AI Engineering for .NET: Agents + RAG

Between the launch of ChatGPT and the GA release of the Microsoft Agent Framework in April 2026, a new class of .NET applications emerged: programs that hold a language model in one hand and a set of tools in the other, and decide for themselves how to string them together to reach a goal. The framework is ready. The guidance mostly isn't — it lives scattered across API docs, conference talks, and GitHub samples pinned to preview builds that broke last Tuesday.

Building AI Agents with C# and .NET 10 is the picture on the puzzle box. Nineteen chapters take you from your first LLM call in C# to a production-hosted, observability-instrumented, safety-filtered multi-agent system, all on the stable, generally available surface of Microsoft Agent Framework 1.14, .NET 10 LTS, and C# 14.

You will learn how to:

• Build your first ChatClientAgent, give it a persona, and expose it over HTTP with Minimal API
• Solve the amnesia problem with sessions, history providers, and context providers
• Get typed C# records back from the model instead of strings you have to regex
• Give agents tools and function calling, with human-in-the-loop approval for the dangerous ones
• Build and consume Model Context Protocol (MCP) servers in C#
• Ground answers in your own documents with retrieval-augmented generation
• Send images, audio, and video to multimodal agents — and treat them as untrusted input
• Coordinate specialists with multi-agent orchestration and typed workflow graphs
• Host agents with ASP.NET Core and .NET Aspire, and instrument them with OpenTelemetry
• Layer guardrails against the OWASP LLM Top 10, from input sanitization to critic agents
• Package domain expertise as Agent Skills and measure quality with local and cloud evaluation

Along the way you build the Contoso FAQ agent, a small but real production system that grows chapter by chapter until it ships: hosted, observable, guarded, and evaluated.

What makes this book different is discipline. Every listing compiles and runs, and every one is paired with its expected console output, so you know what success looks like before you press F5. Chapters follow an Explain–Demonstrate–Practice rhythm. Exercises come in three tiers (Basic, Intermediate, Challenge) with full worked solutions in Appendix E. The companion repository — one solution, 120 runnable projects — works against Azure OpenAI, OpenAI, GitHub Models, or a local Ollama model with a single configuration change, and ships dedicated helpers for Anthropic and Microsoft Foundry.

Written for intermediate C# developers. No AI or machine learning background required: if you can write a class, use LINQ, and wire up dependency injection, you're ready. Coming from Semantic Kernel or AutoGen? Appendices B and C are your migration maps.

Buying for a team? Volume licenses are available from 3 copies, with team pricing up to 40% off for larger groups — one purchase, one invoice, a copy for everyone.

The companion code lives at https://github.com/RachidD68/building-ai-agents-with-csharp-and-dotnet-10

Production-Grade RAG with C# and .NET is the complete guide to building Retrieval-Augmented Generation systems on the Microsoft stack — from an 80-line Hello World to a fully deployed, audited, EU AI Act-aware Azure pipeline.

Most RAG material is written for Python. This book is written for you: a .NET developer with a few years of C# experience who wants to ship something real with your company's data. Across seven parts, twenty-five chapters and 668 pages, you'll build Contoso SmartDocs, a running enterprise project that grows chapter by chapter into a production system spanning six document silos — powered by the Microsoft Agent Framework 1.14, Microsoft.Extensions.AI, Azure AI Search, Qdrant, Neo4j, and .NET 10 / C# 14.

You'll learn to:

• Build the core pipeline component by component: embeddings, chunking with Anthropic's Contextual Retrieval, multimodal content, vector databases, indexing strategies, hybrid retrieval with RRF, reranking, and first-class SSE streaming
• Add query intelligence: metadata filters, query construction, routing, and conversational multi-turn rewriting
• Go beyond vectors with graph databases, hybrid storage, GraphRAG, LazyGraphRAG, and Vectorless RAG
• Serve retrieval as a tool over the Model Context Protocol and orchestrate multi-agent RAG workflows with agentic memory
• Ship it for real: a 100-query golden eval gate on every pull request, latency and cost optimization, drift and model migration, prompt-injection defense, grounded citations, EU AI Act audit logging, and a full Bicep-deployed Azure capstone

Every chapter ships runnable code in the companion repository. Every snippet traces to a specific file. Every sample compiles; all 458 tests pass on a clean clone. The architecture is deliberately clean — small, named interfaces with swappable implementations — so what you learn maps directly onto the system you need to build, not just the one in the book.

Nine appendices round it out: worked exercise solutions, a design-pattern flashcard set, a vector database comparison, embedding benchmarks, a Python-to-.NET Rosetta Stone for readers arriving from LangChain or LlamaIndex, a debugging decision tree, prompt patterns, and the math behind RAG.

This is the production-concerns RAG book the .NET ecosystem has been missing. If you can write C#, you can ship RAG.