This module gets you from zero to a working Claude Code installation. If you're new to terminals and Node.js, check out the non-developer setup guide for a more detailed walkthrough.
Installation
There are two ways to install – pick whichever suits your setup:
# Option A: npm (works everywhere with Node.js)
npm install -g @anthropic-ai/claude-code
# Option B: Homebrew (macOS)
brew install --cask claude-codeTo verify the installation worked:
claude --versionYou should see a version number. If you get "command not found", make sure your npm global bin directory is in your system PATH.
Authentication
This is where people get tripped up, so it's worth reading carefully.
There are two ways to authenticate, and they have different billing implications:
Option 1: Claude subscription (recommended for teams)
If your company has a Claude Team or Enterprise plan, or you have a personal Max or Pro subscription:
- Run
claudein your terminal - Select "Claude.ai account" when prompted
- Follow the browser-based OAuth flow
Usage counts against your subscription's limits. There are no per-token charges, but you may hit usage caps during busy periods.
Option 2: Anthropic API key (pay-per-token)
If you want uncapped usage and are comfortable with pay-as-you-go:
- Go to console.anthropic.com and create an API key
- Run
claudein your terminal - Select "Anthropic API Key" when prompted
- Paste your key
Every request is billed per token. Costs vary by model – Opus is significantly more expensive than Sonnet.
The billing distinction matters. A Claude subscription gives you usage within plan limits. An API key charges per token with no cap. If you're unsure which you're using, run /cost during a session – if it shows dollar amounts, you're on API billing. Within our team, start with the company Claude subscription. If usage limits become a bottleneck, we may move to individual subscriptions or API key access – talk to Amadeus if you're regularly hitting caps.
Initialise Your First Project
Navigate to a project directory and run:
cd your-project
claude /initThis does two things:
- Scans your codebase – reads file structure, package.json, and common patterns
- Creates a
CLAUDE.mdfile – a summary of your project that Claude will reference on every request
You can (and should) edit this file afterwards. The auto-generated version is a starting point, not a finished product – we cover this in detail in the context and memory module.
Set a Budget Limit
If you're using an API key, it's worth setting a safety net straight away:
claude --max-budget-usd 5This caps the current session at $5. If Claude hits the limit, it stops rather than continuing to spend. You can adjust the number as you get comfortable with what typical sessions cost.
Verify Everything Works
Run a simple test to make sure everything is connected:
claude "What files are in the root of this project? List them briefly."If Claude lists your project files, you're good to go. If you see an authentication error, double-check your auth setup. If you see a timeout, check your internet connection.
Quick Orientation
When Claude Code is running, your terminal becomes an interactive session:
- You type prompts at the bottom
- Claude responds with text, and may request to use tools (read files, run commands, etc.)
- Permission prompts appear when Claude wants to modify files or run shell commands – you approve or reject each one
- Shift+Tab cycles between permission modes (Normal → Auto-accept → Plan)
Useful commands to know from day one: /help shows all available slash commands. /model shows which AI model you're using. /cost shows token usage for the current session. /clear starts a fresh conversation.
