# Docs in your AI tools (/dev/ai-tools) The docs have three parts, each for one audience, and each part has its own MCP server and llms files. Add only the one you need: the product guide knows nothing about the code, the developer docs nothing about the API's parameters. | Part | For | MCP server (Streamable HTTP) | llms | | --------------------------- | ---------------------------------------- | ------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------------------------------------- | | [Product guide](/docs) | people using the app | `https://remy-auth-docs.gedw99.workers.dev/api/mcp/docs` (`remy-product-guide`) | [llms.txt](https://remy-auth-docs.gedw99.workers.dev/docs/llms.txt), [full](https://remy-auth-docs.gedw99.workers.dev/docs/llms-full.txt) | | [Developer docs](/dev) | developers building Remy or an app on it | `https://remy-auth-docs.gedw99.workers.dev/api/mcp/dev` (`remy-developer-docs`) | [llms.txt](https://remy-auth-docs.gedw99.workers.dev/dev/llms.txt), [full](https://remy-auth-docs.gedw99.workers.dev/dev/llms-full.txt) | | [API reference](/reference) | developers calling the API | `https://remy-auth-docs.gedw99.workers.dev/api/mcp/reference` (`remy-api-reference`) | [llms.txt](https://remy-auth-docs.gedw99.workers.dev/reference/llms.txt), [full](https://remy-auth-docs.gedw99.workers.dev/reference/llms-full.txt) | [`/llms.txt`](https://remy-auth-docs.gedw99.workers.dev/llms.txt) lists all three. No sign-in: they are public and read-only. ## The tools [#the-tools] Every server has the same four, all read-only: * `search` and `fetch`: the pair ChatGPT needs for chat, deep research and company knowledge ([OpenAI's schema](https://developers.openai.com/api/docs/mcp)). `search` returns `{results: [{id, title, url}]}`; `fetch(id)` returns `{id, title, text, url}`, the page as Markdown. * `list_pages` and `get_page`: Fumadocs' own (`fumadocs-core/mcp`). Each server's instructions say who it is for, so an assistant with all three picks the right one. ## Adding one [#adding-one] | Tool | How | | --------------------- | -------------------------------------------------------------------------------------------------------------------------------------------- | | ChatGPT | Settings → Security and login → Developer mode on; then Plugins → **+** → paste the URL. | | Claude (web, desktop) | Settings → Connectors → Add custom connector → paste the URL. | | Claude Code | `claude mcp add --transport http remy-developer-docs https://remy-auth-docs.gedw99.workers.dev/api/mcp/dev` | | Cursor | `.cursor/mcp.json`: `{"mcpServers": {"remy-developer-docs": {"url": "https://remy-auth-docs.gedw99.workers.dev/api/mcp/dev"}}}` | | VS Code | `.vscode/mcp.json`: `{"servers": {"remy-developer-docs": {"type": "http", "url": "https://remy-auth-docs.gedw99.workers.dev/api/mcp/dev"}}}` | Swap in another row's URL and name for the product guide or the API reference. `npm run docs -- test remote` lists every server's tools. ## Gemini and Google [#gemini-and-google] Google's AI answers do not use MCP servers or llms.txt: Google says so ([Search Central](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide)). | Where | How it sees these docs | | ---------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Google AI Mode, AI Overviews | Google's search index: the HTML pages (`/sitemap.xml`), their titles, descriptions and text. A page must be indexed with a snippet, and the site included in "Search generative AI features" in Search Console. | | The Gemini app | The same Google Search grounding; it takes no MCP server of ours. | | Gemini CLI | An MCP server: `gemini mcp add --transport http remy-developer-docs https://remy-auth-docs.gedw99.workers.dev/api/mcp/dev` | | Gemini API (managed agents) | An MCP server: an `mcp_server` tool with the URL. | | Gemini Enterprise | An MCP server: an administrator adds it as a custom MCP connector (Streamable HTTP). | So for Google the pages themselves have to say who they are for, which is why every part is written for its audience ([Writing docs](./writing-docs.mdx#who-each-part-is-for)): each page's title ends with its part ("… | Remy product guide", "… | Remy developer docs", "… | Remy API reference"), and each part's home page opens by saying who it is for. ## Finding them [#finding-them] Nothing finds an MCP server by searching the web: an assistant needs its URL. These pages, the [landing page](/), `/llms.txt` and each part's sidebar ("For AI tools") give it. Listing them in the official [MCP Registry](https://registry.modelcontextprotocol.io) (`io.github.joeblew999/...`) is the owner's decision; it is not done. ## llms.txt and Markdown [#llmstxt-and-markdown] * `//llms.txt`: every page with its description; Spanish after the part's base (`/dev/es/llms.txt`). * `//llms-full.txt`: every page's text in one file. * Any page as Markdown: add `.md` to its address (`/dev/tooling.md`, `/reference/status.md`). ## Search engines [#search-engines] Google and other search engines read the pages themselves: each language has its own address, linked to the others with `hreflang`, and all of them are in `/sitemap.xml`.