Unlock the Power of LLMs: A Beginner's Guide to MCP Servers in Python

Are you struggling to connect Large Language Models (LLMs) with external data and tools? Do you want to expand LLMs beyond simple text generation to interact with the real world? This guide walks you through creating your own Model Context Protocol (MCP) server in Python, allowing you to integrate LLMs with databases, APIs, and more.

What is Model Context Protocol (MCP) and Why Should You Care?

LLMs like GPT and Claude are great at generating text, but they can’t access your files, query databases, or trigger actions on their own. The Model Context Protocol (MCP) is a standardized way for applications to provide context to LLMs, essentially giving them a "USB-C port" to the outside world. MCP enables LLMs to do more, like querying databases or sending emails.

Key Benefits of Using MCP:

• Extends LLM Capabilities: Connect LLMs to external data and tools.
• Standardized Interface: Provides a universal way for LLMs to interact with different services.
• Automation and Efficiency: Automate tasks by integrating LLMs with real-world actions.

Building Your First Python MCP Server: Step-by-Step

Ready to build an MCP server to bridge the gap between your LLM and external resources? This section guides you through setting up a local MCP server using Python that can query a SQLite database.

Prerequisites:

• Python 3.7+
• SQLite (with a community.db file)
• Cursor Pro or Claude Desktop
• Terminal (macOS/Linux) or PowerShell/CMD (Windows)

Step 1: Setting Up Your Environment

First, create a virtual environment to isolate your project dependencies.

python -m venv mcp-env
source mcp-env/bin/activate  # For Windows: mcp-env\Scripts\activate

Next install the MCP Python SDK.

pip install mcp

Step 2: Grab the Sample SQLite Database

Download the community.db database file, which contains a chatters table with sample data. This database will be queried by our MCP server.

Step 3: Writing Your MCP Server in Python

Create a file named sqlite-server.py and add the code below:

from mcp.server.fastmcp import FastMCP
import sqlite3

# Initialize the MCP server with a friendly name
mcp = FastMCP("Community Chatters")

# Define a tool to fetch the top chatters from the SQLite database
@mcp.tool()
def get_top_chatters():
    """Retrieve the top chatters sorted by number of messages."""
    # Connect to the SQLite database
    conn = sqlite3.connect('community.db')
    cursor = conn.cursor()
    # Execute the query to fetch chatters sorted by messages
    cursor.execute("SELECT name, messages FROM chatters ORDER BY messages DESC")
    results = cursor.fetchall()
    conn.close()
    # Format the results as a list of dictionaries
    chatters = [{"name": name, "messages": messages} for name, messages in results]
    return chatters

# Run the MCP server locally
if __name__ == '__main__':
    mcp.run()

This simple server defines one tool, get_top_chatters, which connects to your SQLite database, retrieves the sorted data, and returns it in an easy-to-read format.

Integrating Your MCP server with Cursor for Enhanced LLM Interactions

Integrating your shiny new MCP server into Cursor allows your LLM to access external data seamlessly. Here's how:

1. Open Cursor Settings: Go to Settings > MCP (Cursor Pro required).
2. Add New Server: Click Add New Global MCP Server to open the mcp.json file.
3. Configure Your Server: Add the server details to the file:

{
  "mcpServers": {
    "sqlite-server": {
      "command": "/path/to/your/project/venv/bin/python",
      "args": [
        "/path/to/your/project/sqlite-server.py"
      ],
      "description": "A simple MCP server to query top chatters from a community database"
    }
  }
}

Important: Replace /path/to/your/project/ with the actual path to your project.

4. Save and Verify: Save the file and return to MCP Settings. You should see a green dot next to your server.

Congratulations! Now you can query the top chatters using Cursor.

Connecting to Claude Desktop for Enhanced LLM Functionality

You can also integrate your MCP Server with Claude Desktop. Here’s how:

1. Open Claude Desktop Settings: Go to Settings > Developer > Edit Config.
2. Add Server Block: Add the following to claude_desktop_config.json:

{
  "servers": [
    {
      "name": "Community Chatters Server",
      "command": "python",
      "args": [ "server.py"],
      "description": "A simple MCP server to query top chatters from a community database"
    }
  ]
}

3. Restart Claude: Save, close, and reopen Claude Desktop for the changes to take effect.

Testing Your MCP Server: Real-World Examples

Now that your server is connected to both Cursor and Claude, it's time to put it to the test.

Testing in Cursor

1. Ask a Question: In Cursor, ask "How many chatters are in the database?"
2. Approve the Tool: A prompt will appear asking for permission to run the tool. Approve the request.
3. Review the Output: Cursor will display the number of chatters, confirming the tool is working correctly.

Testing in Claude Desktop

1. Ask a Question: In Claude Desktop, ask "Show me the list of top chatters."
2. Approve the Tool: Approve the prompt requesting permission to run the MCP tool.
3. Review the Output: Claude will then display the data returned by the MCP server.

FAQs: Addressing Your Key Questions About MCP

Here are some frequently asked questions regarding developing the MCP server.

What is the purpose of the MCP Server in this tutorial?

The MCP Server queries an SQLite database and provides data (number of chatters, names, message counts) to applications like Cursor or Claude Desktop, improving user experience and insights.

How do I integrate my MCP Server with Claude Desktop?

Add the server block to claude_desktop_config.json, then save, close, and reopen Claude Desktop.

What is the significance of the MCP ecosystem?

The MCP ecosystem connects LLMs to external data sources, allowing for more accurate and informative AI responses across various applications.

Conclusion: Your Journey into the MCP World Begins Now

You've now successfully built and integrated a basic MCP server with both Cursor and Claude Desktop and are starting to realize the power of LLMs. This opens a new world of possibilities for automation, data integration, and enhanced LLM interactions. Continue experimenting with different data sources, APIs, and integrations to fully unlock the potential of MCP.