MCP Docs Search Server

What is MCP Docs Search Server?

A lightweight MCP server that searches and retrieves relevant documentation content from popular AI libraries like LangChain, LlamaIndex, and OpenAI using web search and content parsing. It acts as a bridge between LLMs and external documentation sources, enabling language models to query and fetch up-to-date documentation dynamically.

How to use MCP Docs Search Server?

Clone the repository, create a virtual environment with uv, install dependencies (mcp[cli], httpx, beautifulsoup4), and set the SERPER_API_KEY environment variable. Integrate with Claude Desktop by adding the server configuration to claude_desktop_config.json, then use the get_docs tool with a query and library name. Debug in real time using npx @modelcontextprotocol/inspector uv run main.py.

Key features of MCP Docs Search Server

• Searches documentation via the Serper API (Google).
• Extracts clean, human-readable text with BeautifulSoup.
• Provides a structured get_docs tool for LLM agents.
• Supports LangChain, LlamaIndex, and OpenAI libraries.
• Designed for seamless integration with Claude Desktop.

Use cases of MCP Docs Search Server

• Enabling an LLM to fetch and reason over the latest LangChain documentation during a conversation.
• Querying LlamaIndex or OpenAI docs in real time without manual browsing.
• Providing an LLM agent with up-to-date external knowledge for answering technical questions.
• Serving as a reusable documentation bridge for any MCP-compatible client.

FAQ from MCP Docs Search Server

What dependencies does the server require?

It requires Python (managed via uv), the mcp[cli], httpx, and beautifulsoup4 packages, plus a Serper API key stored as the SERPER_API_KEY environment variable.

What libraries are currently supported?

LangChain, LlamaIndex, and OpenAI. More libraries can be added by updating the docs_urls dictionary.

How does the get_docs tool work?

It accepts a query string and a library name, searches for relevant documentation pages, fetches and parses clean text content, and returns the result to the LLM for further reasoning.

Can I debug the server in real time?

Yes. Use npx @modelcontextprotocol/inspector uv run main.py after installing Node.js 18+, then follow the port where the connection is set up.

How do I integrate this server with Claude Desktop?

Add the server configuration to claude_desktop_config.json under mcpServers, specifying the command uv with the directory and run main.py, then restart Claude Desktop.