# Day 2 — The Anthropic API Track

> **Summary — what this page covers**
> The landing page for Day 2. It sets the goal, shows the full schedule, explains how Day 2
> extends the BookTracker app from Day 1 with AI-native features, and points readers to the
> setup page. Keep it tight and exciting — Day 2 is where the C# developer brain lights up.

***Day goal:*** *participants leave Day 2 able to architect, implement, and deploy AI-native features — streaming, tool-calling agents, RAG pipelines, and MCP servers — all in idiomatic, production-quality C#.*

## Day at a glance

| Time | Session | Format |
|---|---|---|
| 9:00 – 10:30 AM | **Section 1:** Official Anthropic C# SDK Fundamentals | Lecture + Demo + Lab |
| 10:30 – 10:45 AM | Morning Break | — |
| 10:45 AM – 12:00 PM | **Section 2:** Streaming, Tool Calling & Extended Thinking | Lecture + Demo + Lab |
| 12:00 – 1:00 PM | Lunch | — |
| 1:00 – 2:15 PM | **Section 3:** RAG Pipelines in .NET | Lecture + Demo + Lab |
| 2:15 – 3:15 PM | **Section 4:** Building MCP Servers in C# | Lecture + Demo + Lab |
| 3:15 – 3:30 PM | Afternoon Break | — |
| 3:30 – 4:15 PM | **Section 5:** AI-Powered Testing + CI/CD | Lecture + Demo + Lab |
| 4:15 – 5:00 PM | **Section 6:** Responsible AI + Wrap-Up | Discussion + Q&A |

## What you'll be able to do (learning outcomes)

By the end of Day 2, you will be able to:

| Outcome | Where it's delivered |
|---|---|
| **Integrate the official Anthropic C# SDK** into ASP.NET Core with correct **DI patterns** | [Section 1](03-section-1-sdk.md) · [Lab 1](04-lab-1-chat.md) |
| **Build streaming chat interfaces** using **Server-Sent Events** in ASP.NET Core | [Section 2](05-section-2-streaming-tools.md) · [Lab 2](06-lab-2-agent.md) |
| **Implement the tool-calling agent loop** in C# for multi-step AI workflows | [Section 2](05-section-2-streaming-tools.md) · [Lab 2](06-lab-2-agent.md) |
| **Build RAG pipelines in .NET** — chunking, embedding, vector storage, retrieval | [Section 3](07-section-3-rag.md) · [Lab 3](08-lab-3-rag.md) |
| **Create custom MCP servers in C#** to connect Claude to your own **databases and internal services** | [Section 4](09-section-4-mcp.md) · [Lab 4](10-lab-4-mcp.md) |
| **Use Claude Code in GitHub Actions** for automated **code review and release notes** | [Section 5](11-section-5-testing-cicd.md) · [Lab 5](12-lab-5-cicd.md) |
| **Apply responsible AI practices** — prompt-injection defense, token budgets, hallucination mitigation | [Section 6](13-section-6-responsible-ai.md) |

> These build on Day 1's foundation: the **SDK** uses the LLM mental model from Day 1 Section 1;
> **MCP servers** (Section 4) are the "build" counterpart to Day 1's "use the GitHub MCP server";
> and **CI review** (Section 5) extends Day 1's hooks/guardrails into the pipeline.

## What you build today

Day 2 takes the **same BookTracker API you already understand** from Day 1 and adds AI-native
capabilities, one section at a time — no new domain to learn, just new powers bolted onto familiar
code. By 5:00 PM the API has a **streaming chat assistant**, a **tool-calling agent** that queries
the real database, a **RAG recommendation engine** grounded in book data, a **C# MCP server**
callable from Claude Code, an **AI-generated test suite**, and a **CI/CD pipeline** with automated
AI code review. See [BookTracker — AI Extensions](02-booktracker-ai-extensions.md) for the full map
of what gets added where.

## How to use these docs

- [Before You Begin](01-before-you-begin.md) — **API access differs from Day 1** (an API key is
  required; Pro/Max does NOT grant API access). Confirm before 9:00 AM.
- Each **Section** page has the concepts, demos, and key takeaways.
- Each **Lab** page is the hands-on build with a committed deliverable.
- The [Models & Cost Reference](14-models-cost-reference.md) is a keeper — pricing and the
  free/local alternatives.