Your First Day With Claude Code

I recently got on a Zoom call with a friend who wanted to learn Claude Code. He had zero coding experience. Never opened a terminal. Didn't know what VS Code was. I figured we'd spend an hour and get him set up.

It was messy. We stumbled through terminal commands, fought with slow hotel Wi-Fi, and when things broke, we screenshotted the errors and fed them right back to Claude to figure out what went wrong. But by the end of the call, he had a working setup.

So I figured I would turn the whole thing into a guide. Every step below is what we actually did on that call, in the order we did it.

What You Need

A Mac, an internet connection, and about 30 minutes of patience. That is really it. You do not need to know how to code or understand what any of this stuff does yet. You just need to be willing to type a few commands and trust the process.

Step 1: Meet the Terminal

Hit Command-Space on your Mac to open Spotlight, then type Terminal and hit enter. A window opens with a blinking cursor. This is your computer, just a different way of looking at it. All the same files and folders you are used to, just text instead of icons.

Here are the three commands that matter right now:

pwd tells you where you are. It stands for "print working directory." Type it and hit enter:

run this in your terminal

pwd

You will see something like /Users/yourname. That is your home folder.

ls lists everything in the current folder:

run this in your terminal

ls

You should recognize some of those names: Desktop, Documents, Downloads. Same stuff you see in Finder.

cd changes into a different folder. A folder is a directory. Type this and hit enter:

run this in your terminal

cd Desktop

Now type ls again. You should see the files that are sitting on your actual desktop right now.

To go back up one level:

run this in your terminal

cd ..

That moves you up to the parent folder.

On the Zoom call, I had my friend do exactly this. Navigate to the Desktop, list the files, see that they matched what was actually sitting on his screen. That is the moment it clicked for him. The terminal is not some secret hacker dimension. It is Finder without the pictures.

Step 2: Download VS Code

Go to Google and search for VS Code for Mac. Download it, drag it to Applications, and open it.

VS Code is a code editor. Think of it like a Word document for code, except it also has a terminal built right into it. That built-in terminal is super important because it means you can see your files on the left side and type commands on the bottom. Everything in one place.

When you first open VS Code, it shows a Welcome tab. Close that. Then go to the menu bar at the top and click Terminal, then New Terminal. Drag that terminal panel up so it takes about a third of the screen.

Quick tip about Macs: the menu bar at the very top of the screen changes depending on which app is selected. If you click on Finder, you get Finder menus. If you click on VS Code, you get VS Code menus. It is easy to accidentally be in the wrong app and wonder why the menus look different.

Step 3: Set Up Your Project Folders

Open Finder and navigate to your home folder. The one with your name. Right-click, make a new folder, and call it apps. All lowercase. This is where all your projects are going to live.

Inside apps, make a folder for your business or agency. Then make one for a client. Keep the names lowercase, no spaces, just a coding convention that will save you headaches later.

Now here is where it gets cool. Go back to VS Code, click File, then Open Folder, and open that apps folder. You should see your business folder and your client folder on the left side. Drag a screenshot or any file from Finder into one of those folders. Watch it show up in VS Code instantly.

Same files, just two different ways to look at them.

Step 4: Install Homebrew

Homebrew is like an app store for the terminal. Instead of going to a website and clicking a download button, you type one command and it handles everything.

Open the terminal in VS Code and go to brew.sh in your browser. Copy the install command from the website and paste it into your terminal. Hit enter.

This is going to take a few minutes. On the call, we were on hotel Wi-Fi and this part took a while. Your Mac might also ask you to install Command Line Tools for Xcode, which is a bundle of developer tools Apple provides. Just say yes and let it download. Enter your Mac password when it asks. It will not show the characters as you type. That is normal. Just type it and hit enter.

When it finishes, Homebrew might tell you to run two extra commands to add it to your path. Copy those and run them. This is just telling your terminal where to find Homebrew.

Step 5: Install Node.js

Claude Code runs on Node.js. You need it installed before Claude Code will work. With Homebrew set up, this is one command:

run this in your terminal

brew install node

Wait for it to finish.

On the call, we tried to install Claude Code first and got a "command not found" error. We screenshotted the error, pasted it into Claude Desktop, and it told us we needed Node. That is actually the right instinct going forward. When something does not work, screenshot it and ask Claude. It has seen every error message a thousand times.

Step 6: Install Claude Code

Now the good part:

run this in your terminal

npm install -g @anthropic-ai/claude-code

When it finishes, launch it:

run this in your terminal

claude

The first time you run it, it will ask you to sign in and authorize it. Follow the prompts. This is the moment where a bunch of text starts flying across the screen and you feel like you are in a movie. That feeling is correct. You are about to have a conversation with an AI that can see, edit, and create files on your computer.

Step 7: Write Your First CLAUDE.md

Before you start giving Claude Code instructions, you want to set up a file called CLAUDE.md in your project folder. This is like an SOP for the AI. It reads this file every time you start a conversation, so it always knows the context.

On the call, we started with a simple Apple Note. Just a few lines:

You are an expert content writer for an electrician company in Las Vegas, Nevada. You always write in a friendly, approachable way.

That is it. That is a starting point. You are telling the AI who it is, what it does, and how it should behave. Just like you would train a new employee.

We even added a silly test instruction: "You use the word 'like' three times in every sentence." That was just for the sake of example, to prove that it actually follows what you tell it. And it did. Every sentence came back stuffed with "like." The instructions work.

Over time, you will add more. I have one for my own content that runs thousands of words. It was built from voice-to-text transcripts of my YouTube videos. Claude analyzed 50,000 words of how I actually talk and reverse-engineered my communication style. The result is a voice profile that captures my personality, my vocabulary, how I open and close content, even my go-to phrases. So when Claude writes for me, it actually sounds like me.

But you do not need all that on day one. Start with three lines. Expand it as you learn what matters.

Step 8: Build Something

This is the magic moment. Open VS Code, make sure your terminal is pointed at one of your client folders, and type claude. Then tell it what to build.

On the call, we were setting up a content writing system for an electrician. But it could be anything. Tell it to build a homepage. Tell it to write a service page. Tell it to create a landing page for a specific offer. Give it whatever instructions are in your CLAUDE.md and say go.

What happens next is wild. Claude Code starts creating files. You can watch them appear in real time in VS Code's file explorer on the left side. HTML files, content files, stylesheets. It is building an actual website right in front of you.

Inside those client folders in apps, those files could eventually be the actual pages that get uploaded to the internet. Real website files, managed from the same place you are having a conversation with an AI.

The Moment It Clicks

On the call, there was a moment about 45 minutes in. We had gone from "What is Terminal?" to having Homebrew installed, Node running, Claude Code authenticated, and a project folder with real files in it. My friend looked at the screen and said: "So by using these notes, I'm basically creating a custom chat."

Yes. Exactly. Each project folder with its own CLAUDE.md is like a custom GPT, except it actually lives on your computer, has access to your files, and can build real things instead of just chatting about them.

The other insight that came up was about version control. I explained how GitHub tracks every change to every file, so you can see how your SOPs evolve over time. He had this moment where he realized: "I'm not just writing instructions. I'm building a system that gets smarter every time I update it."

That is the shift. You go from "I use AI to write content" to "I have a system that knows how my business operates." And it just keeps getting better the more you put into it.

What Is Next

You have got Claude Code installed, a project folder set up, and your first CLAUDE.md written. Here is where to go from here:

Start adding real instructions to your CLAUDE.md. How do you write title tags? What is your meta description format? What tone do you use? Give it examples of your best work and tell it to reverse-engineer your style. The more specific you are, the better the output gets.

If you want to go deeper into what Claude Code can actually do once you are up and running, I wrote a full breakdown of how I use it to run my entire content operation from writing articles to managing YouTube to deploying websites. That article covers the advanced stuff: content pipelines, propagation, SOPs that build themselves.

For faster input, check out Wispr Flow. It lets you dictate your prompts at 120 words per minute instead of typing at 40. When you are iterating quickly with Claude Code, talking is way faster than typing.

And if you are curious about the deeper architecture of how to make AI reliable across sessions, I wrote about splitting your AI workflow into two layers. The short version: let scripts handle the boring repeatable stuff, let Claude handle the thinking. That separation is what makes the whole thing actually work at scale.

The setup is the hardest part, and you already got through it.