Short answer: learn Claude Code by building one small app of your own. First learn six basics: briefing Claude, Plan mode, tests, Git, CLAUDE.md and when to start over. Skip subagents, MCP, hooks and plugins until you need them. Then follow four weeks: stack, structure and agent workflow; data, auth and payments; environments, tests and releases; logs, alerts, backups and security. Start free with my first three videos and the playbook; a course adds reviews of your own app.
What Claude Code is, and how to learn it
Claude Code is Anthropic's AI coding agent. It reads your project, edits files and runs commands. It runs in the terminal, code editors, the browser and the Claude desktop app, which my course uses.
I'm CTO at DeVmand, with 15 years in software. This site was built with Claude Code, and my own build times fell from a month to two days: sendinvoicefast.com took a week, handprints.app two days.
Learn it by building one small app end to end, picking up each skill when the app needs it.
What to learn first
- Briefing Claude. Treat it like a contractor: say what you want, where, and what "done" looks like. A vague task gets a guess.
- Plan mode. Claude reads your project and proposes a plan without editing code. In the desktop app, pick Plan in the mode menu under the message box; after you approve the plan, switch to Accept edits or Manual. Fixing a plan is cheaper than fixing code.
- Checks that pass or fail. A test is a small program that checks one thing your app must do. Claude stops when the work looks finished; without a check, you're the tester. Ask for the command it ran and its output.
- Git. It keeps your project's history. A commit is a saved snapshot of changes. A branch is a separate line of work that doesn't touch the live app until approved. A pull request is a request to merge a branch, reviewed and checked first. Claude runs the commands; you need to know what they mean.
- CLAUDE.md. A text file of rules Claude reads at the start of every session. Anthropic's docs say to add it first: what to put in it.
- When to start over. Esc stops Claude mid-task. After two failed corrections, start a fresh session with one better prompt.
A first task in Plan mode, on Quest Guild, the app I build in the course:
Parents can't delete a quest. Find why, plan the fix and a test that proves it. Show me the test output when done.
More habits: Claude Code best practices.
What you can skip for now
Anthropic's docs say you don't need it all up front: add each feature when a need shows up.
- Subagents, helpers for side tasks. Claude already starts them on its own.
- MCP servers, links to outside tools like Slack. Add one when you keep copying data Claude can't see.
- Hooks, scripts that run at fixed moments, for example to block an action. Add one when a rule must never bend: hooks in plain words.
- Your own skills and plugins. A skill is saved instructions Claude loads when needed; a plugin bundles skills and more into one install. You'll install one skill on day one; writing your own can wait.
- Prompt tricks. ALL CAPS and "think harder" were workarounds for older models; Anthropic says current ones follow plain instructions.
- Writing code by hand. Your job is the one from my first video: know what you want, know your customers, and send bad work back.
What you need before you start
- A paid Claude plan. Claude Code in the desktop app needs one.
- The desktop app, Git and Docker. Docker runs your app and its database the same way on any computer. Claude installs both for you.
- The free MVP System Design skill, the rulebook Claude follows.
- Comfort in a terminal, the text window where you type commands. You don't need to be a developer.
- A small first idea you understand well, not your dream app.
Every step: set up Claude Code, step by step.
A 4-week path from idea to a live app
ProCode follows this path: four weeks, one app, a layer a week. Four weeks is a typical pace, not a schedule.
Week 1: stack, structure, agent workflow
A boring stack and specs agents follow. Your stack is the tools your app is built on; boring ones with long track records let Claude build instead of debating basics. Write the spec, a plain-words description of your app. Set the folder structure and the agent workflow for every change: plan, build, check, commit.
Week 2: data, auth, payments
The parts that break demos when users arrive. Data is what your app stores: the terms in your spec become database tables. Auth is sign-up and access: who can see and change what. Payments go through Stripe, a payments service.
Week 3: environments, tests, releases
Deploys you can roll back in one command. Your app gets three copies (environments): dev on your computer, stage for rehearsal and prod, the live app. Tests catch what agents break. A deploy puts a new version live; one command rolls it back.
Week 4: logs, alerts, backups, security
Know about failures before your users do. Logs record what your app did; alerts tell you when something breaks. Backups are copies of your data, restored once to prove they work. Security means secrets (your app's passwords and keys) and rate limits, which cap how often someone can repeat an action.
Then launch: your app, live in production.
Four traps that sink first apps
- No spec. Claude can read your code, not your mind. Fix: a one-page brain dump, then let Claude grill you until you have
idea.md, one file describing your app: turn your idea into a spec. - Skipping tests. Each change can quietly break something that worked yesterday, and a user finds it first. Fix: a test with every change and a minimum test coverage, the share of code tests run. My playbook fails the build under 99%.
- No version control. Without Git, a bad change has no way back. Claude Code's checkpoints can rewind recent edits, but Anthropic's docs say they don't replace Git. Fix: commit every working step, and send each change through a branch and a pull request, never straight to main, the branch your users run.
- Shipping without backups. One bad command or a failed server can wipe your users' data. A backup you've never restored is a hope. Fix: automatic backups before launch, and one test restore.
Before launch, go through the production-ready checklist.
Free first steps, and when a course is worth it
You can start free, with ProCode Basic, my free course:
- Build an MVP in one week with AI agents (video 1): why one person can now do a team's work.
- Set up Claude Code, step by step (video 2): the app, the plan, Git, Docker and the skill.
- Turn your idea into a spec (video 3): from a brain dump to
idea.md. - The MVP System Design playbook: the skill I made, and the rulebook ProCode is built on. 16 chapters from an empty repo to production, free, MIT license.
Anthropic also has free self-paced courses on Claude Academy, such as Claude Code 101, for a tour of the tool itself.
The free path gives you the rules, but nobody reviews your app. A Claude Code course is worth it when you want someone to. In ProCode Extended you apply the playbook to your own app, with recorded lessons, a weekly live Q&A, two 30-minute 1:1 sessions with me and reviews of your code and architecture. It’s hybrid and you start any day.
Skip it if you want the app built for you, have no app idea yet or want an intro to programming. If the free path gets your app live, you don't need it.
Product facts are from Anthropic's Claude Code docs (overview, best practices, extending Claude Code, desktop app), checked on 7 November 2026. All guides: articles.