MCP Mix Server

You can run a local MCP server that exposes a modular set of tools to a client. This setup lets you test how a client queries, prompts, and analyzes data through a standardized MCP transport, all while keeping the server and client running on your machine.

How to use

Start the MCP server locally and then connect your MCP client to it. The server runs in stdio transport mode, meaning the client communicates with it on the same machine without network endpoints. You will activate a virtual environment, start the server, and from there run a client that talks to the server to perform tasks like data summarization and file analysis.

To verify the server is ready and registered by a compatible client, follow the client’s help section to discover available tools and run sample queries such as summarizing a CSV file or counting rows in a Parquet file.

How to install

Prerequisites you need on your system before starting:

• Python 3.x is installed on your system
• A Python virtual environment tool is available (venv or similar)
• A local command runner is installed (uv)

Step 1: Create and activate a virtual environment if you do not already have one.

source .venv/bin/activate

Step 2: Start the local MCP server using the standard runtime command.

uv --directory . run mcp_server/main.py

Note: There are no outputs from the terminal when the server starts; this is expected and normal.

Configuration and how it works

The MCP server is designed to run locally in stdio transport mode, so the client communicates directly with the server without remote endpoints. Use this setup to test data workflows and tool interactions in a contained environment.

If you want to customize how the server is started from a client, you can reuse the same runtime command pattern shown above, adjusting the directory path to your project layout as needed.

Additional notes

There are no extra outputs to monitor in the terminal once the server is running; this behavior is expected. Use your MCP client to query the server and test tools like summarizing data or counting rows in files.

When you build or extend the client, you can keep using the same stdio transport approach by invoking the client with the server’s main module path, ensuring both run in the same environment.

MCP server configuration example

{
  "mcpServers": {
    "mcp_mix": {
      "type": "stdio",
      "name": "mcp_mix",
      "command": "uv",
      "args": [
        "--directory",
        "{ABSOLUTE_PATH}/mcp-mix-server",
        "run",
        "mcp_server/main.py"
      ],
      "env": []
    }
  }
}

Repository scope and tooling

This setup focuses on running an MCP server locally for experimentation and client testing. Tools and examples shown are intended to help you verify client-server interactions and do practical data tasks with sample inputs.