How to create an MCP server for LM Studio with FastMCP

Starting from version 0.3.17 (b10), LM Studio supports both local and remote MCP (Model Context Protocol servers). This repository is a simple guide to create an MCP server in Python using the FastMCP library.

Prerequisites

To get started, ensure you have the following prerequisites:

Python: Install the latest version from python.org.
LM Studio: Download and install LM Studio from lmstudio.ai/download.
FastMCP library: Open a terminal and run:
```
pip install fastmcp
```
Optional: A code editor such as VS Code for editing your Python scripts.

1. Write your MCP server logic

Create a Python file and implement your MCP server using FastMCP. Here is a minimal example:

from fastmcp import FastMCP

mcp = FastMCP("Demo 🚀")

@mcp.tool
def add(a: int, b: int) -> int:
    """Add two numbers"""
    return a + b

if __name__ == "__main__":
    mcp.run()

This example registers an add method that takes two parameters and returns their sum. You can add more methods as needed.

➡️ You can also copy the server_math.py file from this repository for a more complete example.

2. Configure LM Studio to use your MCP server

In LM Studio, switch to the "Program" tab in the right-hand sidebar. Click Install > Edit mcp.json.

This will open the mcp.json file in the in-app editor. You can add MCP servers by editing this file.

3. Example of mcp.json configuration

Use the example below as a template. Adapt the path in "args" to the actual location of your Python script!

{
  "mcpServers": {
    "Math": {
      "command": "python",
      "args": [
        "C:/Users/.../server_math.py"
      ]
    }
  }
}

4. Start and test your server

Once your mcp.json is configured, LM Studio will automatically launch your MCP server when needed. You can verify that your server is running by interacting with your tool from the LM Studio interface.

To manually test your server, open a terminal and run:

python server_math.py

This should start your MCP server and display its status.

5. Debugging and tips

Double-check the Python executable path and script location in your mcp.json configuration.
Ensure all required dependencies are installed (fastmcp, etc).
For advanced usage, refer to the FastMCP documentation.

MCP Servers