Anthropic's official CLI coding agent. Set up Claude Code, understand its key commands, and ship your first feature with it.

Claude Code: From Zero to Productive in 20 Minutes

Claude Code is Anthropic’s official CLI for AI-assisted development. It runs in your terminal, reads your repo, and can edit files + run commands.

Install

npm install -g @anthropic-ai/claude-code
# Or: brew install claude-code

# Authenticate
claude login

Core concepts

Concept	What it does
`claude` (no args)	Interactive REPL
`claude -p "..."`	One-shot prompt (non-interactive)
`claude "fix the bug in auth.ts"`	Repo-aware task
`/command`	Slash commands inside REPL

Your first session

cd ~/projects/my-app
claude
> /help
> read package.json and explain what this app does
> find all TODO comments in src/
> refactor the validateEmail function to use zod

The CLAUDE.md pattern

Drop a CLAUDE.md in your project root:

# Project: MyApp
- Stack: Next.js 15 + Postgres + Clerk auth
- Conventions:
  - Use Server Actions, not API routes
  - All forms use react-hook-form
  - Never commit to main directly
- Test command: npm test
- Deploy: vercel --prod

Claude reads this on every session. This is the highest-leverage thing you can do.

Key shortcuts

claude --continue — resume last session
claude --resume — pick from history
claude "add tests for X" — task with file context
/clear — reset context window

Key takeaways

CLAUDE.md is your secret weapon
Always run from project root (it indexes the repo)
Use --print (-p) for non-interactive scripting