Back
AzureNovember 18, 20261 sources

Azure AI Gateway reaches GA for governing model and MCP tool traffic across providers

AI Analysis

Azure AI Gateway is now generally available. Microsoft describes it as 'a fully managed Azure service that lets you set up a single, controlled path from your AI applications and agents to models and tools across providers.' The promise is that developers can use whichever models and tools they choose while platform teams govern access, policies and spending from one place. The service covers MCP tools, the Model Context Protocol connectors agents use, as well as model endpoints.

In practice the gateway sits between applications and the model and tool providers. It enforces authentication, rate and token limits, routing and cost attribution centrally, so each team does not hard-code that logic. Microsoft says it is available in 17 Azure regions with 10,000 free requests per gateway each month, a low threshold meant to get teams to try it before they pay. Microsoft also used Ignite to extend 'Copilot in 30' partner trials through March 2027.

The gateway category is getting crowded. Hugging Face just handed its inference routing to OpenRouter, AWS is building AgentCore around identity and memory for agents, and every hyperscaler wants to be the control point between enterprises and a growing number of models. Notably, Microsoft's gateway is model-agnostic in the same week it launched seven in-house MAI models that it would happily route to.

What to watch: pricing beyond the free tier, how deeply it integrates with Microsoft IQ context, and whether multi-cloud customers accept Azure as their central routing point.

Sources
AI Briefing
·Vendors·Curated by AI agents · Updated daily · 2026
Built by Koby Almog