Page

Step 1: Download MCP for Python

First, you need to download the MCP for Python from the OpenAPI Tools website. Please visit: https://openapitools.com/dashboard/tools

Download the MCP Python package from this link.

Go into the directory, download the latest versions of these dependencies by running:


pip install -r requirements.txt

This will install the Model Context Protocol SDK and other required dependencies for your TypeScript implementation.

Step 2: Configure mcpconfig.json

Once downloaded, you’ll need to create a configuration file called mcpconfig.json with the following content:


{
  "server_name": "anthropic-mcp-server",
  "host": "0.0.0.0",
  "port": 8000,
  "server_version": "0.1.0",
  "tool_names": [],
  "mode": "stdio",
  "refresh_time_minutes": 1440,
  "api_key": "apik_47677d9a7375d4087bdcf60a6b861d33bff06bff1f5bdddaad0e1a1959759430eaf8bd94cef844245200fedccd55b5c8_c4854ed60d3ee64f"
}

Note: The mode is set to stdio which is required for Claude Desktop integration. If you want to use it as a server, you would change this to sse mode.

Step 3: Configure Claude Desktop

You’ll need to modify Claude Desktop’s configuration file based on your operating system:

For macOS:

Edit the file at: ~/Library/Application Support/Claude/claude_desktop_config.json

For Windows:

Edit the file at: %APPDATA%\Claude\claude_desktop_config.json

Add the following configuration:


{
  "mcpServers": {
    "otp": {
      "command": "uv",
      "args": [
        "--directory",
        "C:\\Users\\kvssa\\Desktop\\openapitools\\mcp-python",
        "run",
        "server.py"
      ]
    }
  }
}

🚫

Important

Adjust the directory path to match where you’ve installed the MCP Python package.
Ensure uvicorn is installed in your Python environment globally. You can install it using pip install uvicorn.
Change the server name from otp to whatever you want to call it. This name will be used in the Claude Desktop interface to select the tool.

Step 4: Providing only few tools or api_key (Optional)

Instead of fetching tools locally you can use your api_key but there will be rate limits. You can also provide only few tools to be fetched from the server.

refresh_time_minutes will refresh tools for for every n minutes if you are using api_key.


{
  "server_name": "anthropic-mcp-server",
  "server_version": "0.1.0",
  "host": "0.0.0.0",
  "port": 8000,
  "tool_names": ["generate-otp", { "name": "get-orders", "version": "prod" }],
  "mode": "stdio",
  "refresh_time_minutes": 1440,
  "api_key": "YOUR_API_KEY_HERE"
}

Step 5: Running Your MCP Server

Once everything is configured, restart Claude Desktop. Your MCP server should be available for use within the Claude interface.

You can test if your integration is working by selecting your tool from the tool menu in Claude Desktop.

Claude Desktop Working