A free Beginner course on Claude Code — the habits, not the commands

Today we are publishing a free course: Claude Code, from zero to daily driver. It lives at /guides/claude-code, it costs nothing, and it takes an afternoon — five core lessons of about twenty minutes each, a capstone that puts them together, and three standalone electives you can take whenever the need arrives. This post is the announcement; the course is the thing.

Why we built it: Claude Code is where our working day actually happens. The site you are reading ships through agent sessions like the ones the course teaches, under a delivery board that reviews every change before it counts as done. When clients ask how their teams should start, “read the manual” is the wrong answer for a tool with this much surface, and most of what passes for a tutorial is a tour of commands. So we wrote the on-ramp we wished existed when we started.

01Commands versus habits

Most tutorials teach what to type. The commands change every month — flags move, defaults flip, new verbs arrive — and a skill that expires with the release notes was never really a skill. What survives is the operating habits, and the course is built around five of them:

• —Know what session you are in. The version, the model, the account — /status shows all three. It is the same instinct as checking which branch you are on before you commit: ten seconds that prevent an hour of confusion.
• —Decide what the agent may do without asking. Permission modes and plan mode are the difference between delegating and hoping.
• —Take one change from request to commit. Explore, plan, implement, test, review the diff, correct, commit — the full loop, once, slowly, before you ever do it fast.
• —Keep a long session on track. Context, compact, clear, rewind, resume — the commands for a session that has started to wander.
• —Leave the project a memory. A CLAUDE.md that makes the next session follow the rules without being told twice.

The course teaches the commands too, naturally — you cannot set a permission mode you have never heard of. But every command sits underneath a habit, which is why the lessons still read correctly after the flags move.

02What the course covers

• —Lesson 1 — Install, sign in, and look around: the one-line install, first sign-in, workspace trust, and the /status habit.
• —Lesson 2 — Permission modes and plan mode: the four modes, what plan mode is for, and what a plan approval actually looks like.
• —Lesson 3 — Your first change, from request to commit: the full loop on a real repository, with a failing test written first.
• —Lesson 4 — Keep a session on track: context, /compact, /clear, rewind and resume — recovery before you need it.
• —Lesson 5 — Project memory with CLAUDE.md: the file, what belongs in it, and watching a fresh session follow it.
• —Capstone — fix a reported bug, from plan to commit: everything above, unsupervised, on a bug you did not choose.

The three electives — Claude Code in your editor, giving Claude a screenshot, and learning from Claude Code itself — are fifteen to twenty minutes each and standalone: useful the day you need them, skippable until then. The whole level is about two and a half hours with the exercises.

03Every lesson ends in your repository

Reading is not the unit of work; the exercise is. Each lesson ends with one exercise you run in a repository you don’t mind experimenting on, and the level closes with an exit checklist — five things you can demonstrably do by the end, from “choose and switch permission modes” to “write a project CLAUDE.md and watch a fresh session follow it.” The capstone checks what files and git can see; you tick the rest, honestly. That checklist is also the entry test for what comes next: the course ships in three installments, and Beginner is live now. Intermediate — skills, hooks, subagents, MCP and plugins — is in preparation, and Advanced follows it. Each level exits into the next.

04How it was made, and what we verified

We did not write this from scratch, and we are not going to pretend otherwise. The course is our adaptation of Wesam Abousaid’s open curriculum “Claude Code: Everything You Need to Know” (v2026.10), used under its MIT license: the lesson spine, the exercise pattern and the capstone design are his; the phrasing, the opinions and the field notes from daily agent delivery are ours. Every fact in the course was checked against the official Claude Code documentation at v2.1.285 on the stable channel, October 2026 — the version we run. We are not affiliated with or endorsed by Anthropic; we just spend our days in this tool and think it deserves a better on-ramp. If you want the automated per-lesson checks and the purpose-built practice repository, use his upstream template — the course stands on its own, and points back to it.

05The honest caveats

• —Version drift is real. The course was verified against v2.1.285 in October 2026; by the time you read this, a flag may have moved. The habits hold; when a command disagrees with the docs, the docs win.
• —Claude Code is not on the free plan. You need a Claude subscription — Pro, Max, Team or Enterprise — or a Console account, or a supported cloud provider. Lesson 1 walks the sign-in for each.
• —The exercises assume a repository you can afford to break. Make one before Lesson 3, not during it.
• —“Daily driver” does not mean autopilot. What we let agents do unsupervised — and what we never will — is written down elsewhere, and the course teaches the same discipline.

If you have been meaning to learn Claude Code properly — or you manage people who have — start at Lesson 1. Twenty minutes, one sign-in, and the first habit is yours.