GitHub radar
Understand how a coding AI agent works, then build your own
A free 17-lesson Python course that rebuilds a coding AI assistant's inner workings step by step — finish it and you'll know why your agent does what it does, and how to add permissions, memory, and background tasks yourself.
A set of 17 short Python lessons that gradually assemble the inner workings of a coding AI agent like Claude Code — from a simple tool-calling loop to memory and multi-agent teamwork.
When it helps
Useful if you want to understand agent mechanics and build your own deliberately instead of guessing why it misbehaves; skip it if you just want to use Claude Code as a ready-made chat tool.
Pros
- Explains permissions, hooks (automatic rules), and agent memory hands-on
- Shows how subagents — an agent's team of helpers — work
- Each lesson is a separate working piece of code to tinker with
- 79k GitHub stars — vetted by a large community
Cons
- It's a coding course, not a ready app — you must install Python
- Every lesson spends real tokens from your paid Anthropic API key
- You must go in order from lesson 1, can't skip to the middle
How to set it up — step by step
- 1Install Python — the course lessons need it to run.
- 2Open your AI agent (Claude Code or Codex) in your project folder.
- 3Send the agent the text from the block below.
- 4Give the agent your Anthropic API key if it asks — billing is pay-per-use.
- 5Check that the agent shows the result of the first lesson, s01_agent_loop.
- 6Then have the agent work through the lessons in order up to s17.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the Learn Claude Code course from https://github.com/shareAI-lab/learn-claude-code — install dependencies per the README, use my Anthropic API key (ask me for it) and run through the first lesson s01_agent_loop, show me the result.
Runs on a regular laptop — just Python and your own Anthropic API key (calls are billed through that key).
Open on GitHub▌ More finds