How to Build a Custom MCP Server in Python for Claude: Step-by-Step Tutorial

March 16, 2026

TOC

What Is MCP (Model Context Protocol)?

MCP (Model Context Protocol) is an open standard developed by Anthropic that enables AI assistants like Claude to connect with external data sources and tools. Think of it as a universal plugin system — MCP servers expose capabilities (called “tools”) that Claude can call to interact with databases, APIs, file systems, and custom business logic.

Why Build a Custom MCP Server?

Connect Claude to your internal APIs — Let Claude query your company databases, CRM, or internal tools
Automate workflows — Give Claude the ability to trigger deployments, update tickets, or send notifications
Access private data — Connect Claude to data sources it can’t reach through the public internet
Custom business logic — Implement domain-specific calculations or transformations

Prerequisites

Python 3.10+
Claude Code or Claude Desktop installed
Basic familiarity with Python async/await

Step 1: Install the MCP Python SDK

pip install mcp

Step 2: Create Your Server

Create a new file called my_server.py:

from mcp.server.fastmcp import FastMCP

# Create an MCP server
mcp = FastMCP("My Custom Server")

# Define a tool
@mcp.tool()
def get_weather(city: str) -> str:
    """Get current weather for a city."""
    # Your implementation here
    return f"Weather in {city}: 72°F, Sunny"

# Define a resource
@mcp.resource("config://app")
def get_config() -> str:
    """Return application configuration."""
    return "App version: 1.0.0"

Step 3: Configure Claude to Use Your Server

Add your server to Claude’s MCP configuration:

{
  "mcpServers": {
    "my-server": {
      "command": "python",
      "args": ["path/to/my_server.py"]
    }
  }
}

Step 4: Advanced Features

Adding Database Access

import sqlite3

@mcp.tool()
def query_database(sql: str) -> str:
    """Execute a read-only SQL query."""
    conn = sqlite3.connect("mydb.sqlite")
    cursor = conn.execute(sql)
    results = cursor.fetchall()
    conn.close()
    return str(results)

Adding API Integration

import httpx

@mcp.tool()
async def fetch_api_data(endpoint: str) -> str:
    """Fetch data from an external API."""
    async with httpx.AsyncClient() as client:
        response = await client.get(f"https://api.example.com/{endpoint}")
        return response.text

Best Practices

Keep tools focused: Each tool should do one thing well
Write clear descriptions: Claude uses the docstring to understand when and how to call your tool
Handle errors gracefully: Return meaningful error messages, don’t let exceptions crash the server
Use type hints: They help Claude understand what parameters to pass
Implement read-only by default: Be cautious about giving write access to databases or APIs

Real-World MCP Server Ideas

Jira/Linear ticket management
Database query interface for analytics
CI/CD pipeline monitoring and triggering
Slack message sending and channel reading
Custom documentation search
Internal knowledge base access

Conclusion

Building MCP servers is one of the most powerful ways to extend Claude’s capabilities for your specific needs. Start with a simple tool, test it with Claude Code, and gradually add more functionality. The MCP ecosystem is growing rapidly, and custom servers let you stay at the forefront of AI-assisted development.

Let's share this post !