Summary
On August 4, 2026, Databricks made Unity AI Gateway generally available: a runtime governance layer for AI agents with hard spend caps, smart routing, service policies, and PII guardrails, presented as one control plane for AI. Some capabilities (smart routing, service policies, agent services) remain in beta, and certain managed MCP integrations are still rolling out.
What changed
Databricks' Unity AI Gateway reached general availability. It provides runtime governance for agents and LLM access, including hard spend caps, smart routing across models, service policies, and PII guardrails, unifying control over which agents can use which models and systems. Smart routing, service policies, and agent services remain in beta; some managed MCP integrations are still rolling out.
Why it matters
Enterprises scaling agents on their data need spend limits, routing, and data-protection guardrails enforced at runtime, not per-application. Shipping this as a GA control plane inside the Databricks platform positions Unity AI Gateway as the governance chokepoint for agentic workloads on enterprise data.
Evidence excerpt
Unity AI Gateway reached general availability on August 4, 2026: Databricks' runtime governance layer for agents with hard spend caps, smart routing, service policies, and PII guardrails; some capabilities including smart routing, service policies, and agent services remain in beta.