Math MCP Example: Server & Client
#############################################
Directory Structure & Environment Setup
#############################################
Recommended project structure:
iceberg-mcp-main/
├── iceberg_mcp/
│ └── math/
│ ├── math_server.py
│ ├── math_clinet.py
│ └── README.md
├── .venv/ # Python virtual environment (recommended)
└── ... # Other project files
Setting up your Python environment
- Create a virtual environment (recommended):
python3 -m venv .venv
source .venv/bin/activate
- Install required dependencies:
(Make sure you are in the root directory or where your requirements are listed)
pip install mcp fastmcp fastapi-mcp langchain-mcp-adapters uvicorn
# And any other dependencies your project needs
- Set your OpenAI API key:
export OPENAI_API_KEY=sk-...your-key-here...
Overview
- Server: Exposes math tools (add, sub, multiply) and prompt templates using FastMCP and SSE (Server-Sent Events).
- Client: Connects to the server, discovers available tools, and uses an LLM agent to invoke those tools.
Learning Objectives
- Understand how to register and expose tools in a Python server.
- Learn how to connect to a tool server and discover available tools.
- See how an LLM agent can use external tools to answer questions.
- Practice async programming and client-server communication.
#############################################
Server: math_server.py
#############################################
What does it do?
- Registers three math tools:
add, sub, multiply. - Registers three prompt templates for natural language generation.
- Exposes an ASGI app for Uvicorn to serve via SSE.
- Logs every tool and prompt call for transparency.
Key code sections
from mcp.server.fastmcp import FastMCP
import logging
# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger("math_server")
# Create the server
mcp = FastMCP("Math")
app = mcp.sse_app # Expose the SSE ASGI app
# Register tools
@mcp.tool()
def add(a: int, b: int) -> int:
result = a + b
logger.info(f"add({a}, {b}) = {result}")
return result
# ... (sub, multiply, and prompts similar)
print("Registered tools:", mcp.list_tools())
>>>>>>>>>>>>>>>>>>>>>>>> How to run the server===================================================================================
From the iceberg_mcp/math directory:
uvicorn math_server:app --port 3000
Or from the project root:
uvicorn iceberg_mcp.math.math_server:app --port 3000
#############################################
Client: math_clinet.py
#############################################
What does it do?
- Connects to the math server using SSE.
- Discovers available tools.
- Uses a LangChain agent to ask the server to add 3 and 5.
- Prints only the final answer from the agent's response.
Key code sections
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
import asyncio
client = MultiServerMCPClient({
"math": {
"transport": "sse",
"url": "http://localhost:3000/sse"
},
})
async def main():
tools = await client.get_tools()
print("Discovered tools:", tools)
agent = create_react_agent("openai:gpt-4.1", tools)
math_response = await agent.ainvoke({"messages": "use the add tool to add 3 and 5"})
if 'messages' in math_response and len(math_response['messages']) > 1:
ai_message = math_response['messages'][-1]
print(ai_message.content)
else:
print(math_response)
if __name__ == "__main__":
asyncio.run(main())
OpenAI API Key Setup
To use the LLM agent (e.g., GPT-4), you need an OpenAI API key. This is required for the client to access OpenAI's language models.
How to set your OpenAI API key:
- The recommended way is to set the
OPENAI_API_KEY environment variable in your shell:
export OPENAI_API_KEY=sk-...your-key-here...
- Alternatively, you can set it in your Python code (not recommended for production):
import os
os.environ["OPENAI_API_KEY"] = "sk-...your-key-here..."
You must set the API key before running the client, or you will get authentication errors.
>>>>>>>>>>>>>>>>>>> How to run the client################################################################
From the iceberg_mcp/math directory:
Experiment & Learn
- Try changing the numbers in the client prompt.
- Add new tools (e.g., division) to the server and see if the client discovers them.
- Add more prompts or logging to see how the server responds.
- Explore how async programming enables real-time tool discovery and invocation.
Troubleshooting
- If the client prints
[] for tools, check server logs and package versions. - Make sure both server and client use compatible MCP and adapter versions.
- Ensure the server is running before starting the client.
Summary
This example demonstrates how to:
- Build a tool-augmented AI server in Python
- Connect and interact with it using a modern LLM agent
- Use async programming for efficient, real-time communication
Happy learning! # mcp