Installing Claude Code

Let’s walk through installation in detail and get Claude Code configured for the best experience.

Installation

npm install -g @anthropic-ai/claude-code

Tip

If you get permission errors on macOS/Linux, you may need sudo or to fix your npm global prefix. See the npm docs.

First Run

Navigate to any project directory and launch Claude:

cd ~/my-project
claude

On first run, Claude will:

1. Open your browser for authentication
2. Ask you to authorize with your Anthropic account
3. Create a ~/.claude/ directory for global configuration

Key Directories

After installation, Claude Code uses these locations:

Path	Purpose
~/.claude/	Global config, memory, settings
~/.claude/settings.json	Global permission settings
.claude/ (in project)	Project-specific config and memory
CLAUDE.md (in project root)	Project instructions Claude reads on every session

The CLAUDE.md File

This is one of the most important concepts in Claude Code. A CLAUDE.md file in your project root acts as persistent instructions that Claude reads at the start of every conversation.

# Project: My App

## Stack
- Next.js 14 with App Router
- Tailwind CSS
- Prisma + PostgreSQL

## Conventions
- Use TypeScript strict mode
- Components go in src/components/
- API routes go in src/app/api/

## Do NOT
- Install new dependencies without asking
- Modify the database schema without confirmation

Think of it as a README for your AI collaborator. We’ll use CLAUDE.md extensively throughout this course.

Choosing Your Model

Claude Code lets you pick which model powers your session:

• Opus — most capable, best for complex tasks (default on Max plan)
• Sonnet — fast and capable, good for most work
• Haiku — fastest, good for simple tasks

You can set this in your config or switch during a session.

Next Steps

With Claude Code installed and running, we’re ready to learn the fundamentals of working with it effectively.