36 SESSIONS · 6 PHASES · C# / .NET
Building AI Agents with Microsoft Foundry and Microsoft Agent Framework
You’ve used ChatGPT or Claude. Now it’s time to build with AI. This hands-on course takes developers and architects from zero AI development experience to designing, building, and deploying production-ready AI agents.
👩💻
Who It’s For
Developers starting to build AI-powered apps and agents, and architects designing agentic solutions. No prior AI or machine learning experience needed. If you’ve used a chat assistant and can write basic C#, you’re ready.
🧰
What You Need
An Azure subscription, VS Code with the C# Dev Kit and Foundry Toolkit, the .NET 10 SDK, and basic C#. All demos are in C#; a Python edition may follow based on demand.
🚀
What You’ll Build
One project that grows all course long: a console agent that gains tools, memory, and knowledge, joins a multi-agent system, gets a web front end, and ships to production on Microsoft Foundry.
🛠️ Technologies Covered
Microsoft Foundry
Foundry Agent Service
Microsoft Agent Framework
Microsoft.Extensions.AI
DevUI
MCP
A2A
AG-UI
Azure AI Search
ASP.NET Core
Blazor
.NET Aspire
OpenTelemetry
Application Insights
📚 Course Curriculum
36 one-hour sessions across six phases, from your first prompt to production.
PHASE 1 · S1–5
AI Foundations
PHASE 2 · S6–10
Core Building Blocks
PHASE 3 · S11–22
Microsoft Agent Framework
PHASE 4 · S23–25
Multi-Agent Systems
PHASE 5 · S26–29
Agents in Real Applications
PHASE 6 · S30–36
Production Readiness
PHASE 1
AI Foundations
Understand what happens behind the chat box and write your first AI-powered code.
| # | Session | What You’ll Learn |
|---|
| 1 | From AI User to AI Builder | How chat assistants really work, the difference between a chatbot and an agent, and the course roadmap |
| 2 | How LLMs Work (for Developers) | Tokens, context windows, temperature, hallucinations, and cost |
| 3 | Prompt Engineering for Developers | System prompts, few-shot examples, and writing reliable instructions |
| 4 | Tour of Microsoft Foundry | Projects, the model catalog, the playground, deploying a model, and setting up your environment |
| 5 | First AI Code with Microsoft.Extensions.AI | Calling a model from C#, streaming, structured output, and secure authentication with Entra ID |
PHASE 2
Core Building Blocks
Learn the concepts every agent depends on and build your first agent without code.
| # | Session | What You’ll Learn |
|---|
| 6 | Embeddings and Vector Search | What embeddings are and how similarity search works |
| 7 | Classic RAG with Azure AI Search | Grounding AI answers in your own data |
| 8 | Choosing Models | Model families, the model router, and cost, latency, and quality trade-offs |
| 9 | What Makes an Agent | The reasoning loop, tools, memory, and when not to use an agent |
| 10 | No-Code Agents in Foundry Agent Service | Building a working agent in the Foundry portal with built-in tools |
PHASE 3
Microsoft Agent Framework
Start with simple console apps so you can see exactly what an agent does, then move the same agents into DevUI for visual testing and debugging.
| # | Session | What You’ll Learn |
|---|
| 11 | Your First Agent (Console) | Why Agent Framework, how it builds on Microsoft.Extensions.AI, and running your first agent |
| 12 | Conversations in the Console | Multi-turn conversations with sessions, streaming, and structured output |
| 13 | Function Tools (Console) | Giving agents your own C# methods and watching tool calls happen |
| 14 | Approval-Required Tools (Console) | Human-in-the-loop approval before an agent takes a risky action |
| 15 | Introducing DevUI | Moving your console agents into DevUI and inspecting tool calls and traces visually |
| 16 | Memory and Persistence | Storing conversations and giving agents memory across sessions |
| 17 | Context Providers and Agentic RAG | Letting the agent decide when and how to search your knowledge |
| 18 | Consuming MCP Servers and Foundry Toolboxes | Connecting agents to external tools through the Model Context Protocol |
| 19 | Building Your Own MCP Server in C# | Exposing your company APIs as tools any agent can use |
| 20 | Middleware and Skills | Logging, guardrails, and reusable skills for your agents |
| 21 | Workflows Fundamentals | Building sequential and conditional workflows and visualizing them in DevUI |
| 22 | Advanced Workflows | Checkpointing, human-in-the-loop steps, and recovering from failures |
PHASE 4
Multi-Agent Systems
Make specialized agents work together to solve bigger problems.
| # | Session | What You’ll Learn |
|---|
| 23 | Orchestration Patterns I | Sequential and concurrent multi-agent patterns |
| 24 | Orchestration Patterns II | Handoff, group chat, and Magentic-One patterns |
| 25 | Multi-Agent Lab | Building a triage-and-specialists system end to end |
PHASE 5
Agents in Real Applications
Take agents out of the development environment and into services and user interfaces.
| # | Session | What You’ll Learn |
|---|
| 26 | Agents Behind ASP.NET Core Web APIs | Hosting agents inside your own services |
| 27 | Agent-to-Agent (A2A) | Exposing and calling agents across services |
| 28 | AG-UI Fundamentals | How MCP, A2A, and AG-UI differ, the AG-UI event model, and exposing an agent over AG-UI |
| 29 | AG-UI Front Ends | Building Blazor and React clients, with approvals and shared state in the UI |
PHASE 6
Production Readiness
Evaluate, secure, monitor, and deploy your agents with confidence.
| # | Session | What You’ll Learn |
|---|
| 30 | Evaluating Agents | Measuring quality, groundedness, and safety in Foundry |
| 31 | Observability | Tracing with DevUI, the .NET Aspire dashboard, and Application Insights |
| 32 | Responsible AI and Content Safety | Content filters, groundedness, and transparency |
| 33 | Securing Agents | Prompt injection, tool permissions, identity, and networking |
| 34 | Deploying to Foundry Hosted Agents | Running your agents on managed, scalable infrastructure |
| 35 | CI/CD and Cost Management | Deployment pipelines, evaluations as quality gates, and cost tracking |
| 36 | Capstone and Architecture Review | Completing the full solution and reviewing reference architectures |
