What should I build first with Claude Code as a beginner?

Build a one-page website about you or your business, and keep it on your own computer. It is the best first project because you can see the whole result in your browser in one sitting, and you already know the content because it is about you.

I've never written a line of code by hand in my life, and the way I work with Claude Code is the same every time: know what you want, ask for it, then make it better. A one-page site is the smallest project where you can practice all three of those moves and see each one pay off within minutes. The rest of this post gives you the steps in order.

Most beginners pick something big, like an app or a tool that connects to five other accounts, and stall on the setup before they see anything work. A first project should pass four tests: you can finish it today, you can tell whether it worked just by looking, you need no extra accounts, and a mistake costs you nothing.

Those later projects are all worth building. They just make a better second, third or fourth project, once you've felt the loop of asking, checking and fixing.

Claude Code installed and working, either the desktop app (the Code tab) or the version you open from a terminal, which is the window where you type instructions instead of clicking. If you haven't done that yet, start with Can I use Claude Code if I'm not a coder? and come back.

Then spend two minutes jotting down five facts in any notes app: what you do, who you do it for, the three things you offer, how someone should contact you, and the feel you want (calm, bold, playful). You'll hand these to Claude Code in step 2. If you don't have a business yet, build the page about yourself: a personal profile with who you are, what you're good at, and how to reach you.

A folder is where Claude Code keeps everything it creates, so give this project its own empty one.

Anthropic's docs say new terminal and VS Code sessions start in a mode where a separate checker reviews risky actions and routine ones run without asking. That's fine later. While you're learning what Claude Code does, seeing each request is worth the extra clicks.

Don't write the perfect request. Anthropic's best practices page has a move for exactly the moment when you don't know what you want yet: ask Claude to interview you, and have it write down the answers as a spec, which is a short written description of what you're building.

Paste this in, with your own details where the brackets are:

Claude Code comes back with questions, usually multiple choice: which sections the page should have, what the headline should say, what to put on the contact button. Answer them in plain words. If you don't know, say so and ask what it recommends. When it's done it creates a file called SPEC.md in your folder, and it will ask your permission first.

Open SPEC.md (double-click it, or ask Claude Code to show it to you) and read it like you'd read a quote from a freelancer. It's a plain text file, so there's nothing to decode. Look for anything wrong, anything missing, and anything you wouldn't say yourself.

Fix it by talking, not editing: type something like change the headline to [your words] and remove the pricing section, then ask Claude Code to update SPEC.md. This is the cheapest place in the whole project to change your mind, so spend five minutes here.

Anthropic's docs recommend starting a fresh session once the spec exists, so the build begins with a clean slate and a written definition of done. Type /clear, which the quickstart lists as the command that clears the conversation history, or start a new session. Your SPEC.md file stays in the folder.

Then ask for the build. The docs point out that Claude tends to read a request conservatively and build the smallest thing that fits, and that the phrase be ambitious tells it you want the fuller version.

Claude Code will ask permission before it creates files. Approve them. Watch the first minute: if it starts doing something you didn't ask for, press Esc, which stops it without losing what it already did.

Double-click index.html in your folder and it opens in your browser. That's your finished first result. If you can't find the file, ask Claude Code: where did you save the website?

Now do what Anthropic's docs call giving Claude a check to run. Claude stops when the work looks done, so ask it to prove it: go through SPEC.md item by item and tell me where each one is on the page, and flag anything you left out. The docs say to ask for evidence instead of accepting a bare 'it's done,' and this is the beginner version of that.

This is where beginners learn the skill that matters most. Pick three changes you'd actually want and ask for them one per message, refreshing the browser after each one. For example: make the headline shorter, change the main color to match my logo, move the contact button to the top.

One caveat from the docs: the rewind menu restores files Claude edited with its own file tools, but not files that were changed by a command it ran.

A real, finished thing: a one-page website that opens in your browser, a SPEC.md that says what it's supposed to be, and a folder you can reopen any day. You also practiced the whole loop (describe, build, check, change) on a project small enough to finish. One member put it this way: 'I pretty much did nothing besides tell it exactly what I wanted.' That sentence is the entire skill.

Putting the page on the internet, so other people can visit it, is a separate step. Can Claude Code build a website I can sell to clients? covers where this goes next.

Once your first page works, the habit that saves the most time is writing down what Claude Code keeps getting wrong, which is what What should I put in CLAUDE.md, and how do I organize it? covers. And if you'd rather have your first few builds laid out in order with someone to ask, Claude Code Club is $9/month, and that's where I walk members through them.