Microsoft Agent Framework: Build Multi-Agent AI Systems

“This course contains the use of artificial intelligence.”

Move beyond chatbot demos and build an AI agent that performs a complete business workflow.

Microsoft Agent Framework gives developers a foundation for creating tool-using agents and controlled agentic workflows. In this hands-on course, you will learn the framework by building one application from beginning to end: an AI-powered dentist appointment system.

You begin with a single agent and controlled business tools. You then develop appointment search and booking capabilities, introduce specialized agent responsibilities, orchestrate a multi-step workflow, maintain structured state, and require human confirmation before consequential actions.

The course also examines AutoGen migration so developers working with earlier Microsoft agent technologies can plan their transition systematically.

Who this course is for

• Python or .NET developers entering agentic AI development

• AI engineers who want to move beyond chatbot prototypes

• Azure developers building agent-enabled applications

• Software architects evaluating Microsoft Agent Framework

• AutoGen developers planning future agent projects or migrations

What you will learn

• Explain Microsoft Agent Framework's core agent architecture

• Create and configure a tool-using AI agent

• Design typed tools around deterministic business services

• Build an AI-powered appointment scheduling application

• Prevent the model from inventing business-system data

• Maintain structured state across a multi-step process

• Separate responsibilities across specialized agents

• Add human approval before consequential operations

• Design and test failure paths and controlled execution

• Plan an AutoGen-to-Agent-Framework migration

Requirements

Basic programming experience is required. Students should understand functions, APIs, JSON and asynchronous application concepts. Familiarity with Python or .NET is recommended.

You will also require access to a model/provider currently supported by the Microsoft Agent Framework configuration used during the course. Never commit API keys or credentials to the project repository.

No machine-learning mathematics or model-training experience is required.

Final project

You will build a portfolio-ready intelligent dentist appointment system. The application will process a patient's natural-language request, gather required information, search deterministic appointment data through tools, coordinate specialized agent responsibilities, request confirmation, create the appointment and return a patient-friendly confirmation.