What is Unity Catalog in Databricks?

A company's data platform holds thousands of tables, files and models, used by hundreds of people and, now, by AI agents. Without one place that knows what exists and who may touch it, every team keeps its own permissions and nobody can answer "who read the salary table last week?". Unity Catalog is that place on Databricks.

A name for everything

Every asset gets a 3-part name: catalog.schema.object.

SELECT * FROM finance.payroll.salaries;

• A catalog is the top level, often one per business unit or environment (finance, dev, prod).
• A schema groups related objects inside it (payroll).
• An object is a table, a view, a volume of files, a function, a machine learning model or a served model.

The same names work from SQL, Python, notebooks, jobs and dashboards, in every workspace attached to the catalog.

Who may use what

Permissions are granted on those objects:

GRANT SELECT ON TABLE finance.payroll.salaries TO `hr-analysts`;

Beyond plain grants, row filters and column masks hide rows or values from some users, and attribute-based policies apply a rule to every object with a tag, such as "mask anything tagged pii". Because the rule sits in the catalog, not in each tool, an analyst, a notebook and an AI agent asking the same question get the same answer with the same data hidden. For an agent, that makes the catalog a GuardrailGuardrailChecks placed around a model that stop or change what goes in and what comes out, such as a filter on user messages or an approval step before an agent acts. outside the model.

Managed and external

A managed table's files live where Unity Catalog chooses, and it handles their whole life, including deleting them with the table. An external table points at files you manage; the catalog governs access but does not own the files. Newer features need managed tables: Databricks' multi-table transactions, for example, only write to managed DeltaDelta LakeAn open table format that turns Parquet files in cloud storage into a table with transactions, a version history, schema checks, and updates and deletes, so a data lake behaves like a database. or Iceberg tables with catalog commits.

Lineage and audit

Unity Catalog records lineageData LineageThe record of where a piece of data came from and everything it passed through, from source tables to the dashboards, models and AI answers built on it. by itself: which tables fed which, down to the column, through to models and dashboards. When a source column changes, you can see every report and model downstream. An audit log, readable as a system table, records who accessed what.

Open source too

Databricks also publishes Unity Catalog as an open source project, so other engines can use the same catalog interface. Workspaces created since November 2023 have it turned on by default, and most new Databricks features assume it.