + +export const DEFAULT_RETIREMENT_PACING: RetirementPacing = { pauseRatio: 2, maxRows: 2_000 } type Phase = 'documents' | 'embeddings' | 'done' @@ -87,116 +100,132 @@ interface Progress { */ export const retireSearchEmbeddingsMigration: ScriptMigration = { name: '0027_retire_search_embeddings', - async up(sql) { - const hasTargets = await sql.begin('isolation level repeatable read', async (tx) => { - await tx`SET LOCAL statement_timeout = '120s'` - await tx`SET LOCAL lock_timeout = '1s'` - await tx`CREATE TABLE IF NOT EXISTS search_embedding_cleanup_progress ( + up: (sql) => retireSearchEmbeddings(sql), +} + +/** Retires every captured target, resuming the saved cursor, paced by `pacing`. */ +export async function retireSearchEmbeddings( + sql: Sql, + pacing: RetirementPacing = DEFAULT_RETIREMENT_PACING +): Promise { + if ( + !(pacing.pauseRatio >= 0) || + !Number.isInteger(pacing.maxRows) || + pacing.maxRows < ROW_LIMIT.min || + pacing.maxRows > ROW_LIMIT.max + ) { + throw new Error( + `Search retirement pacing needs a pause ratio of at least 0 and ${ROW_LIMIT.min}-${ROW_LIMIT.max} max rows` + ) + } + const pause = (pageMs: number) => sleep(Math.min(pageMs * pacing.pauseRatio, MAX_PAGE_PAUSE_MS)) + const hasTargets = await sql.begin('isolation level repeatable read', async (tx) => { + await tx`SET LOCAL statement_timeout = '120s'` + await tx`SET LOCAL lock_timeout = '1s'` + await tx`CREATE TABLE IF NOT EXISTS search_embedding_cleanup_progress ( id integer PRIMARY KEY CHECK (id = 1), knowledge_base_id text NOT NULL, phase text NOT NULL CHECK (phase IN ('documents', 'embeddings', 'done')), after_id text NOT NULL )` - const [existing] = await tx` + const [existing] = await tx` SELECT knowledge_base_id, phase, after_id FROM search_embedding_cleanup_progress WHERE id = 1 FOR UPDATE` - const [snapshot] = - await tx`SELECT to_regclass('search_embedding_cleanup_targets') AS relation` - if (snapshot.relation) return Boolean(existing) - if (existing && existing.phase !== 'done') { - const [target] = await tx`SELECT id FROM knowledge_base + const [snapshot] = await tx`SELECT to_regclass('search_embedding_cleanup_targets') AS relation` + if (snapshot.relation) return Boolean(existing) + if (existing && existing.phase !== 'done') { + const [target] = await tx`SELECT id FROM knowledge_base WHERE id = ${existing.knowledge_base_id} AND is_search_index FOR SHARE` - if (!target) throw new Error('Cleanup target is no longer a Search knowledge base') - } + if (!target) throw new Error('Cleanup target is no longer a Search knowledge base') + } - /** Creating the snapshot and resetting a legacy cursor commit atomically, once. */ - await tx`CREATE TABLE search_embedding_cleanup_targets (knowledge_base_id text PRIMARY KEY)` - let afterId = '' - for (;;) { - const [page] = await tx<{ after_id: string | null }[]>` + /** Creating the snapshot and resetting a legacy cursor commit atomically, once. */ + await tx`CREATE TABLE search_embedding_cleanup_targets (knowledge_base_id text PRIMARY KEY)` + let afterId = '' + for (;;) { + const [page] = await tx<{ after_id: string | null }[]>` WITH targets AS ( INSERT INTO search_embedding_cleanup_targets (knowledge_base_id) SELECT id FROM knowledge_base WHERE is_search_index AND id > ${afterId} ORDER BY id LIMIT ${SCAN_PAGE_SIZE} RETURNING knowledge_base_id ) SELECT max(knowledge_base_id) AS after_id FROM targets` - if (page.after_id === null) break - afterId = page.after_id - } - const [first] = await tx<{ knowledge_base_id: string }[]>` + if (page.after_id === null) break + afterId = page.after_id + } + const [first] = await tx<{ knowledge_base_id: string }[]>` SELECT knowledge_base_id FROM search_embedding_cleanup_targets ORDER BY knowledge_base_id LIMIT 1` - if (!first) return false - await tx`ANALYZE search_embedding_cleanup_targets` - await tx`ALTER TABLE search_embedding_cleanup_progress + if (!first) return false + await tx`ANALYZE search_embedding_cleanup_targets` + await tx`ALTER TABLE search_embedding_cleanup_progress ADD COLUMN IF NOT EXISTS reindexed_through text NOT NULL DEFAULT '', ADD COLUMN IF NOT EXISTS vacuumed_tables integer NOT NULL DEFAULT 0` - await tx`INSERT INTO search_embedding_cleanup_progress (id, knowledge_base_id, phase, after_id) + await tx`INSERT INTO search_embedding_cleanup_progress (id, knowledge_base_id, phase, after_id) VALUES (1, ${first.knowledge_base_id}, 'documents', '') ON CONFLICT (id) DO UPDATE SET knowledge_base_id = EXCLUDED.knowledge_base_id, phase = 'documents', after_id = '', reindexed_through = '', vacuumed_tables = 0` - return true - }) - if (!hasTargets) return + return true + }) + if (!hasTargets) return - const startedAt = Date.now() - let batches = 0 - let mutated = 0 - let rowLimit: number = ROW_LIMIT.initial - /** The largest limit the run may still try: half of the smallest limit that timed out. */ - let ceiling: number = ROW_LIMIT.max - for (;;) { - /** Timed around the whole call, so the synchronous-replication wait at commit counts. */ - const pageStartedAt = performance.now() - let page: PageResult - try { - page = await retirePage(sql, rowLimit) - } catch (error) { - if (!(error instanceof PageMutationTimeout)) throw error - /** The timed-out page rolled back with its cursor, so it is retried with fewer rows. */ - if (rowLimit <= ROW_LIMIT.min) throw error.timeout - ceiling = halve(rowLimit) - rowLimit = ceiling - logger.warn('Search retirement page timed out; retrying with fewer rows', { rowLimit }) - await sleep(Math.min(performance.now() - pageStartedAt, MAX_PAGE_PAUSE_MS)) - continue - } - if (page.done) break - const pageMs = performance.now() - pageStartedAt - batches++ - mutated += page.mutated - if (page.transition) { - /** A phase change may include a full recheck, which says nothing about page cost. */ - logger.info('Search retirement phase changed', { - transition: page.transition, - batches, - mutated, - }) - } else if (pageMs > SLOW_PAGE_MS) { - rowLimit = halve(rowLimit) - logger.warn('Search retirement page was slow; halving the row limit', { - pageMs: Math.round(pageMs), - rowLimit, - }) - } else if (pageMs < FAST_PAGE_MS) { - rowLimit = Math.min(ceiling, rowLimit * 2) - } - if (batches % 10 === 0) { - logger.info('Search embedding retirement progress', { - batches, - phase: page.phase, - afterId: page.afterId, - mutated, - rowLimit, - elapsedMs: Date.now() - startedAt, - }) - } - await sleep(Math.min(pageMs, MAX_PAGE_PAUSE_MS)) + const startedAt = Date.now() + let batches = 0 + let mutated = 0 + let rowLimit = Math.min(ROW_LIMIT.initial, pacing.maxRows) + /** The largest limit the run may still try: half of the smallest limit that timed out. */ + let ceiling = pacing.maxRows + for (;;) { + /** Timed around the whole call, so the synchronous-replication wait at commit counts. */ + const pageStartedAt = performance.now() + let page: PageResult + try { + page = await retirePage(sql, rowLimit) + } catch (error) { + if (!(error instanceof PageMutationTimeout)) throw error + /** The timed-out page rolled back with its cursor, so it is retried with fewer rows. */ + if (rowLimit <= ROW_LIMIT.min) throw error.timeout + ceiling = halve(rowLimit) + rowLimit = ceiling + logger.warn('Search retirement page timed out; retrying with fewer rows', { rowLimit }) + await pause(performance.now() - pageStartedAt) + continue + } + if (page.done) break + const pageMs = performance.now() - pageStartedAt + batches++ + mutated += page.mutated + if (page.transition) { + /** A phase change may include a full recheck, which says nothing about page cost. */ + logger.info('Search retirement phase changed', { + transition: page.transition, + batches, + mutated, + }) + } else if (pageMs > SLOW_PAGE_MS) { + rowLimit = halve(rowLimit) + logger.warn('Search retirement page was slow; halving the row limit', { + pageMs: Math.round(pageMs), + rowLimit, + }) + } else if (pageMs < FAST_PAGE_MS) { + rowLimit = Math.min(ceiling, rowLimit * 2) + } + if (batches % 10 === 0) { + logger.info('Search embedding retirement progress', { + batches, + phase: page.phase, + afterId: page.afterId, + mutated, + rowLimit, + elapsedMs: Date.now() - startedAt, + }) } - logger.info('Selected Search knowledge bases retired', { - batches, - mutated, - elapsedMs: Date.now() - startedAt, - }) - }, + await pause(pageMs) + } + logger.info('Selected Search knowledge bases retired', { + batches, + mutated, + elapsedMs: Date.now() - startedAt, + }) } /** @@ -380,17 +409,36 @@ async function validateTargetMarkers(tx: TransactionSql): Promise { } } -/** The standalone entry resumes the deployment cursor and journals only a completed retirement. */ +/** + * The operator entry: resumes the saved cursor and, with `--maintenance`, also rebuilds the indexes, + * vacuums, and journals the completed cleanup. `--pause-ratio` and `--max-rows` set the pacing. + */ if (import.meta.main) { + const { values } = parseArgs({ + options: { + maintenance: { type: 'boolean', default: false }, + 'pause-ratio': { type: 'string' }, + 'max-rows': { type: 'string' }, + }, + }) + const pacing: RetirementPacing = { + pauseRatio: Number(values['pause-ratio'] ?? DEFAULT_RETIREMENT_PACING.pauseRatio), + maxRows: Number(values['max-rows'] ?? DEFAULT_RETIREMENT_PACING.maxRows), + } const url = resolveMigrationDatabaseUrl() if (!url) throw new Error('DATABASE_URL is required for Search retirement') const sql = postgres(url, { max: 1, max_lifetime: null, onnotice: () => undefined }) try { - const { runScriptMigrations } = await import('@sim/db/script-migrations/index') - const { retireAllSearchEmbeddingsMigration } = await import( - '@sim/db/script-migrations/0029_retire_all_search_embeddings' - ) - await runScriptMigrations(sql, [retireAllSearchEmbeddingsMigration]) + if (values.maintenance) { + const { runScriptMigrations } = await import('@sim/db/script-migrations/index') + const { retireAllSearchEmbeddings } = await import( + '@sim/db/script-migrations/0029_retire_all_search_embeddings' + ) + await runScriptMigrations(sql, [retireAllSearchEmbeddings(pacing)]) + } else { + await retireSearchEmbeddings(sql, pacing) + logger.info('Search retirement pass finished; run with --maintenance off-peak to complete it') + } } finally { await sql.end() } diff --git a/packages/db/script-migrations/0029_retire_all_search_embeddings.ts b/packages/db/script-migrations/0029_retire_all_search_embeddings.ts index 5cfa5b6b582..1e6d5263256 100644 --- a/packages/db/script-migrations/0029_retire_all_search_embeddings.ts +++ b/packages/db/script-migrations/0029_retire_all_search_embeddings.ts @@ -1,13 +1,23 @@ -import { retireSearchEmbeddingsMigration } from '@sim/db/script-migrations/0027_retire_search_embeddings' +import { + type RetirementPacing, + retireSearchEmbeddings, + retireSearchEmbeddingsMigration, +} from '@sim/db/script-migrations/0027_retire_search_embeddings' import { maintainSearchRetirementMigration } from '@sim/db/script-migrations/0028_maintain_search_retirement' import type { ScriptMigration } from '@sim/db/script-migrations/types' -/** Supersedes single-KB retirement receipts so every deployment receives the expanded cleanup. */ -export const retireAllSearchEmbeddingsMigration: ScriptMigration = { - name: '0029_retire_all_search_embeddings', - supersedes: [retireSearchEmbeddingsMigration.name, maintainSearchRetirementMigration.name], - async up(sql) { - await retireSearchEmbeddingsMigration.up(sql) - await maintainSearchRetirementMigration.up(sql) - }, +/** + * The complete operator-run cleanup: every Search KB's retirement, then index maintenance. It is not + * in the deploy registry; the 0027 entry runs it with `--maintenance` and journals it on success. It + * supersedes single-KB retirement receipts so every database receives the expanded cleanup. + */ +export function retireAllSearchEmbeddings(pacing?: RetirementPacing): ScriptMigration { + return { + name: '0029_retire_all_search_embeddings', + supersedes: [retireSearchEmbeddingsMigration.name, maintainSearchRetirementMigration.name], + async up(sql) { + await retireSearchEmbeddings(sql, pacing) + await maintainSearchRetirementMigration.up(sql) + }, + } } diff --git a/packages/db/script-migrations/index.ts b/packages/db/script-migrations/index.ts index ec10731c9a3..7700a8f80a3 100644 --- a/packages/db/script-migrations/index.ts +++ b/packages/db/script-migrations/index.ts @@ -10,7 +10,6 @@ import { projectionAclSkipUnfilledMigration } from '@sim/db/script-migrations/00 import { knowledgeProjectionAsyncMigration } from '@sim/db/script-migrations/0024_knowledge_projection_async' import { scopeKeywordProjectionsMigration } from '@sim/db/script-migrations/0025_scope_keyword_projections' import { userTableSchemaForWriteMigration } from '@sim/db/script-migrations/0026_user_table_schema_for_write' -import { retireAllSearchEmbeddingsMigration } from '@sim/db/script-migrations/0029_retire_all_search_embeddings' import type { Sql } from 'postgres' import { backfillTableOrderKeys } from './0001_backfill_table_order_keys' import { backfillPausedBillingAttribution } from './0002_backfill_paused_billing_attribution' @@ -59,8 +58,10 @@ export const scriptMigrations: readonly ScriptMigration[] = [ scopeKeywordProjectionsMigration, /** 0026 installs the schema guard every table row write takes before it validates. */ userTableSchemaForWriteMigration, - /** 0029 expands single-KB retirement to every saved Search target and completes maintenance. */ - retireAllSearchEmbeddingsMigration, + /** + * Search retirement (0027–0029) is an operator-run maintenance command, not a deploy step: + * see `search-embedding-retirement.md`. + */ ] /** diff --git a/packages/db/script-migrations/search-embedding-retirement.md b/packages/db/script-migrations/search-embedding-retirement.md index d9c531b50f3..86d48536e5b 100644 --- a/packages/db/script-migrations/search-embedding-retirement.md +++ b/packages/db/script-migrations/search-embedding-retirement.md @@ -1,48 +1,92 @@ # Retiring legacy Search indexes -`0029_retire_all_search_embeddings` runs through the existing script-migration registry without -cleanup flags. It supersedes the single-KB retirement and maintenance entries (`0027`/`0028`), -including databases that already recorded either receipt. It snapshots every knowledge base whose -persisted `is_search_index` marker is true. No Search KB is a completed no-op. Once saved, the -snapshot stays fixed across retries even if another Search KB is created. Ordinary KBs and the -selected KBs' live source/credential configuration, document metadata, and backing files are preserved. - -## Deployment and execution - -The app and workers must already use live Search, and older indexing jobs must be drained before -this cleanup ships: deployment migrations run before the new app switches over. `SIM_SEARCH_LIVE=true` -(the default) makes `isIndexedOrgSearchEnabled()` false. **`SIM_SEARCH_LIVE=false` enables indexed -Search again.** The cleanup does not inspect this flag. Live source setup may still create a Search -KB for configuration; it does not index content. Document uploads, dispatch and queued processing -also honor the indexed-search gate. - -The ordinary migration runner starts the cleanup automatically and continues until it is complete -in the same deployment. There is no page-count or one-minute deferral. A successful run records -`0029_retire_all_search_embeddings` and its superseded names in `script_migrations` only after all -selected KBs have no remaining chunks or unretired documents and index maintenance finishes. -On upgrading a legacy single-KB checkpoint, the snapshot and cursor reset commit atomically. The -scan starts at the beginning once so it includes other KBs behind the old cursor; previous deletes -remain committed. Maintenance checkpoints also reset once because the expanded cleanup creates new -dead entries. A completed legacy checkpoint does not require its former KB to still exist or remain -Search-marked; the new snapshot selects current Search KBs and preserves any KB now marked ordinary. -An unfinished legacy checkpoint still requires its target to remain Search-marked. Subsequent retries -resume the saved scope, phase, cursor and maintenance checkpoints. -The existing maintenance implementation rebuilds HNSW indexes and vacuums affected tables before -deployment continues. +The retirement is an **operator-run maintenance command, not a deploy step**. Deploy migrations no +longer register it: a long cleanup inside the deploy migration generated heavy WAL and stalled +application writes, and it held the release until it finished. It is optional storage reclamation +once live Search is on, so it runs separately, paced, at a time the operator chooses. Self-hosted +operators can run the same command. + +`0029_retire_all_search_embeddings` snapshots every knowledge base whose persisted `is_search_index` +marker is true and supersedes the single-KB retirement and maintenance receipts (`0027`/`0028`), +including databases that already recorded either. No Search KB is a completed no-op. Once saved, the +snapshot stays fixed across runs even if another Search KB is created. Ordinary KBs and the selected +KBs' live source/credential configuration, document metadata, and backing files are preserved. + +## Before running + +The app and workers must already use live Search, and older indexing jobs must be drained. +`SIM_SEARCH_LIVE=true` (the default) makes `isIndexedOrgSearchEnabled()` false. **`SIM_SEARCH_LIVE=false` +enables indexed Search again.** The cleanup does not inspect this flag. Live source setup may still +create a Search KB for configuration; it does not index content. Document uploads, dispatch and +queued processing also honor the indexed-search gate. + +## Running it + +From the repository root, with the migration role's writer DSN on a **direct or session-pooled** +connection (the run holds a session advisory lock and session settings; PgBouncer transaction pooling +is unsupported, and reserving a postgres.js client does not pin a backend through it): + +```sh +# Retire documents and delete their chunks, resuming the saved cursor. Safe to stop and rerun. +MIGRATION_DATABASE_URL= bun run packages/db/script-migrations/0027_retire_search_embeddings.ts + +# Off-peak: finish any remaining retirement, rebuild the HNSW indexes, vacuum, and record completion. +MIGRATION_DATABASE_URL= bun run packages/db/script-migrations/0027_retire_search_embeddings.ts --maintenance +``` + +| Flag | Default | Effect | +| --- | --- | --- | +| `--pause-ratio N` | `2` | After each page, pause N × the page's duration (at most one minute), so the run is busy at most `1 / (1 + N)` of the time. Raise it to go gentler. | +| `--max-rows N` | `2000` | The most rows one page may update or delete (25–8,000). Lower it to make each page lighter. | +| `--maintenance` | off | After retirement, run the index rebuilds and vacuums and journal `0029` with its superseded names. | + +Run it as the migration role: maintenance needs `pg_maintain`, which the application roles lack. Run +it outside peak traffic, and run `--maintenance` in the quietest window you have: concurrent HNSW +rebuilds are long and write a lot of WAL (GitLab, for example, schedules automatic reindexing for +weekends). Keep one run at a time. + +**Pausing.** Ctrl-C is safe at any point. The in-flight page rolls back with its cursor, and an +interrupted concurrent rebuild's leftover index is removed on the next run. Rerun the same command to +resume; completed pages stay committed. + +**Watching.** Every ten pages the run logs the phase, cursor, rows mutated so far and current row +limit; it also logs each halving after a slow page, each phase change, and the start of the completion +recheck. In PostgreSQL, watch for `checkpoint starting: wal` in quick succession, slow checkpoint +sync times, `canceling wait for synchronous replication`, and replica lag. If they appear, stop the +run and resume later with a higher `--pause-ratio` or lower `--max-rows`. + +```sql +SELECT * FROM search_embedding_cleanup_progress; +SELECT name, applied_at FROM script_migrations +WHERE name IN ('0027_retire_search_embeddings', '0028_maintain_search_retirement', + '0029_retire_all_search_embeddings'); +``` + +A plain run does not journal anything; only a `--maintenance` run that finishes records `0029` and its +superseded names. On upgrading a legacy single-KB checkpoint, the snapshot and cursor reset commit +atomically. The scan starts at the beginning once so it includes other KBs behind the old cursor; +previous deletes remain committed. Maintenance checkpoints also reset once because the expanded +cleanup creates new dead entries. A completed legacy checkpoint does not require its former KB to +still exist or remain Search-marked; the new snapshot selects current Search KBs and preserves any KB +now marked ordinary. An unfinished legacy checkpoint still requires its target to remain +Search-marked. + +## How a run paces itself Each page mutates at most a row limit of target rows and reads at most four IDs per row of that limit, never more than 25,000 IDs. Pages execute -sequentially, and each is followed by a pause as long as the page took, up to five seconds, to -leave the primary headroom. Retiring a document is a non-HOT update that writes every index on +sequentially, and each is followed by a pause of `--pause-ratio` times its duration, up to one +minute. Because a page is timed through its commit, a slow synchronous replica or a checkpoint stall +lengthens the following pause by the same factor. Retiring a document is a non-HOT update that writes every index on `document`, and deleting a chunk cascades into its projections, so a page's cost follows the target rows it mutates, not the IDs it reads. A page that reaches the row limit advances the cursor only to its last mutated row; the rest of its scan is read again by the next page. Tying the scan window to the limit keeps that re-reading proportional to the work, even after the limit shrinks. Documents that are already retired never count against the limit. -The row limit starts at 2,000 rows. A page is timed from the start of its transaction through its +The row limit starts at 2,000 rows, or `--max-rows` if lower. A page is timed from the start of its transaction through its commit, including the synchronous-replication wait and any lock-timeout retries. A page slower than -30 seconds halves the limit. A fast page, one under 7.5 seconds, doubles it up to 8,000, which +30 seconds halves the limit. A fast page, one under 7.5 seconds, doubles it up to `--max-rows`, which also widens the scan window, so sparse stretches are not crawled in small windows. The limit never drops below 25 rows. Phase changes do not adjust it. Materialized SQL pages keep the IDs inside PostgreSQL; the migration process receives only a cursor @@ -55,23 +99,6 @@ speed it up. The completion rechecks, which walk every captured KB once, run wit timeout. Brief lock timeouts retry the rolled-back page with bounded backoff for up to one minute. Other errors, or exhausted lock retries, fail the migration without a completion receipt. -Every ten pages the migration logs the phase, cursor, rows mutated so far and current row limit. It -also logs each halving after a slow page, each phase change and the start of the completion -recheck. The deployment job retains its five-hour overall timeout; it is not a runtime estimate. A -large cleanup can need more than one job run, and each run resumes from the saved cursor. - -After interruption or failure, rerun the migration job, or run -`bun run packages/db/script-migrations/0027_retire_search_embeddings.ts` with the writer supplied -through the normal `MIGRATION_DATABASE_URL`/`DATABASE_URL` configuration. Completed pages remain -committed and the failed page is retried from its saved cursor. The standalone command runs both -retirement and maintenance through the successor migration and the same journal. Keep one maintenance worker and monitor primary -latency, WAL, replica lag and available disk. - -Both entry points require a direct or session-pooled PostgreSQL connection, as the deployment -migration runner already does for its session advisory lock and settings. `DATABASE_URL` is a valid -fallback only when it provides that session affinity. PgBouncer transaction pooling is unsupported; -reserving a postgres.js client connection does not pin a backend through a transaction pooler. - The runner-owned `search_embedding_cleanup_targets` table stores the frozen KB set, populated in bounded SQL pages within one repeatable-read transaction. The existing `search_embedding_cleanup_progress` row stores the shared phase and ID cursor; its legacy `knowledge_base_id` remains an informational @@ -110,12 +137,11 @@ indexed Search requires deliberately restoring document eligibility and fully re ## Storage maintenance -After deletion, `0029` invokes the existing maintenance implementation to run `REINDEX INDEX CONCURRENTLY` on each HNSW index of `embedding_search`, +With `--maintenance`, after deletion, `0029` invokes the existing maintenance implementation to run `REINDEX INDEX CONCURRENTLY` on each HNSW index of `embedding_search`, then `VACUUM (ANALYZE, TRUNCATE FALSE)` on the vector and keyword projections, chunk provenance, embeddings, and documents. These operations execute sequentially outside transactions. Rebuilds keep ordinary reads and writes available and require temporary index space and WAL capacity. -They wait for older transactions and can dominate total runtime; the five-hour job deadline still -applies. PostgreSQL's `pg_stat_progress_create_index` and `pg_stat_progress_vacuum` expose progress. +They wait for older transactions and can dominate total runtime. PostgreSQL's `pg_stat_progress_create_index` and `pg_stat_progress_vacuum` expose progress. The existing progress row gains `reindexed_through` and `vacuumed_tables` checkpoints. Completed indexes and tables are skipped on retry; interruption between an operation and its checkpoint may @@ -134,3 +160,20 @@ bucket objects: their application hard-delete path also enqueues identity-bound and applies accounting. Its ordinary scoped mode excludes retired documents, so a follow-up must explicitly support these rows while preserving those side effects. Do not delete source accounts, integration policies or permission grants used by live Search. + +## Why it runs this way + +Long data changes belong outside deploy migrations, in batches, throttled on database health, and +resumable from a cursor: + +- [strong_migrations: Backfilling data](https://github.com/ankane/strong_migrations#backfilling-data) +- [GitLab batched background migrations](https://docs.gitlab.com/development/database/batched_background_migrations/) + and [automatic reindexing](https://docs.gitlab.com/omnibus/settings/database/) +- [Shopify maintenance_tasks](https://github.com/Shopify/maintenance_tasks) +- [gh-ost throttling](https://github.com/github/gh-ost/blob/master/doc/throttle.md) and + [pt-online-schema-change](https://docs.percona.com/percona-toolkit/pt-online-schema-change.html) +- [Stripe: online migrations at scale](https://stripe.com/blog/online-migrations) +- PostgreSQL 17: [WAL configuration](https://www.postgresql.org/docs/17/wal-configuration.html), + [synchronous replication](https://www.postgresql.org/docs/17/warm-standby.html#SYNCHRONOUS-REPLICATION), + [replication statistics](https://www.postgresql.org/docs/17/monitoring-stats.html) +- [PlanetScale: the only scalable delete](https://planetscale.com/blog/the-only-scalable-delete)