Agentic AI Development with Agent Framework, MCP and .NET

This repository contains the source code used in my Udemy course:

Agentic AI Development with Agent Framework, MCP and .NET
https://www.udemy.com/course/agent-development-microsoft-agent-framework-mcp-and-net/?couponCode=LAUNCH

Welcome to the definitive guide on building production-ready agentic AI systems in the .NET ecosystem. This course goes beyond theory and focuses on hands-on development of autonomous multi-agent orchestration for enterprise applications.

Using Microsoft Agent Framework, Microsoft Foundry, the Model Context Protocol (MCP), .NET Aspire, AG-UI, DevUI, and .NET, you will learn how to build robust AI workflows that solve complex business problems.

---

What You Will Master

• Microsoft Agent Framework (MAF): Build sophisticated, stateful AI systems with Microsoft Foundry and Azure OpenAI.
• Multi-Agent Orchestration: Implement Sequential, Concurrent, Handoff, and Group Chat patterns.
• Agentic RAG Systems: Build intent-based retrieval using Qdrant vector databases and TextSearchProvider.
• Protocols & Interoperability: Implement A2A communication, MCP tool exposure, and AG-UI frontend streaming.
• Observability & Visual Testing: Track JSON payloads, token usage, handoff latencies, and tool calls with Aspire + DevUI.
• Enterprise Microservices Integration: Integrate AI agents into MinimalAgent-style microservices and existing APIs.

---

Course Roadmap

Part 1: Core Agent Development

Learn core agent anatomy, Azure OpenAI connectivity, invocation lifecycle, custom function tools, and persistent conversational context with AgentSession.

Part 2: Orchestrating Multi-Agent Systems

Design and implement specialized swarms (triage, finance, compliance) with AgentWorkflowBuilder and production workflow topologies.

Part 3: Advanced Reasoning with Agentic RAG

Build systems where agents decide when retrieval is needed. Integrate Qdrant, embeddings, and semantic search into enterprise-ready AI flows.

Part 4: Agent Communications and Protocols

Expose agents through Web APIs for A2A architectures, connect to local/hosted MCP servers, and stream generative UI experiences with AG-UI.

---

Repository Structure (src)

The repository is organized by section, matching the course progression:

• section4-setup
  • HelloAgent
• section5-getting-started
  • BasicAgentApp, MultiTurn, Streaming, StructuredOutput
  • MinimalAgent (Aspire-style AppHost/WebApi/ServiceDefaults)
• section6-tool-use
  • FunctionCall, CodeInterpreter, ApproveRequiredFunc
  • MinimalAgentWithTools (AppHost/WebApi/ServiceDefaults)
• section7-memory
  • CustomContextProvider, MockCosmosDb
• section8-workflows
  • FirstWorkflow, AgentsWorkflow
• section9-workflow-patterns
  • AgenticWorkflowPatterns
  • MinimalAgentWithWorkflows (AppHost/WebApi/ServiceDefaults)
• section11-agentic-rag
  • BasicTextRAGExample, QdrantVectorStore
• section13-a2a
  • EnterpriseComplianceService
• section14-mcp
  • HostedMcpGovernanceExample, LocalGitHubMcpExample
• section15-ag-ui
  • Server, Client

Solution file:

• src/develop-agents.slnx

---

Technology Stack

• Languages & Frameworks: .NET 10, C#, ASP.NET Core, Blazor Server
• AI & Agents: Microsoft Agent Framework, Azure OpenAI (gpt-5-mini, text-embedding-3-small)
• Cloud & Deployment: Microsoft AI Foundry
• Frontend Protocol: AG-UI
• Orchestration: .NET Aspire
• Observability: Aspire OpenTelemetry, Application Insights
• Vector Data & Storage: Qdrant
• Architecture: Microservices, Clean Architecture, MCP

---

Who This Course Is For

• Software Developers, Architects, and Tech Leads in enterprise environments transitioning into AI Engineering.
• Professionals integrating AI agents into backend microservices, legacy systems, and APIs.
• C#/.NET developers who want to build agentic AI systems without switching to Python.