You are a designer. Or an artist. Or a photographer. Or someone who makes things with their hands and their eyes.

You have heard about Claude Code. You have heard it can build websites, automate workflows, write entire applications. You may have also heard it requires a terminal. It does not. Not any more.

This guide is for you. No programming knowledge. No terminal. No jargon. One hour, and you will have built something real.

What Claude Code actually is

Claude Code is an AI that talks to your computer.

Not metaphorically. Literally. You describe what you want in plain English, and Claude creates files, opens apps, runs things, moves things around, and shows you the result — all on your machine, using your stuff.

Think of it this way:

  • Claude.ai (the chat) is like asking someone to describe how to bake bread. They give you the recipe, but you do the work.
  • Claude Code is like having that person standing in your kitchen, mixing ingredients, adjusting the oven, and handing you warm bread.

You do not need to know where files are. You do not need to open a browser yourself. You do not need to know what any of the technical words mean. You describe, Claude does.

What you need

A computer and an internet connection. That is genuinely it.

Claude Code runs inside the Claude desktop app on macOS and Windows. Linux is not currently supported for the desktop app — if you are on Linux, see the terminal section below. Any machine made after 2018 will handle it comfortably. You do not need to install anything else — no programming languages, no package managers, no command-line tools.

One exception: on Windows, you need Git installed. Download it from git-scm.com and run the installer — the defaults are fine. Most Macs include Git already.

You also need a paid Claude account. Claude Code requires at minimum the Pro plan at $20 per month. There is no free tier. If you are already paying for Claude Pro or Max, you already have access.

Step 1: Download Claude

Go to claude.ai/download and download the app for your system:

  • Mac: Download the universal build. Open the .dmg file. Drag Claude into your Applications folder. Open it from Applications.
  • Windows: Download the installer. Run it. Claude appears in your Start menu.

Sign in with your Claude account. The app opens to a chat window. This is the regular Claude chat — not Claude Code. Not yet.

Step 2: Open the Code tab

At the top of the window, you will see three tabs: Chat, Cowork, and Code.

Click Code.

If it asks you to upgrade, you need a paid plan first. If it asks you to sign in online, do that and restart the app. Once you are through, you will see a clean interface with a prompt box at the bottom. This is Claude Code.

Step 3: Point it at a folder

Claude Code needs to know where to work. It reads and writes files in a specific folder — your project.

Click Select folder and choose where you want your project to live. If you do not have a project yet, create a new empty folder on your desktop. Call it whatever you like. Select that.

You will also see a model selector near the send button. Leave it on the default for now. You can experiment with models later.

Step 4: Build something

Type what you want. In plain English. Be specific about what you want it to look like.

Try this:

Create a full-screen interactive particle animation — hundreds of small dots drifting slowly across a dark background, responding gently to mouse movement. Use soft, warm colours. Make it beautiful and meditative. When you’re done, open it in my browser so I can see it.

Press Enter. Watch.

Claude thinks for a moment, then starts creating. You will see exactly what it is building — changes highlighted in the sidebar. When it finishes, it opens the result in your browser. You did not need to find a file, double-click anything, or know what “HTML” means. You described what you wanted, and there it is.

By default, Claude asks your permission before making changes. You will see Accept and Reject buttons beside each change. Click Accept.

Not quite right? Say so. “Make the particles smaller.” “Use a cream background instead.” “Add a subtle fade trail.” Claude modifies the file based on your feedback. This is a conversation, not a vending machine.

Step 5: The preview

You may have noticed — Claude already opened your creation in a browser. That is what happens when you ask it to. But the desktop app can do something better: show you a live preview right inside the window.

Look for the Preview dropdown in the session toolbar. Claude can open an embedded browser right in the app — you see your work without switching windows. Make a change, see it update instantly. Say “make the particles smaller” and watch them shrink in real time.

This is the rhythm of working with Claude Code. You describe, it builds, you see it, you refine. Everything happens in one place.

Step 6: Set up your project’s memory

Here is the most underused feature in Claude Code: the memory file.

Every time you start a new session, Claude starts fresh — no memory of what you discussed yesterday. A file called CLAUDE.md fixes that. Think of it as a creative brief that Claude reads at the start of every session — your preferences, your style, your rules. You never need to open or edit this file yourself. Just tell Claude what to remember.

Create one now. In your session, type:

Remember these things about me for this project: I’m a designer, not a developer. I prefer clean, minimal aesthetics with lots of whitespace. I like warm colour palettes — creams, soft grays, muted earth tones. Keep everything simple — I should be able to see my work in a browser without any setup. Use smooth, subtle animations. And explain every decision you make as you go — teach me along the way.

Claude creates the file. From now on, every session in this folder starts with Claude reading those preferences. You never repeat yourself.

You can also type /init and Claude will generate a CLAUDE.md automatically by analysing your project. But writing your own — describing your creative sensibilities — is more powerful.

Keep it under 200 lines. Longer files consume context and reduce adherence. Short, specific, well-structured instructions work best.

One more thing to put in your memory file: Add “Explain what you’re doing as you work, in plain language. Teach me along the way.” Claude will narrate its decisions, tell you why it made certain choices, and gradually teach you the concepts without you needing to study them. The AI is not just a builder — it is the most patient teacher you will ever have.

Step 7: Learn the rhythm

A few things that will save you time as you settle in.

Interrupt freely. If Claude is heading in the wrong direction, click the stop button or just type your correction and press Enter. Claude stops and adjusts. You do not have to wait for it to finish.

Give context with @. Type @ followed by a filename to pull a specific file into the conversation. Drag in images, PDFs, screenshots. The more context Claude has, the better the results.

Use skills for repeatable tasks. Type / to browse built-in commands and custom skills — reusable prompts you can invoke any time, like a design review checklist or a deployment step.

Switch permission modes. You started in Ask mode, where Claude waits for approval before every change. Once you trust the rhythm, switch to Auto accept edits — Claude makes changes without asking, and you review after. For big tasks, try Plan mode — Claude maps out an approach without touching any files, and you approve the plan before execution.

What you just did

In roughly thirty minutes, you:

  1. Downloaded an app
  2. Opened a tab
  3. Created something visual through conversation
  4. Set up a persistent creative brief that carries across sessions

You did not learn to code. You learned to vibe code — to describe what you want and let the AI build it. That is a real skill, and it matters.

Common mistakes to avoid

Starting vague. “Make something cool” gives Claude nothing to work with. “Make a single-page portfolio with my name centred in large serif type on a cream background with a subtle particle animation behind it” gives Claude everything.

Correcting endlessly instead of starting fresh. If Claude is going in the wrong direction after two or three attempts, start a new session with a better initial description. A fresh start with a clear prompt almost always outperforms a long, confused conversation.

Not using CLAUDE.md. Every session without it is a cold start. You will repeat yourself constantly.

Treating the output as final. The first result is a draft. Iterate. “Move the text up.” “Make the animation slower.” “Change the background to this specific hex colour.” Each refinement takes seconds.

A note on the terminal

If you are comfortable with the terminal, Claude Code also runs as a command-line tool. The CLI and desktop app share the same engine, the same CLAUDE.md files, the same configuration. Some people prefer it. You do not need to.

If you are curious, see the official CLI quickstart. But everything in this guide — and every article in the code track — works from the desktop app.

Where to go from here

You have the foundation. From here, you can:

  • Build a portfolio site. Describe your layout, your type choices, your colour palette. Claude builds it.
  • Create interactive presentations. Scroll-triggered reveals, animations, full-screen visuals — describe it and Claude builds it.
  • Generate design tools. Colour palette generators, type scale calculators, spacing visualisers — tools that serve your creative process.
  • Prototype ideas. Before opening Figma, describe the concept to Claude Code. See it working in a browser in minutes.

Art & Algorithms publishes guides, tutorials, and prompt packs at the intersection of art and code. Subscribe for the full archive.