A concise, reviewed starting point for building software with AI. The core tools below had a usable current offering and supporting first-party evidence on 2026-09-18. Known sunsets, archived implementations, and unresolved entries are kept out of this shortlist.
| I want to… | Start here |
|---|---|
| See what changed | Recent verified updates |
| Build or improve software | Find tools by task |
| Learn from a first project | Start here · Learn & practice |
Code · App & UI · Context & specs · Test & review · Deploy & run · Workspaces & usage · Community
Recent Verified Updates
Recent reviews and material changes from the last 30 days, newest first. A review date is not a product release date. Lifecycle review and evidence record what was checked.
| Verified on | Update | What changed |
|---|---|---|
| 2026-09-20 | Delta (Zed) | Reviewed addition: multiplayer agent-coding workspace with thread-based review; public beta, free during beta. |
| 2026-09-18 | Shep | Reviewed addition: local-first orchestrator that runs parallel coding agents in isolated Git worktrees through to draft PRs. |
| 2026-09-18 | Lifecycle cleanup | Removed unavailable/new-user-blocked entries, corrected canonical products, and reduced the core tool list from 156 to 40. |
| 2026-09-18 | Pi | Reviewed addition: extensible MIT terminal coding agent with multi-provider models and an SDK. |
| 2026-09-18 | NextReset | Reviewed addition: unofficial public reset history and local countdowns; forecasts are not endorsed. |
| 2026-09-18 | Superagent | Reviewed addition: macOS coding-agent workspace with browser and iOS workflows. |
| 2026-09-18 | Publish.my | Reviewed addition: agent-oriented static-site publishing; email activation required. |
| 2026-09-18 | Agent QA | Reviewed addition: web/mobile test workflows; FSL-1.1-ALv2 is identified. |
| 2026-09-18 | Duckweed | Reviewed addition: local coding-agent terminal workspace; source-available license. |
| 2026-09-18 | Vibe Coding with Confidence (Mahmoud Zalt) | Access clarified: free core chapters and paid advanced chapters. |
New to development: Choose one small project, learn how to save changes with Git, and test one feature at a time. You should be able to open a project folder and run the commands in your chosen tutorial; use its setup chapter if those steps are new.
- Read a first-project guide and choose a result you can explain, such as a local task list.
- Choose one coding assistant or app prototyping tool. Read the access conditions first: a free client can still require a paid model, subscription, or API key.
- Write a short goal and acceptance checks. Build one small step, inspect the changes, run tests, then save a Git checkpoint.
- Use the testing and review tools before sharing. Keep credentials outside prompts and commits, and understand the generated code you rely on.
- Deploy only after checking the hosting/runtime constraints, especially whether the result needs a backend.
Already a developer: Jump to a task above, then use the practical workflows for planning, refactoring, debugging, and validation.
The main page keeps a small set of distinct starting points for each task; it is not a popularity ranking or an exhaustive list. More verified options remain available. Each resource has one primary home. CLI, IDE, Web, Desktop, and MCP describe how you use a tool; they are not quality rankings. Search this page for MCP to find protocol integrations across tasks.
Read the metadata: each verification date links to the lifecycle evidence. It confirms the described offering and access scope as of that date, not a security audit or a promise of future availability. Check pricing means the current price was not comprehensively reviewed. Open-source client licenses do not make model inference free. New-user restrictions and paid chapters are stated explicitly.
Understand a repository, implement a feature, or refactor existing code.
| Resource | When to use it | Access / scope | Verified |
|---|---|---|---|
| Aider · CLI | Terminal pair programming with cloud or local models and Git integration | Open-source client; API or local-model costs | 2026-09-18 |
| Claude Code · CLI | Agentic coding assistant with terminal and project workflows | Check pricing | 2026-09-18 |
| Cline | Coding agent for IDE, terminal, and desktop workflows with file, command, and browser tools | Client plus model provider; check pricing | 2026-09-18 |
| Cursor · IDE | Coding editor and agent for implementation, debugging, and code review | Check pricing | 2026-09-18 |
| Gemini CLI · CLI | Open-source terminal coding agent with API-key, Vertex, or eligible enterprise access; consumer subscription access retired | API/Vertex or eligible enterprise; no retired consumer route | 2026-09-18 |
| GitHub Copilot | Coding assistance and agent workflows across GitHub and supported IDEs | Check pricing | 2026-09-18 |
| Goose · CLI | Extensible local agent with coding workflows, desktop and CLI interfaces, and MCP tools | Check pricing | 2026-09-18 |
| Kiro · IDE | AWS coding agent with IDE and CLI workflows, specifications, and tests | Check pricing | 2026-09-18 |
| OpenAI Codex CLI · CLI | OpenAI coding agent that runs locally in the terminal | Check pricing | 2026-09-18 |
| Pi · CLI | Extensible terminal coding agent and embeddable harness | MIT client; model/provider terms separate | 2026-09-18 |
Create a first application or interface from a description or design.
| Resource | When to use it | Access / scope | Verified |
|---|---|---|---|
| Bolt.new · Web | Natural language app building by StackBlitz | Check pricing | 2026-09-18 |
| Dyad · Desktop | Build applications locally with configurable model providers | Check pricing; local desktop app and model-provider costs. | 2026-09-18 |
| Lovable · Web | Full-stack app generation with Supabase | Check pricing | 2026-09-18 |
| Onlook · Web | Edit application interfaces visually while working with code | Check pricing | 2026-09-18 |
| Replit · Web | Build and iterate on applications with Replit Agent | Check pricing | 2026-09-18 |
| v0 · Web | Vercel's AI for UI/React generation | Check pricing | 2026-09-18 |
Give agents requirements, rules, documentation, and connected project data.
| Resource | When to use it | Access / scope | Verified |
|---|---|---|---|
| Caliber | CLI that generates and syncs AI agent configs for Claude Code, Cursor, and Codex | Check pricing | 2026-09-18 |
| Context7 · MCP | Retrieve library documentation for coding context | Check pricing | 2026-09-18 |
| Filesystem MCP · MCP | Give agents controlled access to project files | Check pricing; reference implementation, not a production-ready guarantee | 2026-09-18 |
| GitHub MCP · MCP | Connect repository, issue, and pull-request workflows | Check pricing | 2026-09-18 |
| Neon · MCP | Connect development workflows to Neon databases | Check pricing | 2026-09-18 |
| Notion MCP · MCP | Official hosted MCP for searching, reading and updating Notion content | Check pricing | 2026-09-18 |
| OpenSpec | Spec-driven development framework for AI coding assistants | Check pricing | 2026-09-18 |
| Supabase MCP · MCP | Official MCP for Supabase schemas, queries and project configuration | Check pricing | 2026-09-18 |
Check behavior, review generated changes, and diagnose failures.
| Resource | When to use it | Access / scope | Verified |
|---|---|---|---|
| Agent QA · MCP | Author and run natural-language web and mobile tests | FSL-1.1-ALv2; source-available with competing-use restriction; model costs | 2026-09-18 |
| Playwright MCP (Microsoft) · MCP | Official browser automation through structured page snapshots | Apache-2.0 client; model terms separate | 2026-09-18 |
| PR-Agent | Community-maintained pull request reviewer, separate from Qodo | Check pricing | 2026-09-18 |
| Qodo | AI code review engine (formerly CodiumAI) | Check pricing | 2026-09-18 |
| Semgrep MCP · MCP | MCP security scanning through the maintained Semgrep CLI | Check pricing | 2026-09-18 |
| Sentry · MCP | Inspect application errors and diagnose failures | Functional Source License; source-available; check hosted terms | 2026-09-18 |
Build, publish, or run code in an appropriate execution environment.
| Resource | When to use it | Access / scope | Verified |
|---|---|---|---|
| Cloudflare · MCP | Manage application deployment and cloud resources | Check pricing | 2026-09-18 |
| E2B | Secure cloud sandbox environment for enterprise-grade AI agents | Check pricing | 2026-09-18 |
| Publish.my · Web | Agent-driven static-site publishing with email activation | Free tier; static sites only; email activation | 2026-09-18 |
| XcodeBuildMCP · MCP | CLI and MCP tools for building, running and debugging Apple-platform projects | Check pricing | 2026-09-18 |
Organize sessions, inspect runs, and understand usage or availability.
| Resource | When to use it | Access / scope | Verified |
|---|---|---|---|
| Delta (Zed) · Desktop | Run agent threads and review their changes in a multiplayer workspace that keeps conversation and code history connected; works with existing Git repositories | Free during public beta; paid plans planned; model costs separate | 2026-09-20 |
| Duckweed · Desktop | Run coding agents, shells, Git diffs and sessions in a local cross-platform workspace | Source-available; see license; model/provider costs separate | 2026-09-18 |
| llm.log | Inspect model costs, tokens and request traces through a local proxy | MIT; model/provider costs separate | 2026-09-18 |
| Parallel Code · Desktop | Run coding agents in isolated Git worktrees and review their changes | MIT; model/provider costs separate | 2026-09-18 |
| Shep · CLI | Orchestrate parallel coding agents in isolated Git worktrees through commit, push, CI watch and draft PRs | MIT client; agent subscription or API costs separate | 2026-09-18 |
| Superagent · Desktop | Use Claude Code or Codex in a macOS workspace with browser and iOS tools | MIT; macOS Apple Silicon; model subscription separate | 2026-09-18 |
| usage · Desktop | View coding-agent quotas from the macOS menu bar or Windows system tray | AGPL-3.0; model/provider costs separate | 2026-09-18 |
| Warp Terminal | Use an agent-oriented terminal and inspect coding workflows | Terminal download; check AI usage pricing | 2026-09-18 |
Choose by what you want to achieve. Guide and paper labels describe the format. Dated conceptual material can remain useful; its old tool examples and benchmark rankings are not current product advice. Check language, prerequisites, and paid sections before starting. Historical videos and additional reading are in the extended catalog.
Follow a setup-to-result sequence. Use the material's own prerequisites and finish a small working application before collecting more tools.
| Resource | When to use it | Access / scope | Verified |
|---|---|---|---|
| AI Book: AI Coding · Guide | Chinese handbook on coding-agent workflows, tool selection, and practices with Cursor, Codex, Claude Code, and Kiro | Free reading; check reuse terms | 2026-09-18 |
| Vibe Coding Manual (Roboco) · Guide | Korean workflow and project-rule templates; dated examples | Free reading; check reuse terms | 2026-09-18 |
| Vibe Coding with Confidence (Mahmoud Zalt) · Guide | Development handbook with readable introductory chapters and paid advanced material | Free core chapters; paid advanced chapters; EN | 2026-09-18 |
Use these when a project already exists and you need to solve a specific problem.
| Task | Working sequence |
|---|---|
| New feature | Goal and acceptance checks → inspect context → small implementation → review and test |
| Refactoring | Capture current behavior → identify a small change → compare behavior → repeat |
| Bug fixing | Reproduce → form a hypothesis → add a regression check → fix and verify |
| Testing | Identify critical behaviors → choose meaningful checks → run and inspect failures |
Full workflows and prompt templates include session setup and reusable playbooks. Keep requirements and decisions in project documents, use sandboxes where appropriate, and review security-sensitive changes before deployment.
| Resource | When to use it | Access / scope | Verified |
|---|---|---|---|
| Agentic Coding Recommendations (Armin Ronacher) · Guide | Practical agent workflows and testing advice; 2025 practitioner perspective | Free reading; check reuse terms | 2026-09-18 |
| Here's how I use LLMs to help me write code (Simon Willison) · Guide | Iterative coding and QA practices; 2025 tool examples are historical | Free reading; check reuse terms | 2026-09-18 |
| Secure Vibe Coding Guide (CSA) · Guide | Security checklist covering secrets, authorization, validation and review | Free reading; check reuse terms | 2026-09-18 |
Understand the tools after trying a small project, or go deeper into evaluation and development practices.
Vibe coding uses natural-language intent to guide AI-generated software. The Model Context Protocol connects agents to tools and data. A coding agent, its underlying model, and its integrations are different choices: changing one does not automatically change the others.
| Resource | When to use it | Access / scope | Verified |
|---|---|---|---|
| Context Engineering Intro (coleam00) · Guide | Project-context and instruction examples for coding agents | Free reading; check reuse terms | 2026-09-18 |
| Model Context Protocol documentation · Guide | Current official introduction to MCP architecture and integrations | Free reading; check reuse terms | 2026-09-18 |
| SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering · Paper | Autonomous agent fixing real bugs using Agent-Computer Interface | Free reading; check reuse terms | 2026-09-18 |
| SWE-bench: Can Language Models Resolve Real-World GitHub Issues? · Paper | Standard benchmark for evaluating AI coding agents | Free reading; check reuse terms | 2026-09-18 |
Background and origin
"Fully give in to the vibes, embrace exponentials, and forget that the code even exists." — Andrej Karpathy, February 2025
For learning and production work, pair natural-language instructions with understanding, review, tests, and clear responsibility for the result.
- Additional verified resources — current alternatives and dated learning material beyond the core shortlist.
- Lifecycle evidence and review decisions — sources, scope limitations, and current canonical names.
- Removed, replaced, and unverified entries — history and reasons, not current recommendations.
- For product support, use the official documentation linked in each resource's evidence record. Open a repository issue to report a status change or correction.
Suggest a resource through an issue. Every addition must pass direct relevance, usable public evidence, distinct value, transparent access/claims, and maintenance or substantive completeness. Paid products and self-submissions follow the same rules; disclose affiliations and material limitations. GitHub stars do not guarantee admission.
Read the curation policy and contribution guide. Clear failures are rejected with reasons; uncertain cases remain open for review. Accepted changes are synchronized in English, Korean, and Japanese and issues are closed only after publishing succeeds.
The weekly and issue workflows use Pi with Kimi or Qwen and Exa Search. Automation and configuration explains the implementation and maintainer controls.
This work is dedicated to the public domain under CC0 1.0. Linked projects and learning materials retain their own licenses and access conditions.