Python MCP Server - Documentation Search

This is a simple MCP (Minimal Command Protocol) server that allows users to search for documentation from popular

libraries such as LangChain, LlamaIndex, and OpenAI using the Serper API. The server fetches search

results and extracts the relevant documentation from the web using HTTP requests and BeautifulSoup.

Features

• Supports searching documentation for LangChain, LlamaIndex, and OpenAI.
• Uses Serper API to perform web searches.
• Extracts and returns the text from the relevant documentation pages.
• Can be used interactively through MCP protocol.

Requirements

• Python 3.12 or higher.
• Serper API Key for performing web searches.
• MCP library for the server and tool execution.
• BeautifulSoup4 and httpx for HTTP requests and HTML parsing.

Installation

Step 1: Install Python Dependencies

To set up the project, clone this repository and install the required dependencies.

git clone 
cd 
pip install -r requirements.txt

Or, if you're not using a requirements.txt file:

pip install httpx beautifulsoup4 mcp python-dotenv

Step 2: Set Up Environment Variables

Create a .env file in the root directory of the project with your Serper API Key.

SERPER_API_KEY=your-serper-api-key-here

Make sure the .env file is loaded using the dotenv package.

Step 3: Run the Server

Once the environment is set up, run the server using the following command:

uv run main.py

The server will start and wait for input.

Usage

Once the server is running, you can use it to query the latest documentation for specific libraries. Here’s an example

of how to query the tool:

Example Query

Start your server with ``uv run main.py``.

In another terminal or from within an interactive MCP client, invoke the get_docs tool to search for documentation:

get_docs("Chroma DB", "langchain")

This will search for "Chroma DB" in the LangChain documentation and return the relevant content.

Available Libraries

• LangChain: Documentation at langchain.com
• LlamaIndex: Documentation at llamaindex.ai
• OpenAI: Documentation at platform.openai.com

Code Overview

- FastMCP: Initializes the MCP server.
- Tools: The server has two tools:
- dummy_tool: A simple tool that confirms the server is up and running.
- get_docs: Fetches the latest documentation for a given query and library (LangChain, LlamaIndex, OpenAI).
- search_web(): Handles searching using the Serper API.
- fetch_url(): Fetches a URL and extracts the text content using BeautifulSoup.

• Loads the Serper API Key from the ``.env`` file.
• Provides debug-level logging for better visibility into server actions and potential issues.`````

Troubleshooting

• Missing API Key: Make sure the Serper API Key is set correctly in the ``.env`` file. You’ll get an error if it's

missing.

• Timeout Issues: If a search or URL fetch times out, try increasing the timeout values or checking the network

connectivity.

• Package Installation Issues: Ensure all dependencies are installed and the correct Python environment is activated.

Contributor

Udayan Sawant