MCP Demo

A demonstration of the Model Context Protocol (MCP) using LangChain and LangGraph. This project showcases how to create and connect to multiple MCP servers (Math and Weather) and use them with a ReAct agent powered by Groq.

Overview

This demo includes:

• Math Server: An MCP server that provides mathematical operations (add, subtract) using stdio transport
• Weather Server: An MCP server that provides weather information using streamable-http transport
• Client: A LangGraph ReAct agent that connects to both servers and can use their tools

Prerequisites

• Python 3.11 or higher
• Groq API key (get one from Groq)

Installation

1. Clone or navigate to this directory:

cd MCP_Demo

1. Install dependencies using pip:

pip install -r requirements.txt

Or if using uv:

uv sync

Setup

1. Create a .env file in the project root:

GROQ_API_KEY=your_groq_api_key_here

1. Make sure you have the required dependencies installed:
   • langchain-groq
   • langchain-mcp-adapters
   • mcp
   • langgraph
   • langchain
   • python-dotenv

Running the Demo

Step 1: Start the Weather Server

In a separate terminal, start the weather server:

python weather.py

The weather server will run on http://localhost:8000/mcp using streamable-http transport.

Step 2: Run the Client

In another terminal, run the client:

python client.py

The client will:

1. Connect to both the math server (via stdio) and weather server (via HTTP)
2. Create a ReAct agent using Groq's Qwen model
3. Execute two example queries:
   • A math calculation: "What is (10 + 20) * 3?"
   • A weather query: "What is the weather in Pune?"

Project Structure

MCP_Demo/
├── client.py          # Main client that connects to MCP servers and runs the agent
├── mathserver.py      # Math MCP server (stdio transport)
├── weather.py         # Weather MCP server (streamable-http transport)
├── requirements.txt   # Python dependencies
├── pyproject.toml     # Project configuration
├── .env              # Environment variables (create this)
└── README.md         # This file

How It Works

1. MCP Servers: Both mathserver.py and weather.py use FastMCP to create MCP servers with different transport mechanisms:

   • Math server uses stdio transport (direct process communication)
   • Weather server uses streamable-http transport (HTTP-based)

2. Client: The client.py uses MultiServerMCPClient from langchain-mcp-adapters to:

   • Connect to multiple MCP servers simultaneously
   • Expose their tools to a LangGraph ReAct agent
   • Use Groq's Qwen model for reasoning and tool execution

3. Agent: The ReAct agent can autonomously:

   • Understand user queries
   • Decide which tools to use
   • Execute tools from either server
   • Provide natural language responses

Features

• Multi-server MCP integration
• Support for different transport mechanisms (stdio and HTTP)
• LangGraph ReAct agent for autonomous tool use
• Groq API integration for fast inference
• Example tools: Math operations and Weather queries

Troubleshooting

ModuleNotFoundError

If you encounter ModuleNotFoundError: No module named 'langchain', make sure you've installed all dependencies:

pip install -r requirements.txt

Weather Server Connection Issues

Make sure the weather server is running before starting the client. The client expects it at http://localhost:8000/mcp.

API Key Issues

Ensure your .env file contains a valid GROQ_API_KEY. You can get one from Groq Console.

License

This is a demo project for educational purposes.