You can build a framework-agnostic AI gateway on Azure App Service — here's how

The agent infrastructure conversation moved this year. AWS shipped Amazon Bedrock AgentCore as a managed agent runtime — per-session microVM isolation, built-in long-term memory, native MCP support, an opinionated policy engine. Cloudflare’s Agents SDK bets on edge-native stateful agents on Durable Objects. Both are real, and both arrive as closed, proprietary runtimes.

So what’s Azure’s answer? Azure already has the pieces — they just don’t ship as one product called AgentRuntime, and that’s the point. The pitch is composable: App Service + API Management + MCP, three services you already have access to, glued together with open standards.

This post walks through a runnable sample of that composition. One App Service hosts both an agent (Microsoft Agent Framework) and the stateless MCP server it calls. APIM sits in front with the AI Gateway policy set — semantic caching, token rate limiting, per-subscription token emission for chargeback. Managed identity end-to-end, no AOAI keys in app code, App Insights wired up. One azd up deploys the lot.

The headline claim is in the title, but the point underneath it is the one that matters: the framework is replaceable, the gateway is the contribution. Swap the Agent Framework module for Pydantic AI or LangGraph and the rest of the architecture is unchanged. That’s what “run anything” means, made literal — and it’s the structural reason the composable bet beats the proprietary-runtime bet for the long tail of App Service workloads.

👉 Read the full article on Tech Community and grab the sample repo on GitHub.

Comments