Documentation MCP Server 📚🔍
@sagacious-satadru
An MCP server that can work with Claude desktop to fetch documentation from langchain, llama-index, and OpenAI.
Overview
What is Documentation MCP Server?
Documentation MCP Server is a Model Context Protocol (MCP) server that enables Claude to search and access documentation from popular libraries like LangChain, LlamaIndex, and OpenAI directly within conversations. It uses the Serper API for Google search integration and BeautifulSoup to extract relevant text from documentation pages.
How to use Documentation MCP Server?
Install the uv package manager, clone the project, create a virtual environment, install dependencies, and set your Serper API key in a .env file. Then add the MCP server configuration to Claude Desktop. After restarting Claude Desktop, ask Claude to search documentation using the get_docs tool, specifying the query, library (langchain, llama-index, or openai), and optional max_chars (default 1000).
Key features of Documentation MCP Server
- Search documentation of LangChain, LlamaIndex, and OpenAI
- Smart HTML extraction with BeautifulSoup
- Configurable result length via max_chars parameter
- Powered by Serper API for site-specific Google searches
- Built with the MCP SDK
- Simple setup with uv and Claude Desktop
Use cases of Documentation MCP Server
- Quickly look up LangChain concepts (e.g., vector stores) during development
- Retrieve LlamaIndex documentation excerpts without leaving the chat
- Access OpenAI API details while building AI applications
- Assist developers by providing contextual documentation in real time
FAQ from Documentation MCP Server
Which libraries are supported by Documentation MCP Server?
The server supports LangChain, LlamaIndex, and OpenAI documentation.
How do I configure Documentation MCP Server?
Install uv, clone the project, create a .env file with your Serper API key, install dependencies, and add the MCP server configuration to Claude Desktop by specifying the command uv --directory /path run main.py in the claude_desktop_config.json.
What parameters does the get_docs tool accept?
The tool accepts query (search term), library (langchain, llama-index, or openai), and max_chars (maximum characters to return, default 1000).
Why might a search return no results?
Check that your Serper API key is correct and that you have an active internet connection.
What should I do if Claude can't find the server?
Verify that the absolute path in your Claude Desktop configuration file matches the actual project directory location.
