News Brief

Databricks extends Unity Catalog governance to agents

Historical record This story connects to a documented entry in the Agentic History timeline. As agents began connecting to many tools, standardized connectors and governance for tool access became as important as model choice.

Databricks described extending Unity Catalog governance to cover AI agents, models, MCP servers, and related assets via Unity AI Gateway.

Databricks said the approach enforces identity-aware access, runtime policies, guardrails, auditability, and observability over model calls and tool invocations.

The post emphasized governing by controlling what agents can access and monitoring what they do.

Databricks also described on-behalf-of identity propagation so agents inherit the invoking user’s permissions.

It said logging is tied to both user and agent identities.

Source: Databricks


← All news · Timeline · Research blog