Brings automated data quality checks into Claude. It scans CSV files and tabular data to discover validation rules without you writing them first. The server exposes scan, validate, and baseline operations. You can profile a dataset to surface type mismatches, nulls, outliers, and format issues, then promote the findings you care about into persistent rules saved as YAML. The validate command enforces those rules in CI pipelines. It also supports drift detection by learning a baseline from reference data and comparing new data against it. The TUI is stripped out in MCP mode, but all the profiling and validation logic works through Claude's tool interface. Built in both Python and TypeScript, with optional LLM enhancement for richer diagnostics.
claude mcp add --transport stdio benzsevern-goldencheck uvx goldencheck