chapter 01core course

Start Here

Install Claude Code and make your first edit

Welcome. Before we touch anything, let's get one thing clear: is not a chat app. It's not ChatGPT with a different skin. What makes it useful is also what makes it different, and that difference is the whole reason this course exists.

The next 15 minutes will have you installing it, signing in, and watching it edit a real file in a real project on your computer. Every step is labeled with exactly where to run what — there's no guessing.

beginner~15 minyou'll need a terminal + a repo you care about

before you startset your expectations

what you should already know

Nothing — this is the start of the course. You're good to go.

new terms you'll meet

by the end of this chapter

Claude Code will be installed on your machine, signed in to your Anthropic account, and will have made one real edit to a file in your own repo. You'll understand what just happened well enough to explain it to a friend.

lesson

Claude Code is an agent, not a chat box

Think of as a pair-programmer sitting next to you. When you type a request, Claude doesn't just write words — it uses : it opens your files, reads them, edits them, runs shell commands, checks the output. That's the definition of an : an AI that can do things on your computer, not just describe them.

Everything is permission-gated. The first time Claude wants to change a file, run a command, or write something new, it asks you. Nothing happens behind your back. You can say yes, no, or "yes for this whole session."

That mental shift — from "AI I copy-paste to" to "AI that has a keyboard of its own" — is the entire point of this chapter. Once it clicks, the rest of the course makes sense.

your move5 steps, ~15 min

Check that you have what you need

Three things. You probably have them all already; we'll check anyway so you're not surprised two steps in.

18 or newer. Node is the thing that runs JavaScript outside a web browser, and Claude Code is written in JavaScript — so it needs Node to run.
A . Anything reasonable works — macOS Terminal, iTerm, Warp, Windows Terminal, Ghostty. If you've ever typed a command into a black window, you have one.
A you care about. Any real project folder — a side project, work code, a blog. Don't make a throwaway one. Claude Code is most impressive when it's looking at code that's genuinely yours.

Why does Claude Code need Node? It's distributed as an package (npm is Node's installer for command-line tools). There are other install methods — Homebrew, a shell script — but this is the path that works on any operating system without surprises.

Check Node:

IN YOUR TERMINAL

If you see v18.x.x or higher, you're good. If it says command not found, install Node from nodejs.org first.

node --version

Install Claude Code

One command installs the globally on your computer. -g means — you'll be able to run claude from any folder, not just from inside a specific project.

IN YOUR TERMINAL

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

Confirm it worked:

IN YOUR TERMINAL

claude --version

If claude --version prints a number — anything like 1.x.x— you're done with this step. If not, open a fresh terminal window first (your shell sometimes needs a restart to see newly-installed commands). Still stuck? See Go deeper → If something went wrong at the bottom of this chapter.

Sign in

Now run Claude for the first time. It'll detect you haven't signed in yet and walk you through in your browser.

IN YOUR TERMINAL

claude

A browser tab opens. Sign in with your Anthropic account (the one you'd use at claude.ai). Once it confirms, come back to the terminal — you should now be in an interactive Claude . Don't do anything yet. Just close it with /exit or Ctrl-D. We'll reopen it properly in the next step.

Verify you're signed in:

IN YOUR TERMINAL

claude auth status

If the browser tab doesn't open automatically, copy the URL from the terminal and paste it into your browser manually. Corporate VPNs and adblockers sometimes block the redirect.

Open a real project and ask Claude about it

This is the first "oh wow" moment. Navigate to a real project folder on your computer and start Claude there.

IN YOUR TERMINAL

Replace path/to/your-project with the actual path to any repo you've got locally.

cd path/to/your-project
claude

Claude is now open in your project. Type the following into the Claude prompt — not your terminal, Claude's prompt:

TELL CLAUDE CODE

Just type the question and hit enter. No command, no syntax — talk to it like a person.

what does this project do?

claude

Watch what happens. Claude will announce it's opening a few files (that's the Read ), possibly run a grep (Grep tool), maybe list the directory (Glob tool), then answer you with a real summary of what your project actually does — not a generic guess from the folder name.

Why is this better than pasting the same question into ChatGPT? Because ChatGPT has to guess based on what you pasted. Claude Code just opens your files. The answer is grounded in your actual code, not an approximation of it.

Ask Claude to make your first edit

Time for the payoff. We're going to ask Claude to make a tiny, reversible change — one line in your README. Type this into Claude's prompt:

TELL CLAUDE CODE

add a single comment at the very top of README.md that says: // installed Claude Code

claude

Claude will propose the edit and ask for your permission. Say y. It applies the change to your real README.md file. Close the session (/exit) and open a new terminal window in the same folder:

IN YOUR TERMINAL

If your project isn't a git repo, that's fine — open README.md in your editor instead and verify the line is there.

git status

You should see README.md listed as modified. A real file on your disk just changed. Claude put it there.

Whether you keep the change (git add README.md && git commit) or revert it (git checkout README.md) is up to you. Either way, you've just run the full loop.

what just happened

Claude Code read your project (using its Read tool), answered a question about it (still Read), then modified a real file (using the Edit tool). It asked permission for every action so nothing happened you didn't approve.

That pattern — agent + tools + permission— is what makes Claude Code different from everything else you've used. The next ten minutes (Chapter 2) is about the mental model for working with it well. After that, we start installing the five things that make the difference between "cute AI toy" and "I actually ship code with this now."

Check your exercises below as you confirm each step works on your machine. Then hit → or the Next button to move to Chapter 2.

exercisestry this in your own repo0/5