How to Install Claude Code, for Designers
I've been noticing something in my conversations with product designers about transitioning to AI tools: an aversion to the terminal. Folks who never coded, or are a bit foreign to the development process (totally understandable, plenty of us are), find its lack of a simple UI and its special syntax threatening.
What I try to explain to them is this: you'll barely use the actual terminal. Day to day it's mostly a gateway, the door you open to get to Claude. Once Claude is running, you're just talking to a chatbot that happens to be dressed like a terminal. Think of it as a chatbot with Game Boy graphics, which honestly makes it kind of charming. The terminal itself is for the occasional computer-wide install, or a command Claude asks you to run, and if you're ever unsure, ask Claude and it'll walk you through it.
In this post I'll help you install Claude Code using the terminal. But before we install anything, let me hand you the two ideas that make the terminal make sense.
It's a conversation: The terminal is a text box that talks back. You type a command, hit return, the computer answers, and then it waits for your next one. That's the entire interaction model. Kinda like LLMs. We all spent the last few years learning to type instructions into a box and read the response. The terminal was prompting before prompting was cool. It's like that band from Williamsburg, Brooklyn that was using an electric bagpipe before anyone knew what Williamsburg was.
You're always standing somewhere. A terminal window is always inside a folder, the same folders you see in Finder. Type ls (list) and it tells you what's around you. Type cd Documents (change directory) and you walk into your Documents folder. Type cd .. and you step back out. It's Finder without the pictures. Or, closer to the truth, it's a text adventure game: you are in a folder called Desktop. You see 14 files.
This matters because commands do their thing wherever you're standing. It's the first thing that trips everyone up, and it's also the key to everything later: when you run Claude Code inside a folder, that folder becomes its workspace. Where you stand is what it sees.
That's it. Conversation plus location. Put those together and you realize the terminal is basically an old school LLM. The same type-and-it-answers loop, except instead of natural language it demands a special syntax, and instead of inferring your context it has to be told exactly where you're standing. The world's most pedantic chatbot. Everything else is vocabulary, and you'll see in a minute why you don't need to memorize the vocabulary.
Now, the actual install.
First, open the Terminal. Press Cmd+Space, type "terminal", hit return. It's been on your Mac this whole time, sitting there since the day you unboxed it.
What you'll see is an empty window with a short line of text and a blinking cursor. That's it. It can't do anything until you type, and you can't break something by looking at it. That cursor has been blinking the exact same way for about fifty years, which goes to show that some things don't need a redesign.
Now, the copy-paste. Copy this, paste it into that window, and hit return:
curl -fsSL https://claude.ai/install.sh | bash
In human terms, this line says: "Go to claude.ai, pick up the installer, and run it." That's all. No Homebrew, no Node, none of the developer machinery you may have heard muttered about (if those words mean nothing to you, even better; they're not invited).
Text will scroll by for a bit. Let it. When the blinking cursor comes back, it's done. You just installed a command line tool. You did it!
Next, give your projects a home. Before you launch anything, one bit of housekeeping that future you will thank you for. Everything you build with Claude lives in a folder somewhere, and you don't want that to become a scatter of mystery folders all over your Mac. So make one folder where all your Claude work will live, with one folder per project inside it. Paste these in one at a time, hitting return after each:
mkdir -p ~/claude-projects/first-project
cd ~/claude-projects/first-project
Two new words, but you can already read them. mkdir means make directory: it creates a claude-projects folder with a first-project folder inside it, in one go (the ~ is shorthand for your home folder). And cd you already know. You just walked in. A tip while we're here: keep folder names lowercase, with hyphens instead of spaces. The terminal takes spaces personally.
Don't believe these are the same folders Finder sees? Type open . (open, space, period) and the folder you're standing in pops open in Finder. Same place, two views.
Btw, you could create the folders in Finder and then just point the Terminal to them via cd. Both work, it's whatever is most comfortable for you.
From here on, that's the ritual: new project, new folder inside claude-projects, walk in, then start Claude. Where you stand is what it sees.
Then, say the magic word. Type this and hit return:
claude
Your browser will pop open and ask you to log in with your Claude account. You'll need a Pro or Max plan for this part; the free plan doesn't include Claude Code. Log in, come back to the terminal, and there it is: a welcome screen, in the black window, with your name on it.
(If instead you get command not found, don't panic. Quit Terminal and open it again; it doesn't know a new tool exists until it restarts. A fresh window also forgets where you were standing, so cd back into your folder and type claude again. Computers gonna computer.)
Before you type anything else, notice what just happened. The window looks the same, same black rectangle, same blinking cursor, but Claude has taken over. You're not talking to the terminal anymore; you're talking to Claude, inside the terminal. Which means the pedantic rules no longer apply. Natural language works now, and typos are forgiven. You can ask what a command means before you run it, or tell it to go look at a folder somewhere else entirely, no cd required. The old school LLM just got its natural language upgrade.
Finally, ask it for something. Anything. You're standing in your first project folder with Claude running inside it, so try this: "Create a page called hello.html with my name on it in ridiculously big type, then open it in my browser."
Watch what happens. It won't just do it; it'll ask permission first. Approve it. And a few seconds later your browser opens: a page, with your name on it, that exists because you described it. It's sitting in first-project right now, exactly where you were standing.
That page is the ah ha moment, by the way. The first time the machine makes a thing because you said the thing. In last week's post I called this drawing with words, and I spent two thousand words on what it means for our discipline. But reading about it is like reading about swimming. This is me pushing you into the pool.
Should you learn more terminal after this? Commands, flags, all that? Honestly, you don't have to. I'm no shell wizard myself; I still look up how to unzip certain files, true story. The beautiful twist is that the tool you just installed is also the best terminal teacher ever made. Don't know how to do something? Ask Claude, in the terminal, in plain English. The scary black window now speaks your language.
That's pretty much it! The terminal is the oldest interface on your Mac, older than the Mac itself, actually. And it now runs the most advanced tool I've ever used. Funny how that works.
The cursor's blinking… Type something.
One more thing, and it's a shameless plug. Once this clicks, you'll end up running a few Claudes at once, and juggling black windows gets old fast. And honestly, typing in the terminal is the one truly archaic part of all this; editing a sentence in there feels like carrier pigeon fuel. So I built Rigadigdig, a macOS app that runs many Claude Code sessions side by side, where typing feels like any normal text editor and a calm sidebar shows which Claude needs you. File it away for the day you wanna level up.