Learn to ship autonomous agents, RAG pipelines, and AI tools using Microsoft.Extensions.AI, Microsoft.Agents.AI, and Model Context Protocol — with 37 runnable code samples.
The Problem
Most AI books teach you OpenAI's API in Python. They skip over:
- How to route between 5 different providers (OpenAI, Azure, GitHub Models, Ollama, Anthropic) with one interface
- How to build a RAG pipeline that doesn't hallucinate on page 50 of a document
- How to deploy autonomous agents that your team can actually maintain
- How to package AI tools as MCP servers and drop them into production
- How to defend against prompt injection, monitor token costs, and test non-deterministic systems
Meanwhile, your .NET team is stuck reinventing these wheels or shipping Python sidecars.
Meet From Problems to Patterns
This is the only book that covers the full Microsoft AI stack in production depth:
- Microsoft.Extensions.AI — One interface across every provider; middleware for caching, logging, cost tracking, and fallback logic
- Microsoft.Agents.AI — Build autonomous agents with persistent sessions, human approval flows, and multi-agent orchestration
- Model Context Protocol (MCP) — Ship AI tools as NuGet packages; secure, auditable, enterprise-ready
- .NET 9 + real C# code — 37 runnable companion projects. Every example builds. Every example ships.
What You'll Build
Real patterns for real problems:
✅ Semantic Search Engine — embeddings, vector similarity, and ranking
✅ Production RAG Pipeline — ingestion, chunking, retrieval, inline citations, LLM-as-judge evaluation
✅ Vision AI Document Processor — read receipts from photos, extract structured data
✅ Autonomous Expense Report Agent — handles approvals, maintains context across sessions
✅ Multi-Agent Workflows — agents coordinating with each other to solve complex problems
✅ MCP Servers as NuGet Tools — deploy to Azure Container Apps, lock down security layers
✅ Fully Offline AI App — Ollama integration, no API keys, no cloud costs, data stays local
What You'll Learn
Six chapters. Six patterns. Production-ready.
- Foundations — What generative AI is, why it's probabilistic, how to reason about it
- Microsoft.Extensions.AI — Streaming, structured output, function calling, prompt engineering, context windows
- RAG End-to-End — The enterprise pattern that powers 90% of production AI (100 pages, every decision explained)
- Microsoft Agent Framework — Sessions, approval workflows, graph execution, A2A communication
- Model Context Protocol — Build servers, choose transports, secure for enterprise
- Production Patterns — OpenTelemetry tracing, Polly resilience, cost-aware routing, responsible AI, testing non-deterministic systems
Plus 4 appendices: package reference, model quick reference, provider support matrix, Extensions.AI API reference.
Who This Is For
You're a .NET developer or solution architect with solid C# experience. You don't need an ML background. You need to ship AI features — now — and they need to hold up in production.
You're tired of:
- Shipping Python sidecars just to call an LLM
- Reinventing RAG chunking strategies
- Wondering how to deploy agents safely
- Learning "AI for Python devs" books that skip the .NET tooling
This book is written for you.
The Stack (Validated May 2026)
- .NET 9
- Microsoft.Extensions.AI 10.5.2
- Microsoft.Agents.AI 1.3
- ModelContextProtocol 1.2
- Real production patterns you can deploy tomorrow
Get It Now
Ready to build AI that ships? Stop waiting for Python docs to translate to C#. Read the book written for .NET teams, by someone who's shipped this stack to production.
