digitaloceanapp

DigitalOcean tools and workflows for your AI assistant.

Connect your assistant to DigitalOcean's official MCP servers to explore your infrastructure, review App Platform deployments, and investigate your cloud costs. digitaloceanapp supplies the client configuration and reusable instructions—called skills—that guide those tasks.

Your assistant connects directly to DigitalOcean, either through its hosted servers or its official MCP package running on your computer. This project runs no backend and receives none of your credentials or account data.

Release candidate: 1.0.0-rc.3. Client configurations are experimental templates with automated format and packaging coverage. Synthetic workflows have been exercised by a separately authorized evaluator, with results and limits recorded. No authenticated client/mode combination is certified yet; live client testing remains pending. The V3 roadmap tracks deployment preflight, recovery reviews, proposed-change costs and easier setup.

Get started

You'll need Node.js 22 or newer, npm, and an MCP-compatible AI client.

git clone https://github.com/DangerMouseUK/digitaloceanapp.git
cd digitaloceanapp
npm ci
node bin/digitaloceanapp.js setup

Choose your client, select Remote OAuth and the Core preset, then pick a new output directory. Setup creates the configuration, skills, and an INSTALL.md with instructions for your client.

Follow that file to install the skills and merge the configuration into your client. Setup doesn't change your existing settings. Once installed, connect the servers and sign in through DigitalOcean when your client prompts you.

Preview the services, skills, installation destinations and authentication steps first:

node bin/digitaloceanapp.js setup --client codex --dry-run

Add --json for a structured preview. Dry runs write no files and do not check the proposed output directory's writability.

Try asking:

What do I have running on DigitalOcean?

Review my App Platform apps. Are there any deployment problems or resources worth checking for oversizing?

Why did my latest deployment fail?

The included skills also cover account audits, recent deployment history, infrastructure relationships, and official documentation research. Review and troubleshooting skills are written to inspect resources without changing them; resource changes require a specific request.

Dedicated database, Kubernetes, Droplet and networking reviews help investigate infrastructure concerns. Select the infrastructure preset for those services. Try “Why can app-a not reach db-a?” or “Review cluster-a and report what you can verify about its node pools.” Tools and evidence available through your client determine coverage.

V3 also supports “Check this app before release,” “Review backups for these databases and Droplets,” and “What would adding another app instance cost?” These reviews identify evidence gaps without deploying, restoring or resizing resources. Findings stay in your AI client; the utility stores configuration and skills only.

Choose a client template

Client	Setup guide
Codex	Configuration and skills
VS Code	Configuration and skills
Cursor	Configuration and skills
Claude Code	Configuration and skills
Claude Desktop	Connectors and local MCP
Windsurf	Configuration and skills
ChatGPT	Remote connections and plugin setup
Other portable plugin hosts	Plugin package

Connection modes and skill installation vary by client. ChatGPT uses remote connections; Claude Desktop configuration is provided for Windows and macOS. See the compatibility notes for the current limits.

Select your services

Start with the services you use. Connecting every server adds tools your assistant may not need.

Preset	Includes
core — default	Accounts, App Platform, Insights, Documentation
app-platform	Core plus Container Registry, Spaces, and Networking
infrastructure	Accounts, Droplets, Databases, Kubernetes, Networking, Volumes, NFS, Insights, and Documentation
full	All 21 services in the reviewed registry

You can also choose individual services. For example, to generate a Cursor configuration for App Platform and documentation:

node bin/digitaloceanapp.js setup --client cursor --services apps,docs --output ./do-config

The output directory must be new, and its parent must already exist. Run node bin/digitaloceanapp.js services for the full list, or browse the service reference.

Authentication

Remote OAuth is the default. Your client opens DigitalOcean's sign-in flow; you don't give this project a token. Public Documentation MCP needs no account credentials.

For local MCP, select --mode local. The generated configuration runs the pinned official DigitalOcean package on your computer. Supply DIGITALOCEAN_API_TOKEN through the client's supported secret input or launching environment.

Remote API-token configurations are also available for Codex, VS Code, Cursor, Claude Code, and Windsurf. Setup only writes references to credentials, never their values. See authentication, remote mode, or local mode for instructions.

Check your setup

node bin/digitaloceanapp.js validate --path ./do-config
node bin/digitaloceanapp.js doctor --path ./do-config

validate checks the generated files. doctor also checks local prerequisites and whether a required token variable is present, without displaying it. Both run offline; confirm the actual connection in your client's MCP settings.

Run node bin/digitaloceanapp.js --help for all options. If something isn't working, start with troubleshooting.

For structured redacted output, add --json to validate or doctor. Reports include stable check codes, pass/fail outcomes and corrective actions. Authentication remains unverified; see the diagnostic format.

Upgrade an existing bundle

Inspect changes, then generate a replacement in a new directory:

node bin/digitaloceanapp.js upgrade --path "../My old bundle" --dry-run
node bin/digitaloceanapp.js upgrade --path "../My old bundle" --output "../My new bundle"

The upgrade preserves your selected services, client, platform and connection mode. Read the generated MIGRATE.md and review flagged differences and customizations before manually installing. Your old bundle and client settings remain intact. See upgrading for accepted versions and review limits.

V3 bundles record fingerprints of generated files so later upgrades can distinguish release changes from local customizations. Earlier bundles still need manual comparison. Add --json to an upgrade dry run for a redacted change list.

Permissions and privacy

DigitalOcean's tools can change and delete infrastructure. The skills guide your assistant's behavior, but they cannot enforce read-only access. Use appropriately scoped DigitalOcean access and your client's tool approval controls.

The project has no telemetry, proxy, or account database. Your AI client and DigitalOcean handle your requests under their own policies. Resource coverage, metrics, and billing detail depend on the tools DigitalOcean exposes; missing data should be reported as missing, not treated as zero usage.

Read the security model, privacy statement, and known limitations. Report vulnerabilities through private security reporting.

Contributing

Improvements to skills, client setup, and service coverage are welcome. See CONTRIBUTING.md for the development workflow. Coding agents should follow the repository's AGENTS.md.

npm ci
npm run check

Configuration and service-documentation changes are generated from the registry with npm run generate. The test suite covers client formats, CLI behavior, credential handling, and package installation. Live client checks are tracked separately in the acceptance record.

For bugs and feature requests, open an issue. Please leave credentials and account data out of reports.

License

MIT. This is an independent community project, not an official DigitalOcean product or an endorsed integration.