Welcome: What You'll Be Able to Do

Not long ago, "AI agent" mostly meant a demo that worked once on stage. Today, agents write production code, research topics while you sleep, reconcile spreadsheets, triage inboxes, and run multi-step workflows while their owners do something else. The gap between people who know how to work with agents and people who don't is real — and it has surprisingly little to do with being technical.

This course closes that gap. By the end, you'll be able to hand real work to agents with confidence, check that work quickly, and build your own agents when the off-the-shelf ones aren't enough. This first lesson shows you how the course is laid out and how to get the most out of it.

Who this course is for#

You don't need to be a programmer to start. The course is designed around three kinds of learners, and most people are a mix:

If you've never written code, you'll still get a lot out of the first half of the course, and the coding modules are written so you can follow along by asking an agent to do the typing. That's not a cheat — it's the whole point.

What you'll be able to do#

Here's what "done" looks like. After completing the course you will be able to:

• Explain how agents work — the model, the loop, the tools — well enough to predict when they'll succeed and when they'll struggle.
• Delegate real work to chat and coding agents using task descriptions that agents can actually finish.
• Build software agentically, from an empty folder to a deployed app, reviewing the agent's work instead of typing every line.
• Engineer context: write system prompts, specs, project instruction files, and reusable skills that make agents consistent.
• Connect agents to your tools with the Model Context Protocol (MCP), and build your own MCP server.
• Write an agent from scratch in code, then level up to an agent SDK.
• Coordinate multiple agents for research, coding, and parallel work.
• Measure and improve agents with evals, traces, and a debugging playbook.
• Run agents safely, with sandboxes, least-privilege permissions, and defenses against prompt injection.
• Ship agents to production and turn them into products people pay for.

How the course is structured#

The course has 12 modules and 65 lessons. Each module builds on the last, but every lesson is written to stand on its own so you can jump to what you need.

Modules are numbered from 0, matching the table above:

• Module 0: Start Here — Orientation, setup, and your first agent task.
• Module 1: How Agents Work — The mental model: LLMs, the agent loop, tools, context, and model choice.
• Module 2: Working with Coding Agents — Claude Code, Codex, Cursor and friends, used well.
• Module 3: Prompting & Context Engineering — Giving agents exactly the right information.
• Module 4: Tools & MCP — Connecting agents to the rest of your world.
• Module 5: Build Your First Agent — Writing the loop yourself in TypeScript (with Python equivalents).
• Module 6: Multi-Agent Systems — Orchestrators, subagents, and parallel work.
• Module 7: Agents for Everyday Work — Research, automation, data analysis, writing, scheduled agents.
• Module 8: Evals & Reliability — Proving your agent works.
• Module 9: Safety & Security — Powerful agents without nasty surprises.
• Module 10: Shipping to Production — Architecture, cost, deployment, monitoring, products.
• Module 11: Capstone Projects — Four complete agents built end to end.

Lessons take 5–17 minutes to read. Most include an exercise you can finish in under 30 minutes.

How to get the most out of each lesson#

Reading about agents is like reading about swimming. The understanding only clicks when you get in the water. Every lesson has three kinds of material:

• Concepts explain why something works, so you can adapt when tools change.
• Prompts in highlighted cards are ready to copy and paste. Use them verbatim first, then make them your own.
• Exercises turn the lesson into a skill. Don't skip them — even a five-minute attempt beats a careful re-read.

Here's the first prompt of the course. Paste it into any AI assistant to turn it into a study partner for this course:

I'm taking a course on using and building AI agents. As I go, I'll paste in
concepts I'm learning. For each one:
1. Explain it back to me in one paragraph, as simply as possible.
2. Give me one concrete example from my own work. (I work as: [your role].)
3. Ask me one question that checks whether I really understood it.
Keep your answers short. Start by asking what I'm working on this week.

A note on tools changing fast#

Agent tools release new features constantly. This course is built around principles that have stayed stable while the tools churned: clear delegation, good context, well-designed tools, verification, and safety. When we show a specific product, we focus on the parts that transfer.

When something in a lesson doesn't match what you see on screen, that's normal. Ask your agent: "This tutorial says X, but I see Y — what changed and what's the equivalent now?" Working through small mismatches like that is itself an agent skill.