a hands-on course

Build an
AI Agent

(AI Agents)

Give Claude a loop and real tools, and watch it plan, act and finish a task on its own — in Python, in about half an hour.

A free, hands-on tutorial: turn a chatbot into an AI agent. Wrap the model in a loop, give it real Python tools, and let it plan, act and observe until a task is done — first by hand to see the mechanism, then with the SDK's tool runner. The capstone of the build track and the direct sequel to the Claude API app course. (Needs an Anthropic API key and spends a few cents — an agent makes several calls per task.)

in the real world

Agents are the frontier of applied AI: coding assistants, research helpers, customer-support bots and computer-use tools are all this same loop with better tools. Build it once here and every 'AI agent' headline stops being magic. (Heads up: this one needs an Anthropic API key and costs a few cents to run.)

Start with the agent loop →

Part 1 — From a tool call to an agent

1

What makes it an agent

An agent is a model in a loop: plan → act → observe → repeat, until the job is done. The mental model, and your setup.

2

Build the agent loop

Write the while-loop by hand: call the model, run whatever tools it asks for, feed results back, repeat until it's done.

Part 2 — Give it real tools

3

Give it real tools

Hand it more than one tool — read a file, do math, look something up — and let it pick and even call several at once.

4

Let the SDK drive the loop

Now that you've built the loop yourself, swap it for the SDK's tool runner — decorate your functions and it runs the whole cycle.

Part 3 — Make it trustworthy

5

Keep it on track

The judgment that separates a toy from a useful agent: gate risky actions, handle tool errors, and cap the loop.

6

Cost, models & where next

An agent makes many calls, so cost and model choice matter more. Track spend, tune effort — and see the four ways to build agents.