Azure API Management Adds a Dedicated AI Gateway Tier
Microsoft’s public-preview AI Gateway reorganizes Azure API Management around models, MCP servers, and tools rather than ordinary APIs. It federates Azure, Bedrock, Vertex, OpenAI-compatible, and Anthropic providers; exposes policy cards for quotas, token limits, content safety, and fallback; and exports token metrics through OpenTelemetry.
The useful architectural split is central platform control with team self-service: a platform group publishes approved models and tools while applications consume them. The sharp caveat is credential scope—one runtime key reaches every model and tool on a gateway—while preview availability, quotas, and pricing remain unsettled. Reactions also question whether a gateway should govern only traffic or the full lifecycle and auditability of agent work.