Course catalog
Build, preview, and ship from the desktop app, then add terminal power-ups.
Claude Code is a coding agent you can run from the desktop app or the terminal, and most of what makes it good at your project is configuration you write yourself. This course starts in the app: your first desktop session, reading what the agent is doing as it works, and checking its result in a live preview before you accept it. Then you teach it your project with the files that shape everything — CLAUDE.md, .claude/settings.json, .claude/commands/*.md, and .claude/agents/*.md — extend it with MCP servers, skills, and hooks, and ship from a session: review the diff, open a pull request, and watch CI. It closes with terminal power-ups: the install, @ file refs, the slash commands you reach for daily, headless runs in CI, rewind, and the context hygiene that keeps a long session cheap.
Section 1
Start Claude Code in the desktop app's Code tab: choose the environment, folder, mode and model, steer a running session, and check what it did in the transcript and panes.
Section 2
Let Claude run your app in the desktop Browser pane, point at what is wrong precisely, and turn console errors and failed checks into a fix loop you can trust.
Section 3
7-day free trial on every plan for first-time members · Cancel anytime
Get startedTeach Claude Code how this repository works: concise CLAUDE.md guidance, enforceable settings, reusable Skills, and a read-only planning mode.
Section 4
Review the diff, open a pull request and let CI and Auto-fix run, move long work to cloud sessions, and follow it from your phone, with GitHub and your host connected on purpose.
Section 5
Move to the terminal program: install the Claude Code CLI, launch it in the right directory, read the terminal transcript, and learn the keys and slash commands you reach for every day.
Section 6
Choose the right extension surface: isolated subagents and worktrees, discoverable Skills, external MCP tools, and deterministic Hooks.
Section 7
Review and ship agent work with independent evidence: inspect diffs, automate bounded runs, recover safely, and manage context deliberately.