Welcome to the course

You opened your terminal, installed an SDK, and thought: I'll figure out the rest as I go. That works until you need retries, evals, tool calls, a RAG pipeline, and an agent loop that doesn't melt down in production. This course exists so you don't have to figure all of that out the hard way.

This is a hands-on, code-first course for engineers building real applications on the Claude API. No decks full of marketing diagrams. No "imagine a future where..." No hand-holding through basic programming. You write code, you run it against the live API, you see the response, and you move on.

What you're going to build

Over thirteen modules you'll go from your first API call to a working agent that uses tools, retrieves documents, and makes decisions on its own. Along the way you'll build and break enough small programs to develop the thing the rest of this course is really about: the intuition for what Claude is good at and how to ask it for that thing under load.

The arc looks like this:

1. Accessing the API — authentication, the Messages API, multi-turn conversations, system prompts, temperature, streaming, structured data.
2. Prompt evaluation — eval workflows, test datasets, model-based grading, code-based grading. How you know a prompt is actually good.
3. Prompt engineering — clarity, specificity, XML structure, few-shot examples. The techniques that move outputs from "demo" to "ship."
4. Tool use — letting Claude call your functions. Schemas, multi-turn tool loops, multiple tools, fine-grained control, first-party tools.
5. Retrieval-augmented generation — chunking, embeddings, BM25, and a hybrid multi-index pipeline.
6. Advanced features — extended thinking, images, PDFs, citations, prompt caching, code execution, the Files API.
7. Model Context Protocol — building MCP servers that expose tools, resources, and prompts to any client.
8. Claude Code — the CLI, how to drive it, and how to extend it with your own MCP servers.
9. Agents and workflows — parallelization, chaining, routing, environment inspection, and the line between a workflow and a true agent.

How this course works

Every module is built around code you run. Every concept lands in a working example you can modify. Every exercise is a concrete thing to build, not a reflection prompt.

Assume fluency. You already know Python, you already know how to read a stack trace, you already know what an environment variable is. This course respects your time and spends none of it explaining any of that.

Expect to run the code. Reading the lessons gets you maybe a third of the way. The other two-thirds come from typing it out, breaking it, and fixing it.

By the end of Module 13 you should be able to sit down with a blank file, an API key, and a real product requirement, and know exactly how to start.