Intro: What Claude Code is — and how to install it

Before the agentic loop, before CLAUDE.md, before subagents — there is a five-minute setup. This guide covers it completely.

Claude Code is not Claude.ai

Claude.ai is a web chat interface. You open a browser, type a message, read a response. It cannot access your local files, cannot run code on your machine, cannot make changes to your codebase.

Claude Code is different in kind, not just degree. It is a local agent that runs in your terminal. It reads your files. It executes commands. It writes code, runs tests, and makes changes — on your machine, in your project, with your actual codebase.

The browser interface and the terminal agent share the same underlying model, but the capabilities are not comparable. Claude Code is a tool for software development. Claude.ai is a general-purpose chat interface.

What you need before installing

Three things:

1. Node.js 18 or higher. Claude Code is distributed as an npm package. Check your version:

node --version

If you see v18.x.x or higher, you are ready. If not, install Node.js from nodejs.org — the LTS version is the right choice.

2. An Anthropic account. Claude Code authenticates against Anthropic's API. You need an account at console.anthropic.com. A Claude Pro subscription or API access works — Claude Code prompts you to authenticate on first run.

3. A project directory to work in. Claude Code uses your current directory as its working context. It reads files relative to where you launch it. Open a terminal in the root of a project you want to work on.

Install

One command:

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

This installs Claude Code globally. You can run it from any project directory without reinstalling.

Verify the install worked:

claude --version

First run

Navigate to a project directory and launch Claude Code:

cd your-project
claude

On first run, Claude Code opens a browser window for authentication. Sign in with your Anthropic account. The authentication token is stored locally — you will not need to do this again.

After authentication, you see the Claude Code prompt:

Claude Code v1.x.x
Type a task, question, or /help for commands
>

You are now running a local AI agent with full access to your project files and terminal. Part 1 explains what it does with that access — and why understanding that matters before you send your first real task.

The VS Code extension

Claude Code also has a VS Code extension (Part 2). It is not required — everything works from the terminal. The extension adds a visual interface: plan review, diff approval, checkpoint rewind. For engineers learning Claude Code, the VS Code interface makes the agentic loop more legible. For CI pipelines and automated workflows, the CLI is the right tool.

Install both. Use the extension for most development work. Use the CLI for scripting and automation.

What is actually happening when you run claude

Claude Code loads your project context — directory structure, key files, any CLAUDE.md configuration — and makes it available to the model. When you send a task, the model can read files, run commands, write changes, and observe results in a loop. Every step is shown in the terminal.

Part 1 covers the loop in detail. The important thing to understand now: unlike a chat interface, Claude Code can act on your codebase. The first task you send will change files, run commands, or both. That is the point.

Next in this series: Part 1 — The agentic loop: why Claude Code is not autocomplete