# AI & LLM access _Read these docs from an agent_ Every page is also served as Markdown, indexed for LLMs, and exposed by an MCP server. Point your agent at whichever fits: a page's [Markdown twin](/ai.md#markdown), the [llms.txt](/ai.md#llms) index, or the [MCP server](/ai.md#mcp). ## Markdown twins _Any page as clean Markdown_ Append `.md` to a doc URL, or send `Accept: text/markdown` on the page URL. Both return the page as Markdown, with links rewritten to the other `.md` pages. #### EXAMPLE ```bash # the .md twin of any page curl https://medius.k4tech.net/library/clip.md # or content negotiation on the page URL curl -H "Accept: text/markdown" https://medius.k4tech.net/library/clip ``` ## llms.txt _The index and the full corpus_ [/llms.txt](https://medius.k4tech.net/llms.txt) lists every page as a Markdown link, following the [llms.txt](https://llmstxt.org/) convention; [/llms-full.txt](https://medius.k4tech.net/llms-full.txt) is the whole documentation in one file. ## MCP server _Search and fetch the docs from a coding agent_ A read-only [MCP](https://modelcontextprotocol.io) server at `https://medius.k4tech.net/mcp` (Streamable HTTP, no auth). It exposes `search` and `fetch` for ChatGPT deep research, plus `search_docs`, `get_page`, and `list_pages` for other clients. #### CLAUDE CODE ```bash claude mcp add --transport http medius-docs https://medius.k4tech.net/mcp ``` #### CURSOR / WINDSURF / VS CODE ```json { "mcpServers": { "medius-docs": { "url": "https://medius.k4tech.net/mcp" } } } ``` #### CLAUDE DESKTOP / CLAUDE.AI / CHATGPT Add a custom connector with the URL `https://medius.k4tech.net/mcp`.