mcpdoc

What does this tool actually do?

This tool helps AI coding assistants like Cursor or Claude Code read documentation from llms.txt files. It gives developers full control over which documentation sources are used and allows auditing of every tool call. You can point it to any llms.txt URL or local file, and it will fetch the linked pages securely.

Deep dive

mcpdoc is an open-source MCP (Model Context Protocol) server that bridges the gap between AI-powered IDEs and structured documentation. It allows developers to define a list of llms.txt files—standardized website indexes for LLMs—and provides a simple `fetch_docs` tool that reads URLs within those files. This enables MCP host applications like Cursor, Windsurf, and Claude Code/Desktop to retrieve context for tasks with full transparency and auditability. The server implements strict domain access controls for security: remote llms.txt files automatically restrict fetching to their own domain, while local files require explicit domain whitelisting. It supports both SSE and stdio transports, making it easy to integrate into various workflows. Built by LangChain, it is ideal for teams that want to ensure their AI assistants use only approved documentation sources.

Additional use cases

• Enable context-aware code generation by feeding relevant docs to LLMs.
• Standardize documentation retrieval across multiple AI tools in a team.

Who should use it

Tags