Stats Compass

The stats-compass MCP lets you run data analysis and machine learning workflows through a modular server that you connect to with MCP clients. It exposes a range of data-handling capabilities—from loading and cleaning data to transforming, exploring, visualizing, and building predictive models—so you can interact with your data using familiar tools and interfaces across different clients.

How to use

You set up the MCP server and then connect from your preferred MCP client (for example, a desktop client, an editor with built-in MCP support, or a web-based interface). Once connected, you can load datasets, perform cleaning and transformations, run exploratory data analysis, create visualizations, and execute machine learning workflows. Start by launching the server, then connect your client to the server endpoint. You can upload local files when operating remotely, export results, and reuse intermediate data across steps.

How to install

Prerequisites: Python is required to run the MCP server locally. Ensure Python is installed on your system and available on your PATH.

pip install stats-compass-mcp

stats-compass-mcp install --client claude

stats-compass-mcp install --client vscode

claude mcp add stats-compass -- uvx stats-compass-mcp run

Restart your client and start asking questions about your data.

## Additional setup and deployment notes

You can run the MCP server in remote server mode to accept connections from multiple clients or from a browser-based upload/download workflow. Start the server with a dedicated port to listen for client connections.

stats-compass-mcp serve --port 8000