Catalog

The catalog is your workspace's inventory of governed metadata: the tables and columns discovered across your connected platforms, plus the vocabulary you curate on top of them. That vocabulary is what canonical policies target, so catalog curation is governance work, not housekeeping. The catalog is organized into four tabs — Tables, Collections, Taxonomy, and Scenarios.

Tables

The Tables tab opens with a stats row — Catalog tables, Governed by a policy, Classified tables, and Data sources — followed by a search box and the table list grouped by database and schema. Search is instant and spans table names, paths, column names, comments, and tags; the query lands in the URL, so a filtered view is shareable. Each row shows the table's full path, column count, and sensitivity, and opens the table's detail page.

Table detail

The detail page keeps two signals about every column deliberately separate:

• Classification — what the data itself has been identified as, from pattern-based scans, AI-assisted classification, or a manual choice per column.
• Governance — what published policy decisions say about the asset, including policy-derived meaning such as sensitivity, category, domain, owner, steward, purpose, and lineage.

Keeping them separate means a column classified as PII with no policy yet is flagged on its own merits instead of blending into "ungoverned."

Around the columns list you'll find a table metadata card for a user-authored description and tags, a classification summary, an Associated Policies card listing the policies that actually govern this table, and a table info card with identity facts. Editors can edit descriptions, tags, and classifications inline; viewers see everything read-only. A table-level classify action offers a pattern-based scan (admins — it runs through a data-source sync) and AI-assisted classification (editors and above).

Collections

Collections are named, reusable sets of tables that a policy can target as a group — for example, every table that makes up a customer view, regardless of which schema or platform each lives in. The sample workspace ships a "Customer 360" collection built this way. Editors create collections and manage their membership from a picker over the current catalog. Targeting a collection keeps a policy attached to intent ("the customer view") rather than to a list of table paths that drifts; see policy YAML for how a policy references one.

Taxonomy

The Taxonomy tab manages the classification types available to your workspace. A built-in taxonomy covers common types — identity and contact PII among them — each carrying a category, a sensitivity level, and a PII flag. Alongside it, editors can define custom types with their own label, category, sensitivity, PII flag, and description. Custom types behave exactly like built-ins: they appear in the per-column classification selectors and in classification summaries, and policies can target them. This tab is the single home for taxonomy management; the corresponding Settings address redirects here for older links.

Scenarios

Scenarios are the controlled vocabulary of data uses that deployable instructions and decisions reference — the canonical set is fixed platform-wide, so answers stay comparable across workspaces. What you curate here are workspace labels: your own terminology mapped onto a canonical scenario without introducing a new runtime key. See scenarios for the vocabulary itself.

Where catalog content comes from

Tables and columns arrive by sync from your data sources: each connector's latest completed sync populates its slice of the catalog, and the browser reads across every active connector. If you have nothing connected yet, the sample workspace loads a complete, already-classified demo estate instead. Either way, curation is recorded: metadata edits, classifications, collection changes, and scenario labels all appear in the workspace activity trail.