Govern AI models, tools, and agents with Azure API Management | OD831

Anish Tallapureddy and Mike Budzynski explain how Azure API Management’s AI Gateway can act as a governance layer for AI workloads, covering endpoint exposure, request routing, policy-based guardrails, cost controls, and production telemetry for models, tools, and agents across providers.

Overview

AI workloads often behave differently from traditional APIs: they can involve multiple models, tools, and agents across providers, which makes production concerns like routing, cost control, safety, and observability harder.

This session shows how to use Azure API Management (APIM) with its AI Gateway capabilities to:

What the AI Gateway is used for

The presenters position APIM’s AI Gateway as a universal governance layer for AI traffic, focusing on enterprise needs such as:

Model governance and policy enforcement

The session includes a model governance demo centered on using APIM policies to enforce guardrails when calling AI endpoints.

Key themes called out:

New features highlighted

The video calls out newer capabilities in the AI Gateway area, including:

Tool governance (including MCP servers)

Beyond models, the session covers governing tools that agents can call, including:

Tool discovery and agent integration

A demo shows how AI Gateway can support:

Enterprise catalog and agent governance announcements

The session also mentions enterprise-level governance concepts such as:

Resources