MCP Server

A minimal Model Context Protocol (MCP) server that exposes documentation-search tools for any AI agent. Use this server to fetch and search docs for popular libraries (LangChain, LlamaIndex, OpenAI) and wire them seamlessly into your agents.

Features

• Doc Search Tool: Provides a get_docs tool to search and fetch documentation based on a query and library.
• Web Scraping: Uses Serper API for search results and BeautifulSoup for fetching page text.
• Easy Integration: Decorate functions with @mcp.tool() and run the server over stdio transport.

Prerequisites

• Python 3.8 or higher
• A Serper API key (set in .env as SERPER_API_KEY)

Installation

1. Clone this repo:

   git clone https://github.com/<your-username>/MCP-Server-Latest_Doc-.git
   cd MCP-Server-Latest_Doc-
   

2. Create a virtual environment & install dependencies:

   python3 -m venv .venv
   source .venv/bin/activate
   pip install -r requirements.txt
   

3. Add your Serper API key in a .env file:

   echo "SERPER_API_KEY=your_api_key_here" > .env
   

Usage

Running the MCP Server

Execute the main script to start the MCP server on stdio:

python main.py

Using the get_docs Tool

Within any AI agent that supports MCP, the get_docs tool will automatically register. It expects two arguments:

1. query: The search term (e.g., "Chroma DB").
2. library: One of langchain, llama-index, or openai.

The tool returns the concatenated text of the top search results from the specified library docs.

Integrating into an AI Agent

Any agent framework that implements the MCP transport can import and run this server. For example, in Python:

from mcp.server.fastmcp import FastMCP
from dotenv import load_dotenv
import os

load_dotenv()

mcp = FastMCP("docs")

# ... define @mcp.tool() functions here ...

if __name__ == "__main__":
    mcp.run(transport="stdio")

After starting this script, your agent can call the get_docs tool over the stdio channel.

Example: Using with Claude Code CLI

First, install the Claude Code CLI:

npm install -g @anthropic-ai/claude-code

Add the MCP server to Claude:

claude mcp add

Specify the uv binary path (find with where uv):

claude mcp configure --binary-path /full/path/to/uv

Set the working directory of the server:

claude mcp configure --workdir /path/to/MCP-Server-Latest_Doc-

Run the MCP server:

python main.py

Confirm registration:

claude mcp list

For debugging, you can use the Inspector:

npx @modelcontextprotocol/inspector uv run main.py

License

MIT ©