Claude Code 101 Is Free. Here's What It Doesn't Cover.

What does Claude Code 101 cover?

Five sections on Anthropic's course portal: what Claude Code is and how it works, installing it and writing your first prompt, daily workflows (the explore, plan, code, commit loop, plus context management and code review), and a customization section with dedicated lessons on CLAUDE.md (the instruction file the agent reads at startup), subagents (helper agents a main agent can hand work to), skills (packaged instructions for repeatable tasks), MCP (a standard for connecting the agent to outside tools), and hooks (scripts that run automatically on events, like session start). Then a quiz. Anthropic pitches it at new developers and at experienced engineers who have not used an AI coding agent.

So the building blocks are all introduced. This matters, because most criticism of free vendor courses claims they skip the good parts. This one does not.

What about the rest of the free lineup?

It goes further than most people realize. Claude Code in Action, the follow-up course, covers long hands-off sessions: steering, permission modes, routines, headless runs (the agent working unattended, with no terminal window open to watch), and code review in GitHub Actions. The catalog holds more than twenty free courses, including separate introductions to MCP, subagents, and agent skills. If your question is "how do I use this tool well," Anthropic has answered it for free, and no paid course should pretend otherwise. If the question is which course to take at all, free or paid, our eight-course comparison covers the field.

So what is missing?

The system. Every course in the free lineup teaches the tool inside a developer's workflow: you open it, direct it, ship code, close it. None of them has a lesson dedicated to the thing you can build around the tool and keep:

• Memory that survives the session. There is no lesson dedicated to an assistant that remembers yesterday, this month, or your preferences, and none on retrieval (letting the agent search its own accumulated notes).
• An assistant with an identity. A CLAUDE.md lesson teaches the file. There is no lesson dedicated to a named assistant with its own memory, routines, and daily briefing that exists whether or not you opened the terminal today.
• A factory. There is no lesson dedicated to a standing roster of agents with an orchestrator (the agent that assigns work to the others), a task board, and shared memory, working through a queue while you are elsewhere.

That is not a criticism. Anthropic teaches Claude Code the way a manufacturer documents a power tool. What you build with the tool is your project, not their syllabus.

The tool is not the system

The harness is the machinery around a model: what it can see, touch, and remember, and when it runs. Every course above teaches you to operate inside someone's default harness. Building your own is a different skill, and it is the one that compounds: your memory, your skills, and your workflows survive model changes and tool changes because they are plain files on your machine.

How to use both

Take Claude Code 101. It is free, and the tool fluency transfers directly. New to agents entirely? Read our free primer first: why chat has a ceiling, and how assigning work differs from asking questions. Then, when you want the persistent layer, that is what we teach: Level 1 builds your assistant with memory around Claude Code or OpenCode, about 10 hours, no coding required. Level 2 scales it into the factory. What a dark factory is explains where that ends up.