Know Your Data Inside Out: Data Catalog & ERD

As organizations grow, so does the complexity of their data. Tables multiply, column names diverge across teams, and tribal knowledge gets lost when people change roles. The Data Catalog solves this by automatically documenting everything the moment you connect a source.

GradientHarbor scans your databases and builds a searchable catalog of every table and column. But raw metadata alone is not enough — so we use AI to generate plain-English descriptions for each element. A column named `cltv_90d` gets annotated as "Customer lifetime value over the trailing 90-day window, in USD." Your team spends less time guessing and more time analyzing.

The interactive Entity-Relationship Diagram (ERD) gives you a visual map of how tables connect. Click any table to see its columns, data types, and relationships. Zoom in on a specific schema or zoom out to see the full landscape. It is the fastest way to onboard new analysts or debug a tricky JOIN.

Descriptions are fully editable. If the AI gets something wrong, fix it once and the correction persists. You can also add tags, owners, and custom metadata to organize tables by domain — finance, marketing, product, and so on.

The catalog powers the rest of GradientHarbor. When the AI agent writes SQL, it references catalog descriptions to choose the right tables. When you search for data, semantic search uses those descriptions to find results by meaning, not just column names.

Connect your first data source and the catalog builds itself. No setup scripts, no YAML files, no manual entry.