This project provides an open-source Model Context Protocol (MCP) server to expose llms.txt documentation files to IDEs and AI applications, enabling auditable context retrieval for developers. It addresses the opacity of existing tools by offering fine-grained control over data sources and tool calls.

How It Works

The server acts as a bridge, allowing applications like Cursor, Windsurf, and Claude to query specified llms.txt files. It implements a fetch_docs tool that retrieves content from URLs within these files. A key design choice is strict domain access control, where only explicitly allowed domains can be queried, enhancing security and preventing unauthorized data access.

Quick Start & Requirements

• Install uv: curl -LsSf https://astral.sh/uv/install.sh | sh
• Run server: uvx --from mcpdoc mcpdoc --urls "LangGraph:https://langchain-ai.github.io/langgraph/llms.txt" --port 8082
• Connect inspector: npx @modelcontextprotocol/inspector
• Official Docs: https://github.com/langchain-ai/mcpdoc

Highlighted Details

• Supports multiple IDE integrations (Cursor, Windsurf, Claude Code/Desktop).
• Allows configuration via CLI, YAML, JSON, or programmatic Python usage.
• Implements domain access control for security.
• Uses Server-Sent Events (SSE) or stdio for transport.

Maintenance & Community

This project is part of the LangChain AI ecosystem. Further community and roadmap details are available via the LangChain community channels.

Licensing & Compatibility

The project is licensed under the MIT License, permitting commercial use and integration with closed-source applications.

Limitations & Caveats

The README notes that as of March 21, 2025, Claude Desktop and Claude Code may not support global rules, requiring manual prompt appending for rule enforcement. Python version compatibility issues might arise with certain integrations, necessitating explicit Python executable path specification.