Setup is the most engineer-dependent part, so it's the best place to ask for help. Everyone starts by asking.
AI tools
You have three ways to run your AI assistant:
- Claude Code desktop app (easiest): a standalone app, no code editor needed.
- Claude Code in the terminal: the command-line version.
- A code editor like VS Code or Cursor, with the Claude Code extension.
You sign in with a Claude account. The desktop app uses a personal subscription (Pro or Max). To use API credits your organisation provides, use the terminal or an editor. Any of them works; the steps are the same, only the interface differs. (A common setup: Claude Code on the desktop app for building, and Codex for a second opinion on review.)
Getting your computer ready (the app or editor, plus a few tools the project needs) is a one-time setup an engineer can walk you through. To do it yourself, Claude Code's setup guide covers it. Some repos also ship a design harness you install once; on a dx-harness repo, see Install.
Product repo
Most of this happens once. The first time, ask an engineer to walk through it with you.
1. Get access
An engineer adds you to the repo (on GitHub or GitLab) so you're allowed in.
2. Clone it
Cloning downloads the code to your computer.
- On the repo's web page, select Code and copy the HTTPS address (it starts with
https://). - Run
git clone [address], or ask your AI to do it. - You sign in once; your computer remembers it after that (Keychain on a Mac, credential helpers on Windows and Linux). SSH, the
git@…address, is an alternative some teams prefer.
Keep your access safe. Your login is personal, so never share it or use a teammate's. Never put a token or key into a commit, an AI prompt, Slack, or email.
3. Get the .env values
A .env file holds the app's settings and secret keys, like a database address or an API key, kept out of the repo on purpose.
- Your engineer sends the real values through a secure channel, like a password manager or vault, not plain Slack.
- Paste them into your local
.env. Without them, the app won't start.
4. Start the preview
Ask your AI for the command to run the project; it can read the repo and tell you exactly.
- The app opens in your browser at an address like
http://localhost:5173. That's your preview: your own copy, where you view and test your changes. - Start it fresh each session. If it stops, that's normal; ask your AI to start it again.
- Some apps need other pieces too, like a database; an engineer helps set those up the first time.