Claude Code Tutorial: Setup and Your First Project

Claude Code is an AI coding agent that works in your terminal, your editor, or a desktop app. Setup takes a few minutes: install it with one command, run claude in a project folder, and log in. This claude code tutorial covers that setup and walks you through a small first project, so you can see how the tool behaves before you trust it with anything important.

What you need before you start

You need a paid Claude account. One pricing breakdown lists Claude Pro at $20/month as the entry point and says the Free plan has no Claude Code access. Max tiers sit above that. Anthropic changes plan names and limits fairly often, so check the pricing page before you pay. You can also use an API account, which bills by usage.

You don't need Node.js for the recommended install. On Windows, the official terminal guide says you need Windows 10 version 1809 or later. Git for Windows is optional, and without it Claude Code uses PowerShell.

Step 1: Pick how you want to run it

There are a few ways in, and they all run the same Claude Code. Per the official overview, you can use the terminal CLI, a VS Code or Cursor extension, a desktop app, a JetBrains plugin, or the web.

• Terminal: the most documented route, and the one most tutorials assume.
• VS Code or Cursor extension: a chat panel next to your files. Handy if you already live in an editor.
• Desktop app: the docs say it lets you skip the terminal entirely. If a black window with a blinking cursor makes you nervous, start here.

The rest of this guide uses the terminal, since the commands carry over to the other options.

Step 2: Install it

Open a terminal and run the line for your system. These come from the official docs:

System	Command
macOS / Linux / WSL	curl -fsSL https://claude.ai/install.sh | bash
Windows PowerShell	irm https://claude.ai/install.ps1 | iex
Homebrew (macOS)	brew install --cask claude-code

The docs also list WinGet and Windows CMD options, plus apt, dnf and apk on Linux. Check the install page if one of those fits you better.

When it finishes, confirm it worked:

claude --version
claude doctor

If PowerShell says claude isn't recognized, the terminal guide says the install folder isn't on your PATH yet. Close the window and open a new one first. That fixes it more often than you'd expect.

Step 3: Make a project folder and log in

Make an empty folder and start Claude Code inside it. I'd also run git init now. That gives you a safety net, because you can see exactly what changed and roll back.

mkdir habit-tracker
cd habit-tracker
git init
claude

On first run, Claude Code prompts you to log in. A browser window opens, you authorize your account, and you come back to the terminal. You'll also pick a text theme and see some security notes. The short version of those notes is that Claude can make mistakes, so use it on code you trust.

Step 4: Learn the three controls that matter

You don't need the whole command list yet. Learn these first.

Slash commands

Type / to open the menu. Anything starting with a slash controls the session instead of asking Claude to write code. The official cheatsheet lists the useful ones: a command to list everything, one to wipe the conversation and start fresh (project memory stays), and one to roll the conversation or code back to an earlier checkpoint.

Permission modes

By default Claude asks before it edits files or runs commands. Press Shift+Tab to cycle modes. The cheatsheet describes the cycle as Default, then Accept Edits, then Plan Mode. Plan Mode is read-only: Claude explores and proposes but won't edit files or run commands. Use it whenever you're unsure what you're asking for.

CLAUDE.md

This is a markdown file Claude reads automatically at the start of every session. It holds your project's conventions, commands, and constraints, so you don't repeat yourself. Running /init in an existing project has Claude look at the code and draft one for you. In an empty folder there's nothing to analyze yet, so we'll write a short one by hand.

Step 5: Build a small first project

Pick something you can finish in one sitting. A habit tracker in plain HTML, CSS and JavaScript works well. It has no build step, no accounts and no database, so there's little to break.

Press Shift+Tab until you're in Plan Mode, then paste this:

I'm a beginner. I want a single-page habit tracker that runs by opening index.html in a browser. No frameworks, no build tools.

Features:
- Add a habit with a text box
- A checkbox per habit for today
- Habits and check-ins saved in localStorage
- A delete button per habit

Before writing code, give me a short plan: which files you'll create, what each does, and anything you're unsure about. Keep it as simple as possible.

Read the plan. If it proposes React, a backend, or six files, push back: "Simpler. One HTML file, one JS file, one CSS file." Plans are cheap to change and finished code isn't, which is why this step is worth the extra minute.

When the plan looks right, press Shift+Tab to leave Plan Mode and say: Implement the plan. Approve the file edits as they come. Then open index.html in your browser and try it.

Step 6: Write a short CLAUDE.md

Before you iterate, give Claude some memory. Ask for it, or create CLAUDE.md yourself:

# Habit tracker

- Plain HTML/CSS/JS. No frameworks, no build step.
- Open index.html directly in a browser to test.
- Data lives in localStorage only.
- Keep changes small. Explain what you changed in one or two sentences.
- Don't add dependencies without asking.

Keep it short. One cheat sheet suggests staying under 200 lines. Long files bury the rules that matter. Add lines only when you catch yourself repeating an instruction.

Step 7: Iterate in small bites

Now improve it one change at a time. Good second prompts:

• "Add a streak counter per habit. Show me the diff before applying."
• "The delete button has no confirmation. Add one."
• "Explain how the localStorage code works, line by line. I want to understand it."

Commit after each change that works:

git add -A
git commit -m "Add streak counter"

If a change breaks something, you can rewind the session or revert in git. Ask Claude to write the commit message too, if you like. Run /clear between unrelated tasks so old context doesn't leak into new work.

Common beginner snags

• Asking for too much at once. "Build me a full app with logins and payments" gives you a pile of code you can't check. Ask for one feature, test it, then ask for the next.
• Approving edits without looking. You don't need to read every line. Do read the file names and the summary.
• Hitting usage limits. Claude Code shares your plan's limits with the Claude apps, so a long chat session and a long coding session draw from the same pool. Short, focused prompts help.
• Skipping version control. It takes 30 seconds, and you'll be glad of it the first time something goes sideways.

Where to go next

Once the habit tracker works, try the same flow on a project you care about: Plan Mode first, a short CLAUDE.md, small prompts, frequent commits. After that, look at the cheatsheet for hooks, skills, and MCP servers. You don't need any of them to ship something small.

FAQ

Is Claude Code free?

No. The Free Claude plan doesn't include it. You need a paid plan such as Pro, or an API account billed by usage. Check Anthropic's pricing page for current prices, since plans and limits change.

Do I need to know how to code to follow this claude code tutorial?

No, but it helps to be curious about what the code does. Ask Claude to explain anything you don't understand, and keep projects small enough that you can test each change yourself. The more you build, the faster you'll spot when something looks wrong.

Should I use the terminal, VS Code, or the desktop app?

Use whichever you'll actually open. The terminal has the most documentation, an editor extension keeps your files visible beside the chat, and the desktop app avoids the command line entirely. They all run the same Claude Code, so you can switch later without relearning the basics.

Cover photo: Vagaro on Unsplash