AI governance with Unity AI Gateway

Important

This page covers the new Unity AI Gateway (visible in the sidebar of the UI), which is currently in Beta. Account admins can enable access to this feature in the account console Previews page. See Manage Azure Databricks previews.

For details on the previous version of AI Gateway (not Unity AI Gateway), see AI Gateway for serving endpoints.

Unity AI Gateway is the Azure Databricks governance solution for enterprise AI. Built on Unity Catalog, it extends governance beyond your data and AI assets to the runtime interactions between models, agents, MCP servers, and tools. Control which AI services teams can use, route and manage AI traffic, set guardrails, and monitor usage from one control plane.

Get started

Set up and apply AI governance across your AI assets, traffic, and service behavior.

Topic	Description
AI governance guide	An end-to-end path for administrators to govern access to AI assets, traffic and cost, and the content of requests and responses.
Tutorial: Govern a coding agent's GitHub MCP access	Restrict a coding agent's access to GitHub MCP tools using Unity Catalog permissions and a built-in service policy.
Tutorial: Implement guardrails on a model service with service policies	Implement guardrails on a model service using built-in and custom service policies.

Control which AI services teams can use

Register AI assets as Unity Catalog securable objects, then grant and revoke access with the same privileges you use for tables and volumes. Agents are governed through these same securables: an agent is registered as a Unity Catalog model, and the tools it calls are governed as MCP services, functions, and connections.

Topic	Description
Models	Govern registered ML models, including Azure Databricks-hosted foundation models, with Unity Catalog privileges.
Foundation model permissions	Restrict which Azure Databricks-hosted foundation models your organization can access, account-wide or per group.
MCP tools	Govern MCP servers registered as Unity Catalog securables, with tool filtering and service policies.
Custom tools	Govern the Unity Catalog functions that agents use as tools, with the same privileges you use for data.
HTTP connections	Govern the Unity Catalog connections used to reach external APIs and MCP servers.
Create model services	Define and share model services as Unity Catalog securable objects across workspaces.

Route and manage AI traffic

Unity AI Gateway routes requests to your model and MCP services from a central control plane, so you can manage capacity, availability, and spend across providers.

Topic	Description
Apply rate limits	Enforce consumption limits on model services and MCP services to manage capacity and cost.
Configure traffic splitting and fallbacks	Distribute requests across multiple model backends and add failover to increase availability.
Manage budgets	Monitor spend and set per-user thresholds and hard caps across Azure Databricks-hosted and external providers.

Note

Unity AI Gateway features don't incur charges during Beta.

Set guardrails and access policies

Service policies, also called guardrails, control how each request and response proceeds, based on its content and on who is making the call.

Topic	Description
Service policies for AI securables	How service policies govern the content of requests and responses to AI services, using built-in and custom policies.
Create and attach a service policy	How to write a service policy function and attach it to an MCP Service or Model Service.

Monitor usage, cost, and risk

Track activity, spend, and outcomes across all Unity AI Gateway services.

Topic	Description
Monitor usage	Track requests, token usage, and latency for model services using system tables.
Analyze cost	Attribute Azure Databricks cost to services, target models, principals, and tags.
Audit requests and responses	Log requests and responses to Unity Catalog Delta tables for monitoring and debugging.

Model serving endpoints (previous)

The previous version of AI Gateway has governance features for model serving endpoints at the workspace level, including external model endpoints, Foundation Model API endpoints, and custom model endpoints.

Topic	Description
AI Gateway for serving endpoints	Learn about AI Gateway features for serving endpoints, including supported features and limitations.
Configure AI Gateway on model serving endpoints	Configure AI Gateway features such as usage tracking, payload logging, rate limits, and guardrails on a model serving endpoint.
Monitor served models using AI Gateway-enabled inference tables	Monitor served models using AI Gateway-enabled inference tables.