# Workshop Agenda ## AI-Powered .NET Development with Claude & Claude Code **Presenter:** Woody · [Intelligently.dev](https://intelligently.dev) · **Event:** Visual Studio LIVE! · **Dates:** July 9–10, 2026 · **Hours:** 9:00 AM – 5:00 PM each day · **Venue/Room:** [add] · **Registration:** [add link] A hands-on, two-day workshop for .NET developers. **Day 1** makes Claude Code a daily driver for real C# work. **Day 2** uses the Anthropic C# SDK to build AI-native features — streaming, agents, RAG, and MCP servers — in production-quality ASP.NET Core. Every lab builds on one evolving sample application, **BookTracker**, so the skills compound across the two days. > **Before you arrive — access matters and the two days differ.** > **Day 1 (Claude Code)** requires a **paid Claude plan (Pro minimum)** — Claude Code is not on the > Free tier. **Day 2 (the API)** requires a separate **Anthropic API key** (pay-as-you-go from the > Console); a Pro/Max subscription does **not** grant API access. You'll also need the **.NET 10 > SDK**, **Node.js 18+**, **Git**, **Docker Desktop** (Day 2), and **WSL2** on Windows. Full setup > on each day's *Before You Begin* page. A free/local path (Ollama) is provided for Day 2's > embeddings so no OpenAI key is required. --- ## Two days at a glance | | Day 1 — Claude Code Track | Day 2 — Anthropic API Track | | --------------------- | -------------------------------------------------------- | --------------------------------------------------- | | **Theme** | Operate Claude Code as a daily driver | Build AI-native features in C# | | **You leave able to** | Steer, explore, build, refactor, and debug with an agent | Ship streaming, tool-calling, RAG, and MCP features | | **Sample app** | Configure & improve BookTracker | Extend BookTracker with AI capabilities | --- ## Day 1 — The Claude Code Track *Tuesday, July 9 · 9:00 AM – 5:00 PM* → details: [Day 1]() | Time | Session | Focus | Hands-on lab | | ------------- | ----------------------------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | | 9:00 – 10:30 | **Section 1 — Foundations + Claude Code CLI** | How LLMs work (tokens, context, grounding) → integration decisions; install & authenticate; operate from the terminal; first CLAUDE.md; codebase exploration | **Lab 1:** working session + committed CLAUDE.md | | 10:30 – 10:45 | *Morning break* | | | | 10:45 – 12:30 | **Section 2 — Steering Claude Code** | The steering decision (when it loads · compaction · context cost · authority); CLAUDE.md, **Rules**, **Skills** (folders), **Subagents** (markdown), **Hooks**, plugins | **Lab 2:** a `.claude/` setup — rules + skill folder + guardrail hook + plugin | | 12:30 – 1:30 | *Lunch* (Lab 2 continues, self-paced) | | | | 1:30 – 3:15 | **Section 3 — IDE Integration + MCP Ecosystem** | Operate Claude Code in **VS Code** and **JetBrains/Rider**; session management; MCP architecture; the GitHub MCP server | **Lab 3:** GitHub MCP session — query the repo and create an issue | | 3:15 – 3:30 | *Afternoon break* | | | | 3:30 – 5:00 | **Section 4 — AI in the SDLC** | Specification-first development; refactoring at scale with approval gates; testing; **debugging** from stack traces; architecture analysis and ADRs | **Lab 4:** spec → feature → tests → debug, committed | **Day 1 outcomes** — you will be able to: - **Install, configure, and operate Claude Code** from the terminal, VS Code, and JetBrains (Rider/IntelliJ) - **Use Claude Code to accelerate** architecture design, code construction, refactoring, testing, and debugging - **Explain how LLMs work** — tokens, context windows, turns, grounding — and apply that to practical integration decisions --- ## Day 2 — The Anthropic API Track *Wednesday, July 10 · 9:00 AM – 5:00 PM* → details: [Day 2]() | Time | Session | Focus | Hands-on lab | | ------------- | ----------------------------------------------------------- | ------------------------------------------------------------ | ------------------------------------------------------------ | | 9:00 – 10:30 | **Section 1 — Official Anthropic C# SDK Fundamentals** | The official SDK with correct DI patterns; `AnthropicClient` as a singleton; multi-turn history; prompt caching; Polly retries | **Lab 1:** `/api/chat` with history and prompt caching | | 10:30 – 10:45 | *Morning break* | | | | 10:45 – 12:00 | **Section 2 — Streaming, Tool Calling & Extended Thinking** | Streaming chat with **Server-Sent Events** in ASP.NET Core; defining tools; the **tool-calling agent loop** for multi-step workflows; extended thinking | **Lab 2:** streaming endpoint + multi-tool agent | | 12:00 – 1:00 | *Lunch* | | | | 1:00 – 2:15 | **Section 3 — RAG Pipelines in .NET** | Chunking, embeddings (OpenAI **or free/local Ollama**), Qdrant vector storage, cosine retrieval, augmented prompts | **Lab 3:** `/api/recommend` RAG engine with grounded answers | | 2:15 – 3:15 | **Section 4 — Building MCP Servers in C#** | The ModelContextProtocol C# SDK; typed tools; connecting Claude to your **database and internal services** via DI; Streamable HTTP; testing from Claude Code | **Lab 4:** `BookTracker.Mcp` with 3+ tools, called from Claude Code | | 3:15 – 3:30 | *Afternoon break* | | | | 3:30 – 4:15 | **Section 5 — AI-Powered Testing + CI/CD** | AI-generated xUnit suites; edge-case discovery; TestContainers; **Claude Code in GitHub Actions** for automated **code review and release notes** | **Lab 5:** AI tests + CI review + release-notes workflow | | 4:15 – 5:00 | **Section 6 — Responsible AI + Wrap-Up** | **Prompt-injection defense**, **token budgets**, **hallucination mitigation**, enterprise governance, and the path to production | Optional closing security audit | **Day 2 outcomes** — you will be able to: - **Integrate the official Anthropic C# SDK** into ASP.NET Core with correct DI patterns - **Build streaming chat interfaces** using Server-Sent Events in ASP.NET Core - **Implement the tool-calling agent loop** in C# for multi-step AI workflows - **Build RAG pipelines in .NET** with chunking, embedding, vector storage, and retrieval - **Create custom MCP servers in C#** to connect Claude to your own databases and internal services - **Use Claude Code in GitHub Actions** for automated code review and release notes - **Apply responsible AI practices** — prompt-injection defense, token budgets, and hallucination mitigation --- ## How the two days connect The agent skills from Day 1 set up the API work on Day 2: the **LLM mental model** (Day 1, Section 1) underpins the **SDK** decisions; **using** the GitHub MCP server (Day 1, Section 3) sets up **building** your own MCP server (Day 2, Section 4); and Day 1's **hooks and guardrails** (Section 2) extend into Day 2's **CI review** and **responsible-AI** controls. Bring your own project — the last exercise of Day 1 points Claude Code at real code, and everything on Day 2 transfers directly to it. *New here? Start with [Day 1 → Before You Begin]() and, for Day 2, [Before You Begin]().*