LangGraph

You run a focused MCP server that lets you define and audit the llms.txt files you want to use for context retrieval. This server exposes a simple, auditable toolset so applications can fetch documentation URLs from your chosen sources and integrate those results into conversations or analyses.

How to use

You connect MCP tools to your client apps to access llms.txt sources and fetch documentation content. Start the MCP server, then configure your client (Cursor, Windsurf, Claude Code/Desktop, or Claude Code) to use the server. You will be able to list available document sources, fetch docs from those sources, and base responses on the retrieved content.

How to install

Prerequisites you need before installation: a working Node.js and npm, or the uv runtime for running MCP tooling.

Install the uv runtime using the official script:

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

Choose your llms.txt source. For example, select the LangGraph llms.txt file.

Run the MCP server locally to test with your llms.txt URL. This starts a server that you can connect to from MCP clients.

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

Open the inspector to verify the server is responding.

npx @modelcontextprotocol/inspector

Configure your client to connect to the running MCP server. Below is an example for the Cursor client.

{
  "mcpServers": {
    "langgraph_docs_mcp": {
      "command": "uvx",
      "args": [
        "--from",
        "mcpdoc",
        "mcpdoc",
        "--urls",
        "LangGraph:https://langchain-ai.github.io/langgraph/llms.txt",
        "--transport",
        "stdio",
        "--port",
        "8081",
        "--host",
        "localhost"
      ]
    }
  }
}

Additional setup to connect clients

Cursor, Windsurf, Claude Desktop, and Claude Code can be configured to connect to your MCP server. Each client has its own steps to register MCP servers and to enable tool usage in chats. After adding the server, open the chat interface and select the MCP agent to start using the tools.

Available tools

list_doc_sources

Lists the available llms.txt sources that your MCP server can reference, enabling you to audit which document sources are in use.

fetch_docs

Fetches the content from the URLs listed in a provided llms.txt source, returning the processed documentation content for use in downstream tasks.