mcp-server

Pallavi-Pandey/mcp-server

3.2

If you are the rightful owner of mcp-server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

This project demonstrates how to build a simple MCP server that allows AI assistants to search through documentation for popular AI/ML libraries.

Tools
1
Resources
0
Prompts
0

MCP Server Example: Documentation Search Tool

This project demonstrates how to build a simple MCP (Model Control Protocol) server that allows AI assistants to search through documentation for popular AI/ML libraries including LangChain, LlamaIndex, and OpenAI.

Features

  • Search documentation for multiple libraries (LangChain, LlamaIndex, OpenAI)
  • Uses Google Search API via Serper to find relevant documentation pages
  • Extracts and returns text content from documentation pages
  • Built with FastMCP for easy integration with AI assistants

Prerequisites

  • Python 3.10 or higher
  • Serper API key (for Google Search functionality)

Installation

  1. Clone this repository
  2. Create a virtual environment: 
    python3 -m venv .venv
source .venv/bin/activate
  3. Install dependencies: 
    pip install -e .
  4. Create a .env file with your Serper API key: 
    SERPER_API_KEY=your_api_key_here

Usage

Run the MCP server:

source .venv/bin/activate
python3 main.py

The server will start and listen for requests on the standard input/output (stdio) transport.

How It Works

The server exposes a single tool called get_docs that takes two parameters:

  • query: The search query (e.g., "Chroma DB")
  • library: The library to search in (e.g., "langchain")

When called, the tool:

  1. Constructs a site-specific search query
  2. Sends the query to Google via the Serper API
  3. Fetches the content of the top search results
  4. Returns the extracted text content

Supported Libraries

  • LangChain: langchain
  • LlamaIndex: llama-index
  • OpenAI: openai

License

MIT