Quick answer
Use AWS Glue Data Catalog when the authoritative metadata and S3 query path already live in AWS and governance is built around IAM Identity Center, Lake Formation, and Athena. Use Databricks Unity Catalog when the domain is governed and queried through Databricks and OAuth or PAT matches the deployment.
Both integrations are in preview and support the same high-level Quick workflow: natural-language discovery, bulk creation of catalog representations, and semantic inheritance. They are not interchangeable governance paths.
| Decision | AWS Glue Data Catalog | Databricks Unity Catalog |
|---|---|---|
| Discovery auth | Service role or IAM Identity Center | PAT or OAuth 3LO |
| Query path | DirectQuery; Glue metadata commonly paired with Athena for S3 data | DirectQuery through the Databricks connection |
| Per-user identity | Trusted identity propagation with IAM Identity Center and Lake Formation | OAuth 3LO with end-user identities |
| Catalog authority | Glue and its governed metadata | Unity Catalog and its governed metadata |
| Quick fallback without propagation | Quick RLS and CLS | Quick RLS and CLS |
What is common
For either catalog, an author can choose Explore data, describe a use case, review discovered assets, create DirectQuery Datasets, review inherited or inferred relationships, and create a multi-dataset Topic.
The upstream catalog remains the metadata source of truth. Inherited table and column descriptions are read-only in generated Datasets. Authors can choose Sync semantics to refresh them.
The Agent is advisory. Review every table, description, key, relationship, and Topic before making it available to end users. Follow the setup workflow.
Choose Glue when
- Glue is already the approved catalog for the domain;
- data access is governed through Lake Formation;
- IAM Identity Center supplies workforce identity;
- Athena is the accepted query path for cataloged S3 data;
- the platform team owns cross-account roles, Lake Formation grants, and S3 Access Grants.
Trusted identity propagation has prerequisites beyond selecting a checkbox. The Quick account must use IAM Identity Center, Glue must be governed by Lake Formation, identities and groups need grants, and the query results path may require S3 Access Grants.
Choose Databricks when
- Unity Catalog is the semantic and governance authority;
- the approved query path remains in Databricks;
- OAuth 3LO is available for end-user enforcement;
- the team already operates workspace, catalog, schema, and table permissions;
- a PAT is acceptable only for the bounded preview or service path where selected.
PAT or OAuth can support the agentic discovery flow, but OAuth 3LO is the documented path for per-user identity propagation. Do not equate a successful shared connection with correct user-level authorization.
Test the same control matrix
For either choice, validate:
- discovery shows only the intended catalog boundary;
- the author can access only approved assets;
- DirectQuery reaches the correct source and query engine;
- generated Dataset definitions match upstream descriptions;
- Topic relationships reproduce known totals;
- user A cannot see user B’s restricted rows or columns;
- switching to SPICE or adding transformations disables the controls AWS says it disables;
- semantic sync behavior and ownership are documented.
The identity propagation guide explains why discovery authentication and query-time enforcement are different decisions.
Avoid catalog duplication
Do not choose a second catalog only to make the Quick preview easier. If the organization already has authoritative definitions in one platform, copying them into another can create ownership ambiguity and semantic drift. Quick is documented as a consumer of upstream metadata, not a replacement enterprise catalog.
Frequently asked questions
Does Amazon Quick support both Glue and Databricks catalogs?
Yes. The current preview supports AWS Glue Data Catalog and Databricks Unity Catalog for agentic discovery, bulk creation, and semantic inheritance.
How does authentication differ?
Glue discovery supports a service role or AWS IAM Identity Center; Databricks supports PAT or OAuth 3LO. Per-user propagation requires IAM Identity Center plus Lake Formation for Glue or OAuth 3LO for Databricks.
Should catalog choice be based on the Quick Agent?
No. Begin with the authoritative catalog, query engine, identity system, permissions, and operating ownership already used for the data domain.
Official sources
- Amazon Quick: Supported catalogs
- Amazon Quick: Identity propagation
- Amazon Quick: Glue trusted identity propagation
Source check: August 4, 2026. Both integrations are preview features; recheck authentication, identity, and dataset behavior before deployment.