Getting Started: Microsoft Agent Framework in .NET

Integrating language models into .NET applications is often treated as a simple API integration, but building reliable, autonomous agents requires managing complex state, conversation history, and tool execution. This course teaches you how to build these systems from the ground up using the Microsoft Agent Framework. You will start with the core concepts of LLMs and agents before writing your first implementation, configuring local language models, and testing with the DevUI. From there, we cover how to handle multi-turn conversations, manage context limits with chat reduction, and persist agent sessions locally and to Redis.

Once the fundamentals are in place, you will enhance your agents by building custom tools, handling complex parameters with validation, and implementing human-in-the-loop interactions. We then explore how to enforce structured outputs and manage prompt templates to ensure reliable responses. Because real-world agent systems require complex coordination, you will learn how to orchestrate sequential, concurrent, and conditionally routed workflows, alongside managing workflow state and custom events. Finally, we focus on making your agents production-ready. You will trace execution with OpenTelemetry, monitor token usage and costs, and secure your application against prompt injection while filtering outputs for content safety, culminating in a complete, hardened agent system design.