Data Catalog is a governed inventory of your organization's data assets. Instead of hunting through databases, warehouses, and SaaS apps to figure out what data exists and whether it can be trusted, you connect those sources once, and the catalog automatically discovers, organizes, and documents everything inside them — down to the individual field.
Overview
A catalog does three things: it finds your data, it explains how that data relates to and flows into other data, and it lets your team attach context — descriptions, owners, sensitivity labels, business terms — that turns a raw table into something people, and increasingly AI agents, can understand and rely on. You organize catalogs by domain or environment, so a marketing catalog and a finance catalog can each hold their own sources without getting tangled together.
Discovering your data automatically
You start by connecting a source — a database, warehouse, or application — through a connector and a connection. Once added, the catalog runs a metadata sync in the background: it crawls the source and indexes its structure as browsable assets, following the hierarchy the source system already uses — application, database, schema, table, field. You don't model or type any of this by hand; the sync discovers it and reports whether each pass completed or failed.
Once a source finishes syncing, you get a searchable table of every asset the catalog knows about, plus an expandable tree mirroring the source hierarchy so you can drill from an application down to a single field. Filter from the tree or search across everything.
Understanding relationships between assets
Assets aren't isolated entries — each one knows its place in the structure around it. A field belongs to a table, a table to a schema, a schema to a database, and a database to the application it synced from. Fields can also carry relationships like foreign keys, showing how tables are meant to connect. That structural map lets you jump from a single column up to its source system, or across to everything else sharing its parent.
Tracing lineage
Lineage shows where an asset's data comes from and where it goes, as an interactive graph with upstream sources on one side and downstream consumers on the other. Available on tables and fields, you expand either direction to walk the chain — including field-level flows, showing exactly which column feeds which. Edges come from the pipelines and integrations that actually move your data, so the graph reflects real, recorded flows.
That's what makes lineage useful day to day: impact analysis, checking what would break downstream before changing something, and root-cause tracing, walking upstream from a bad number to find where it went wrong. An empty graph usually just means no recorded pipeline has touched that asset yet.
Enriching assets with metadata
Discovery gets you the raw structure; metadata is what makes an asset trustworthy and findable. Every asset can carry:
Tags — free-form labels for organizing assets your own way.
Classifications — labels for the sensitivity or category of the data.
Glossary terms — links to defined business terms, so people find data using business language instead of raw table names.
Owners — the people accountable for an asset.
Certificates — a Verified or Not Verified badge signaling whether an asset has been reviewed and can be trusted.
Together with a description and quick facts pulled from the source — record counts, size, primary key — this metadata separates a catalog from a plain schema browser, turning a synced table into a documented, governed asset.
Making data trustworthy for AI applications and agents
A governed catalog isn't only for manual browsing. Because every asset carries structure, lineage, and metadata in one place, AI applications and agents can query it to find the right dataset, understand what a field means, check whether it's verified, and see where its values came from before using it. Glossary terms give agents business-friendly names to search by, classifications and certificates let them respect trust boundaries, and lineage lets them judge whether a number is safe to rely on — the same checks a careful analyst would make, available programmatically.
Data Catalog also pairs with the separate Data Quality module, which layers monitoring and scoring on top of catalog assets — but the catalog itself stands on its own as the map of what data you have, how it connects, and whether it can be trusted.
Notes
Assets appear only after a source's metadata sync completes — a newly connected source can take a little while before its data is browsable.
An asset's name mirrors the source system and can't be edited; only its description and governance metadata can be changed.
Lineage is only as complete as the pipelines feeding it — assets with no recorded activity show an empty graph until one runs.