Centralize governance for AI agents with Databricks AI Gateway—control LLM access, enforce guardrails, track usage, and gain full visibility across models, tools, and APIs.

databricks

Databricks is expanding AI Gateway to serve as a unified governance layer for agentic AI workflows. The release adds support for governing MCP server usage alongside LLM access, enabling consistent access controls, audit logging, cost tracking, and observability across the full agent lifecycle. Key new capabilities include fine-grained permissions with on-behalf-of user execution for MCP calls, LLM-judge-powered guardrails (PII detection, prompt injection, hallucination guard, custom rules), end-to-end observability via Unity Catalog system tables and Delta-based inference tables, OpenAI-compatible unified APIs for provider portability, and automatic failover between models. Cost attribution can be sliced by team, user, model, or provider.

AI Gateway: A Governance Layer for Agentic AI