Your First MCP Server in Python: Connect Custom Tools to Claude🚀

Did you know you can create your own tools for Claude Desktop without consuming your token quota? I'll show you how to build a Model Context Protocol (MCP) server in 5 minutes.

What is MCP? 🤔

Model Context Protocol is an open standard that allows AIs to access external tools. Think of it as creating "plugins" for your favorite AI.

Key benefits:

✅ Free: Doesn't consume AI tokens
✅ Reusable: Works with multiple clients (Claude, Windsurf, etc.)
✅ Extensible: Connect APIs, databases, or any Python code

Quick Setup 🛠️

# 1. Create project
mkdir my-mcp-server && cd my-mcp-server
python -m venv venv

# 2. Activate environment
# Windows: .\venv\Scripts\Activate.ps1
# macOS/Linux: source venv/bin/activate

# 3. Install dependencies
pip install mcp fastmcp

The Code: 2 Tools in 30 Lines 🎯

#!/usr/bin/env python3
from mcp.server.fastmcp import FastMCP

mcp = FastMCP()

@mcp.tool()
def calculator(a: int, b: int) -> str:
    """Adds two numbers."""
    result = a + b
    return f"OK {a} + {b} = {result}"

@mcp.tool()
def generate_greeting(name: str, language: str = "english") -> str:
    """Generates greetings in different languages."""
    greetings = {
        "english": f"Hello {name}, have a great day!",
        "spanish": f"Hola {name}, que tengas un excelente dia!",
        "french": f"Bonjour {name}, passez une excellente journee!"
    }
    return greetings.get(language.lower(), greetings["english"])

if __name__ == "__main__":
    print("MCP Server started...")
    mcp.run()

Connect with Claude Desktop 🤖

Run the server:

   python server.py

Configure Claude Desktop: Edit %APPDATA%\Claude\claude_desktop_config.json:

   {
     "mcpServers": {
       "MY-SERVER": {
         "command": "C:\\path\\to\\your\\venv\\Scripts\\python.exe",
         "args": ["C:\\path\\to\\your\\server.py"],
         "cwd": "C:\\path\\to\\your\\project"
       }
     }
   }

Test it!
- "Add 15 and 27" → Uses calculator
- "Greet Ana in Spanish" → Uses generate_greeting

Key Concepts 📚

Tools vs Prompts

Tools (@mcp.tool()): Execute code, return data
Prompts (@mcp.prompt()): Generate text templates

When to use each?

Tool: Calculate, query APIs, process data
Prompt: Generate emails, documents, templates

Real-World Use Cases 🎯

# Connect with external API
@mcp.tool()
async def get_weather(city: str) -> dict:
    """Gets current weather."""
    # Your logic here
    pass

# Query database
@mcp.tool()
def find_customer(id: int) -> dict:
    """Finds a customer in the database."""
    # Your logic here
    pass

Common Troubleshooting 🔧

UTF-8 Error: Avoid emojis in prints or configure encoding.
Won't connect: Verify absolute paths in configuration.
Tools don't appear: Restart Claude Desktop after changes.