Claude Code

Anthropic’s agent. The one we teach most, in app or terminal.

Claude Code is Anthropic's coding agent, the one we teach most at AI Clubhouse workshops and the one behind most of the demos you'll see at our meetups. It reads your entire project, plans before acting, edits across many files, runs and fixes its own commands, and drives Git so you never memorize a Git command. It needs Claude Pro ($20/month).

Two ways to run it

The app: download the Claude desktop app from claude.com/download, sign in, and open the Code tab. Visual diffs, multiple sessions side by side, no terminal required. (The same app contains Cowork, the non-coding side of Claude, which gets its own chapter next section.)

The terminal: one paste.

curl -fsSL https://claude.ai/install.sh | bash

irm https://claude.ai/install.ps1 | iex

Then start it inside any project folder and log in when prompted:

cd my-project
claude

Three commands to know on day one

• –/init reads your project and writes it a memory file (CLAUDE.md), so every future session starts already knowing your setup.
• –Shift+Tab toggles plan mode. It writes out what it intends to do and waits for your approval before touching anything. Default to this for everything bigger than a one-file change.
• –/compact compresses a long session so quality stays high. Reach for it when a conversation starts feeling foggy.

A first prompt that shows what it can do:

Build me a personal decision journal web app. I log a decision, my prediction, and a confidence score; later I record what actually happened and see my hit rate. Clean, minimal design. Pick the easiest stack to run locally, and start it when you're done so I can try it.

Notice the shape of that prompt: outcome-focused, opinionated about the feel, silent on the tech, and it ends by asking for the running app rather than the code. That shape is the whole craft, and the next section drills it.

AI Clubhouse · Vibe Coding Guide