automadocs / mcp / index.md
00
MCP
Let your coding agent read your docs
Your agent searches, reads and cites the same living documentation your team reads, instead of guessing from stale files.
- Cursor
- Claude Code
- VS Code
- Codex
- Zed
01
Install
Add AutomaDocs to your agent
First set your key in your shell: export AUTOMADOCS_API_KEY="YOUR_API_KEY"
or add manually: ~/.cursor/mcp.json
Remote endpoint: https://api.automadocs.com/api/mcp/rpc. It accepts your key as Authorization: Bearer or X-API-Key. Keys stay in your own agent config; this page never sees them.
02
Tools
What your agent can do
- 01
list_repos
See which repositories your key can read, with sync status and page counts. Every other tool starts from a repo_id found here. - 02
search_docs
Find where something is implemented or documented by meaning, not just keywords. Each hit carries a citation such as src/auth.ts:10-24. - 03
read_doc
Read one generated documentation page in full, by id or by source file path, when a snippet is not enough context. - 04
get_coverage
See the coverage score and which functions and endpoints still have no documentation. - 05
ask_question
Ask a question in plain language and get an answer grounded in retrieved code, with its sources listed. Counts against your AI usage.
03
Setup
From key to first answer
- 01
Create an API key
Make a key in API settings. Read access is enough for every tool except ask_question, which also needs chat. - 02
Add the server to your agent
Pick your agent above, copy the snippet, and set AUTOMADOCS_API_KEY in your shell. Cursor and VS Code also have a one-click install button. - 03
Ask about your code
Try “search the docs for how webhooks are verified” and the agent calls the tools itself.