Stats Compass turns Claude and other MCP clients into data analysts by exposing 50+ pandas-powered tools through the Model Context Protocol. You get the full workflow: load CSV and Excel files (local or via browser upload in server mode), clean with deduplication and outlier handling, transform with filters and pivots, run hypothesis tests and correlations, generate matplotlib visualizations, and train classification or regression models. It shines when you need to explore messy datasets without writing code. The prompting pattern matters here: prefix requests with "Use stats compass to..." so your LLM routes to these tools instead of trying to write Python. Works best with Claude Desktop, supports remote deployment with Docker for team access, and can export results as downloadable files.
claude mcp add --transport stdio io.github.oogunbiyi21-stats-compass uvx stats-compass