You already build APIs, deploy to Azure, and wire up dependency injection in your sleep. Now it's time to add AI agents to your toolkit.
The Microsoft Agent Framework (MAF) — the unified successor to Semantic Kernel and AutoGen — brings AI agent development into familiar .NET territory. No ML expertise needed. If you know how to build an ASP.NET Core service, you already have the foundation to build production-ready AI agents. The framework uses the patterns you work with daily: DI, middleware pipelines, configuration, and OpenTelemetry — applied to agents that can reason, use tools, and collaborate.
In this session, you will see a demo on how to build an Intelligent Order Fulfillment System — a multi-agent workflow where specialized agents triage orders, check inventory, and route shipments. You'll compose them using MAF's built-in orchestration patterns (sequential, handoff, concurrent) and connect them to business systems through Azure Functions for .NET as remote MCP servers — secured with Entra ID and deployed on Flex Consumption.
What you'll take away:
- Create agents using the AsAIAgent() extension and Microsoft Foundry model endpoints
- Compose multi-agent workflows with streaming, state management, and human-in-the-loop checkpoints
- Expose your existing .NET logic as MCP tools using [McpToolTrigger] on Azure Functions
- Observe agent behavior with OpenTelemetry tracing on the Aspire Dashboard
- Deploy the full stack to Azure with azd up — Entra ID auth and Managed Identity included
