How to install Claude Code and build your first thing

If you have an idea and no developer, an AI coding assistant is the cheapest way to find out whether the thing can exist. Claude Code is one of those assistants. This guide assumes you have never installed a developer tool, never opened a terminal, and would like to keep it that way for as long as possible.

What Claude Code actually is

Claude Code is an AI assistant that works directly with the files on your computer. You describe what you want in ordinary sentences. It reads the relevant files, writes or changes the code, runs the project, and tells you what it did. You approve or reject the changes.

Two things make it different from chatting with an AI in a browser:

• It sees the whole project. You do not paste files back and forth. It opens what it needs.
• It can act. It creates files, installs the things a project needs, runs the app and reads the errors, then fixes them.

The trade-off is that it lives in a terminal, the plain text window that developers type commands into. If that sounds unpleasant, skip to the version without a terminal further down.

What you need before you start

• A reasonably modern computer. macOS 13 or newer, Windows 10 (build 1809) or newer, or a recent Linux. At least 4 GB of memory.
• An internet connection. The assistant thinks on Anthropic's servers, not on your laptop.
• A paid Claude plan. Claude Code needs a Pro, Max, Team or Enterprise subscription, or a Claude Console account billed per use. The free claude.ai plan does not include it.
• About 30 minutes for the first run, most of it waiting and reading.

Step 1: open the terminal

The terminal is an app that is already on your computer. It shows a blank window where you type commands and press Enter.

• Mac: press Command and Space together, type Terminal, press Enter.
• Windows: press the Start button, type PowerShell, press Enter.
• Linux: you already know.

You will see a line ending in a symbol such as %, $ or >. That is the prompt. It is waiting for you. Nothing you type here can break your computer unless you tell it to delete things, and nothing in this guide does.

Step 2: install Claude Code

Copy the line for your system, paste it into the terminal, press Enter.

Mac and Linux:

curl -fsSL https://claude.ai/install.sh | bash

Windows PowerShell:

irm https://claude.ai/install.ps1 | iex

Text will scroll past. That is normal. Wait until the prompt comes back.

If you are on a Mac and already use Homebrew, brew install --cask claude-code works too. On Windows, winget install Anthropic.ClaudeCode is an alternative. The direct installer above updates itself in the background; the package managers do not, so you would have to upgrade by hand.

Check that it worked

claude --version

You should see a version number followed by (Claude Code). If instead you see command not found, close the terminal window completely, open a new one, and try again. That fixes it most of the time.

Step 3: log in

Type:

claude

The first time, it opens your browser and asks you to sign in to your Claude account. Approve it, return to the terminal, and you are in. Your login is remembered, so this happens once.

Step 4: give it a folder to work in

Claude Code works inside one folder, which developers call a project. Everything it reads and writes lives there. Make an empty one and go into it:

mkdir my-first-app
cd my-first-app
claude

mkdir makes a folder, cd moves into it. Now the assistant is running, and its world is that folder.

Step 5: ask for something small

Type a request in plain English and press Enter. Resist the urge to describe your whole product. Start with something you can check with your own eyes in two minutes:

Make a single web page that asks the visitor for their email address and shows a thank-you message. Keep it in one file I can open in my browser.

The assistant will tell you what it plans to do, create the file, and often offer to open or test it. Look at the result. If it is not what you meant, say so in the same plain language:

The thank-you message should replace the form instead of appearing underneath it.

This loop, describe, look, correct, is the entire job. You are not programming. You are reviewing.

Commands worth knowing on day one

Step 6: build up in small steps

The most common beginner mistake is one enormous request: "build me a booking system for hairdressers with payments and text messages". You will get something, it will be wrong in ways you cannot see, and you will not know which part to fix.

Ask for one visible thing at a time, and check each one before moving on:

1. A page that lists the services and prices.
2. A form to pick a date and time.
3. Saving the booking so it survives a refresh.
4. An email confirmation.
5. Payment.

After every step, open the thing and use it as a customer would. When something breaks, describe the symptom rather than guessing at the cause: "after I pick a time and press book, the page goes blank" is far more useful than "the database is broken".

What it costs

Claude Code is included in the Claude subscription plans. Pro is the entry point and is enough for evenings and weekends; Max exists for people who use it all day. With a Console account you pay per use instead, which is unpredictable while you are learning.

Budget separately for the things your app itself uses once it is live: hosting, a domain, and, if your app calls an AI service for its own features, the per-use cost of that. Those are covered in what it costs to build and run an app with AI.

Three habits that save you

• Read before you approve. The assistant explains what it is about to change. Skim it. If it mentions deleting or rewriting something you care about, say no and ask why it wants to.
• Keep a copy. Ask it early on: "set up git so I can undo changes". Then any mistake is one sentence away from being reversed.
• Test like a stranger. The assistant checks that the code runs. Only you can check that the thing makes sense to a human being.

If the terminal is a step too far

There is a desktop application that gives Claude Code the same abilities behind a normal window, and a version that runs in your browser. Search the official documentation for the desktop app if you would rather point and click. Everything in this guide applies; only the place you type changes.

The step before all of this

An AI assistant will build whatever you describe, quickly and without arguing. That is exactly why it is worth knowing, before you start, whether the thing is worth building: which parts genuinely need AI, which are ordinary software, what could go wrong, and what the smallest useful first version is.

That is the question the free check on this site answers, and the one the guides on validating an idea and whether your idea needs AI go into. Spending twenty minutes there will save you a week of building the wrong thing.