Databricks Unity AI Gateway: Agent Governance with On-Behalf-Of User Execution, MCP Permissions, and Cross-Model Cost Attribution

Databricks Blog·medium signal

Databricks extended Unity Catalog into agentic AI with Unity AI Gateway, applying enterprise governance (permissions, auditing, policy controls) to how agents access LLMs and MCP servers. The key feature is on-behalf-of (OBO) user execution: when an agent calls an MCP server, it runs with the requesting user's exact permissions, not a shared service account — if a user can't access a Salesforce record, neither can their agent. Also includes end-to-end observability across LLM and MCP calls with cost attribution per model, team, and workflow.