Model Context Protocol (MCP) Multi-Agent Demo
This project demonstrates how to set up and communicate with custom Model Context Protocol (MCP) servers in Python. It showcases multi-agent orchestration using LangChain, Groq, and MCP adapters, enabling both local and remote tool integration.
Features
- Custom MCP Servers: Math and Weather agents, each as independent MCP servers
- Multi-Transport Communication: Local (stdio) and remote (HTTP) transports
- LangChain Integration: Unified agent interface for tool invocation
- Async Orchestration: Efficient, non-blocking agent communication
Components
1. mathserver.py
A custom MCP server providing math operations (add, multiply) via stdio transport.
2. weather.py
A custom MCP server providing weather information via HTTP transport - (Static content for demo).
3. client.py
A Python client that connects to both servers, discovers their tools, and invokes them using a LangChain agent powered by Groq.
Setup Instructions
-
Clone the repository
-
Install dependencies:
pip install -r requirements.txt
- Set up environment variables:
- Create a
.envfile with yourGROQ_API_KEY:GROQ_API_KEY=your_groq_api_key_here
- Run the servers:
- Start the weather server (in one terminal):
python weather.py - The math server is started automatically by the client when needed.
- Run the client:
python client.py
Example Output
Math Response: The answer is 900
Weather Response: The weather in delhi is sunny
Learning Outcomes
- How to build and register custom MCP servers
- How to enable communication between agents using stdio and HTTP
- How to orchestrate multi-agent workflows with LangChain
Requirements
- Python 3.8+
- See
requirements.txtfor Python dependencies
License
MIT