home / mcp / mcp llms-txt documentation server

MCP LLMS-TXT Documentation Server

Provides a local MCP server to fetch and audit documentation sources from llms.txt files for LM tooling.

Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
  "mcpServers": {
    "ankitrautela333-langchain-mcp": {
      "command": "uvx",
      "args": [
        "--from",
        "mcpdoc",
        "mcpdoc",
        "--urls",
        "LangGraph:https://langchain-ai.github.io/langgraph/llms.txt LangChain:https://python.langchain.com/llms.txt",
        "--transport",
        "stdio"
      ]
    }
  }
}

You can run a focused MCP server that lets you select llms.txt sources and expose a simple fetch_docs tool to audit and retrieve documentation URLs. This gives you full control over which sources are queried and how tool calls are performed, making it easier to audit and reason about the context returned by documentation lookups for LM tooling.

How to use

Start by launching the local MCP server using the standard MCP command. You will then connect your MCP client (such as Cursor, Windsurf, or Claude Code/Desktop) to the running server and begin querying documentation sources defined by your llms.txt URLs.

How to install

Prerequisites: you need Node.js or a runtime that supports the MCP tooling, plus the uv tool for running MCP servers.

1. Install uv if you do not have it already.

curl -LsSf https://astral.sh/uv/install.sh | sh

Additional sections

Run the MCP server locally using a command that specifies your documentation sources. The server is started with the from mcpdoc context and uses the LangGraph and LangChain llms.txt files as sources. You can also specify how the server communicates with clients.

uvx --from mcpdoc mcpdoc \
    --urls "LangGraph:https://langchain-ai.github.io/langgraph/llms.txt" "LangChain:https://python.langchain.com/llms.txt" \
    --transport sse \
    --port 8082 \
    --host localhost

Open the MCP inspector in your environment to test and verify tool calls.

Configuration and client setup

To connect a client like Cursor, Windsurf, or Claude Desktop, add a local MCP server entry that uses the same mcpdoc configuration and sources.

Available tools

list_doc_sources

Tool that lists the available llms.txt sources configured for the MCP server.

fetch_docs

Tool that reads and fetches documentation from URLs listed in the specified llms.txt sources.