Hosting an AG-UI Agent on Microsoft Foundry

This article builds on a previous post about creating an AI agent server with AG-UI and the Microsoft Agent Framework. It explains AG-UI as a protocol for streaming typed events between frontend and agent backend via Server-Sent Events, with CopilotKit as the runtime. The focus is on deploying the agent to Microsoft Foundry (Azure AI Foundry) for production use, leveraging its hosted agents feature. Foundry provides managed HTTPS endpoints, Entra ID authentication, auto-injection of environment variables, telemetry via Application Insights, and autoscaling. The article details the two protocols (Responses and Invocations) for interacting with hosted agents, and notes the new MicroVM-based architecture announced in April 2026. It is a technical tutorial for developers deploying AI agents.