ai-doc-creator 2.3.1

AI-powered documentation generator for repos and local projects — MCP server + CLI, any LLM provider or keyless via MCP host sampling.

AI Document Creator

Ask your AI assistant to document any codebase — and get a full set of docs with architecture diagrams, delivered as files or a pull request.

You say:

"Document the repo https://github.com/user/repo and open a PR with the docs."

You get:

• 📄 A markdown doc for every file — what it does, key functions, how to use it
• 📖 A polished README for the whole project
• 📊 Architecture diagrams (Mermaid) — a project map and a module-dependency chart, generated from real imports so they're always accurate
• 🔀 Delivered where you want: local folder, inline in chat, or a GitHub pull request

Works inside Claude Code, Claude Desktop, Cursor, Codex, Antigravity — any app that supports MCP — plus a plain CLI and a GitHub Action.

Get started in 1 minute

In Claude Code

claude mcp add ai-doc-creator -- uvx --from ai-doc-creator ai-doc-creator-mcp

Then just ask: "Document this project." That's it.

In Claude Desktop, Cursor, or any other MCP app

Add this to the app's MCP settings (Cursor: ~/.cursor/mcp.json):

{
  "mcpServers": {
    "ai-doc-creator": {
      "command": "uvx",
      "args": ["--from", "ai-doc-creator", "ai-doc-creator-mcp"]
    }
  }
}

You need uv and git installed — nothing else.

Don't want to install anything?

Point your MCP app at the hosted server instead — see the Usage Guide.

Do I need an API key?

No. Pick whichever fits you:

You have...	What to do
Nothing — just an AI app	Ask: "Use start_doc_job to document <repo>, write the docs yourself batch by batch, then finish the job." Your app's own AI writes the docs — free.
An API key (Anthropic/OpenAI/Azure/Bedrock)	Add it to the config — fastest, everything generated in one call
Ollama on your machine	provider="ollama" — free and fully offline

Pick a documentation style

Add profile to your request (or --profile in the CLI):

• readme (default) — classic project README
• api — API reference with signatures and parameters
• architecture — components, data flow, and diagrams
• tutorial — a guided walkthrough for newcomers

💡 Diagrams render as pictures on GitHub and in IDE markdown previews. A plain terminal shows their Mermaid source — that's normal.

The CLI (no AI app needed)

pip install ai-doc-creator
export ANTHROPIC_API_KEY=sk-ant-...          # any supported provider
ai-doc-creator --repo https://github.com/user/repo --output docs
ai-doc-creator --path . --profile architecture

Docs that update themselves (GitHub Action)

- uses: dharmikraval1/ai-document-creator@v2
  env:
    ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}

Every push regenerates your docs and opens a PR. Full workflow →

Host your own public server

One click. Safe defaults are pre-configured: your API keys are never spent on other people's requests (they bring their own key in a header, or their app's AI writes the docs), rate limiting is on, and auto-deploy tracks main.

Good to know

• Incremental: re-runs only process files that changed (content-hash cache) — fast and cheap.
• Your key stays private: keys travel in config/headers only — never in chat messages, logs, or output.
• Hardened: SSRF guards, request sandboxing, size/time/rate limits. Details in SECURITY.md.

Learn more

• 📖 Usage Guide — step-by-step setup for every app, all options, troubleshooting
• 🤝 Contributing · 🔒 Security policy · 🗂 Design docs

License

MIT — free for any use.