Setup and First Agentic Edit

What Claude Code Actually Is

Most AI coding tools work like this: you paste code → AI suggests edits → you copy-paste back. You're the integration layer between AI output and your actual files.

Claude Code flips that. It runs in your terminal with access to your file system. When you give it a task, it:

1. Reads the relevant files in your project (automatically)
2. Plans what changes need to happen
3. Makes the edits directly to your files
4. Runs your test suite to verify the changes work
5. Tells you what it did and asks if you want to continue

You're not copy-pasting. Claude Code is pair programming — it just happens to be on your side of the keyboard.

Installation

You need Node.js (v18+) and an Anthropic API key. If you don't have an API key, get one at console.anthropic.com — free tier gives you $5 of credits to start.

# Install Claude Code globally
npm install -g @anthropic-ai/claude-code

# Verify installation
claude --version

# Set your API key (first-time setup)
export ANTHROPIC_API_KEY=sk-ant-your-key-here

# Or add it permanently to your shell config
echo 'export ANTHROPIC_API_KEY=sk-ant-your-key-here' >> ~/.zshrc
source ~/.zshrc

Your First Agentic Command

Navigate to any existing project — it can be anything: a React app, a Python script, a Node API. Then run:

cd your-project

claude "explain this codebase to me in 3 sentences"

Watch what happens. Claude Code doesn't ask for a file. It reads your project structure — package.json, entry points, directory layout — and synthesizes an explanation. It didn't need you to paste anything.

Now try something that requires actual changes:

claude "add a dark mode toggle to the header component"

Claude Code will:

1. Find your header component file
2. Read it
3. Determine what dark mode implementation makes sense given your existing styles
4. Show you the planned changes (before making them)
5. Ask for confirmation
6. Apply the edits

Agentic Mode vs. Asking Mode

Claude Code has two modes you'll use constantly:

Agentic mode (default)

Give it a task. It reads files, plans, edits, and runs commands autonomously. It will ask for confirmation before making changes unless you use --dangerously-skip-permissions (more on that in Day 4). Best for: adding features, refactoring, fixing bugs.

Interactive mode

Just run claude with no arguments to open an interactive session. You can ask questions, get explanations, and then decide whether to let it make changes.

# Start interactive session
claude

# Inside the session, you can type naturally:
> How is the authentication currently implemented?
> What files would I need to change to add OAuth?
> Go ahead and add GitHub OAuth

# Exit with
> /exit

Understanding What Claude Code Sees

Claude Code reads your project by looking at:

• File tree: It maps your directory structure on startup
• Key config files: package.json, tsconfig.json, .env.example, requirements.txt
• Files you mention: It reads them fully when you reference them
• CLAUDE.md (if present): Project-level instructions you write — covered in Day 2
• Files it needs to edit: It reads these before making changes

It does not read every file in your project by default — that would be wasteful and expensive. It reads what it needs, when it needs it.