Multi-Server Model Context Protocol (MCP) Demo

A complete demonstration of building and integrating multiple Model Context Protocol (MCP) servers with a LangGraph ReAct agent powered by Groq and real-time OpenWeatherMap data.

This project showcases how an LLM agent can dynamically discover and consume tools across different MCP servers and transport protocols (stdio and streamable-http) in a single unified workflow.

🏗️ Architecture

flowchart LR
    subgraph Agent["LangGraph ReAct Agent (client.py)"]
        ChatGroq["LLM: Groq (Qwen / ChatGroq)"]
        MultiMCP["MultiServerMCPClient"]
    end

    subgraph MathServer["Math MCP Server (mathserver.py)"]
        MathTools["Tools: add, sub, mul, div"]
    end

    subgraph WeatherServer["Weather MCP Server (weatherserver.py)"]
        WeatherTools["Tool: get_weather (OpenWeatherMap API)"]
    end

    MultiMCP -- "stdio transport (subprocess)" --> MathServer
    MultiMCP -- "streamable-http (http://127.0.0.1:8000/mcp)" --> WeatherServer

Components

1. Math MCP Server (mathserver.py):

   • Built with FastMCP.

   • Runs over stdio transport.

   • Exposes arithmetic tools: add, sub, mul, div.

   • Automatically launched as a background subprocess by the MCP client.

2. Weather MCP Server (weatherserver.py):

   • Built with FastMCP running over streamable-http transport at http://127.0.0.1:8000/mcp.

   • Uses httpx to fetch live meteorological data from the OpenWeatherMap API.

   • Exposes the get_weather tool (returns temperature, weather description, feels like, and humidity).

3. Orchestrator Client (client.py):

   • Connects to both MCP servers simultaneously using MultiServerMCPClient from langchain-mcp-adapters.

   • Assembles the tools and powers a LangGraph ReAct agent using Groq's fast LLM inference (ChatGroq).

Related MCP server: AIE8-MCP Server

📁 Project Structure

Mcp_demo/
├── client.py           # ReAct Agent & Multi-Server MCP client orchestrator
├── mathserver.py       # FastMCP server running on stdio transport
├── weatherserver.py    # FastMCP server with OpenWeatherMap API on HTTP transport
├── requirements.txt    # Python package dependencies
├── pyproject.toml      # Project configuration & metadata
├── .env.example        # Environment variables template
├── .gitignore          # Git ignore rules (protects .env and caches)
└── README.md           # Project documentation

🚀 Getting Started

1. Prerequisites

• Python 3.10+

• Groq API Key

• OpenWeatherMap API Key (free tier)

2. Installation

Clone the repository and install dependencies:

git clone https://github.com/sreeshanthkprakash-stack/MCP-Server.git
cd MCP-Server

# Create and activate virtual environment
python -m venv .venv
# On Windows:
.venv\Scripts\activate
# On Linux/macOS:
source .venv/bin/activate

# Install dependencies
pip install -r requirements.txt

3. Environment Configuration

Create a .env file in the root directory (or copy from .env.example):

GROQ_API_KEY=your_groq_api_key_here
OPENWEATHER_API_KEY=your_openweather_api_key_here

🧪 Running the Application

Step 1: Start the Weather MCP Server

Since the weather server communicates over HTTP, start it first in a terminal:

python weatherserver.py

The weather server will start listening at http://127.0.0.1:8000/mcp.

Step 2: Run the Agent Client

Open a second terminal window (ensure your virtual environment is active), and run:

python client.py

Expected Output Example

math_response The result of the expression ((15*20)+200)/4 is 125.
weather_response Weather in Hyderabad: haze, temperature 29.5°C (feels like 32.1°C), humidity 65%

📦 Dependencies

• mcp: Official Model Context Protocol SDK (FastMCP)

• langchain-mcp-adapters: Connects LangChain to MCP servers

• langgraph: Graph-based agent orchestration runtime

• langchain-groq: High-speed LLM inference with Groq

• httpx: Async HTTP client for weather API requests

• python-dotenv: Secure environment variable loading

📄 License

This project is open-source and available under the MIT License.