Appearance
Day 1: The Agent Loop
Before you build anything, you need to understand the one mechanical feature that separates an agent from a chatbot: the loop. Act, observe, decide, repeat. Everything else in this course — tools, memory, security, deployment — is infrastructure in service of making that loop reliable.
Today you will build a working agent from about twenty lines of Python, set up Claude Code as your development toolkit, and learn the prompting and context-window fundamentals that determine whether an agent produces consistent output or creative surprises. Three lessons, each building on the last.
What you will cover
- What Makes an Agent — the one mechanical difference between a chatbot and an agent, built in twenty lines of code.
- Your Development Toolkit — Claude Code setup, CLAUDE.md for persistent project memory, custom commands, file injection, auto-approve mode, and git as your undo button.
- Prompting & Context Engineering — the craft underneath every good agent: system instructions, context windows, caching, and getting consistent output.
- Model Parameters & Tuning — temperature, top-p, and the handful of settings that actually change behavior.
By the end of today, your agent runs, your tools are configured, and you understand what is happening each time the model decides to take another step.