Learn Claude Code: from your first prompt to your first paid build
Written by Waqas Munawar. Last updated 18 September 2026
Claude Code is a coding tool you talk to in plain English. You tell it what you want, it writes the files, runs the commands and shows you what changed before anything is saved.
You do not need to know how to code to start. You do need to know what you want built, and you need to be able to tell when it got it wrong. That second part is the actual skill, and most of this page is about it.
What Claude Code actually is
It is a program that runs in your terminal. That is the black window on your computer where you type commands instead of clicking things. If you have never opened it, that is fine, you will use about four commands on this page.
The important difference from a chat window is what it can reach. Claude in a browser can only see what you paste into it. Claude Code reads your project files itself, edits them, runs commands, reads the error that comes back, and fixes it. You are handing over the work rather than the question.
It also runs in VS Code and JetBrains, as a desktop app on Mac and Windows, and in your browser at claude.ai/code. Start with the terminal anyway. It is the version every tutorial assumes, including mine.
What you need before you start
Three things, and you probably have two of them:
- A terminal. Terminal on Mac, PowerShell on Windows. Already installed on both.
- A folder to work in. An empty one is fine. This is your project.
- A Claude account. Pro, Max, Team or Enterprise, or a Claude Console account with credits.
You do not need Node, Python, or any of the things a tutorial from last year told you to install first. Git is worth having early, for a reason I get to further down.
Your first twenty minutes with Claude Code
Here is how it works, start to finish.
1. Install it
On Mac, Linux or WSL, paste this into your terminal:
curl -fsSL https://claude.ai/install.sh | bash
On Windows, open PowerShell and paste this instead:
irm https://claude.ai/install.ps1 | iex
Then check it landed:
claude --version
You should get a version number followed by (Claude Code). If you get "command not found", close the terminal, open a new one, and try again. That fixes it nine times out of ten.
2. Log in
Type claude and press enter. It opens your browser and asks you to sign in. You do this once.
3. Open your project
Move into the folder you want to work in, then start a session there:
cd my-project
claude
Whatever folder you are standing in is what it can see. That is worth knowing on day one, because it is the reason people wonder why it cannot find their files.
4. Ask before you build
Your first message should not be a request. Ask it to look around first:
what does this project do?
Then ask for one small thing. Not the whole app. One thing:
add a page that says hello and shows today's date
It will show you the change before saving it. Read that. Say yes. You now have a working build and you have seen the whole loop: you ask, it plans, it edits, you approve.
The one file that changes everything: CLAUDE.md
Here is the thing nobody tells you when you start.
Claude Code does not remember you. Every time you open it, it is a new person walking into your project for the first time. They are good, they are fast, and they know nothing about what you decided yesterday.
A file called CLAUDE.md in your project folder fixes that. Every session reads it before it does anything. Whatever is in there is already in their head.
Make an empty text file with that exact name and put four things in it:
- What the project is, in two sentences.
- The rules it must follow. Things like which framework to use, or never to touch a certain folder.
- The commands that run and test the thing, so it does not have to guess.
- Anything you have now told it twice.
That last one is the rule that matters. The second time you correct the same mistake, stop correcting it and write it down instead. Thirty seconds of typing saves you the same argument every session for the rest of the project.
What to build first
Pick something with an end.
That sounds obvious and it is the single most common reason people stall. "Build an app" has no finish line, so there is never a day when you are done and there is never a thing you can show anyone. Something small and finished beats something ambitious and abandoned, every time.
Good first builds: a one page site that takes bookings. A script that renames three hundred files properly. A form that emails you when someone fills it in. Each of those is an evening, and each of them is a thing somebody would pay for.
Four things that cost beginners the most time
Asking for too much in one message. "Build me a booking system with payments and login" gets you a pile of code nobody understands, including you. Ask for one piece, check it, then ask for the next.
Not reading what it wrote. Approving changes on autopilot feels fast for about two days. Then something breaks and you are debugging a project you have never read. Skim every change. You do not need to understand every line, you need to notice when it touched something you did not ask about.
No CLAUDE.md. Covered above, and still the most skipped step.
Starting a new project every time one gets hard. I did this for about a year. I have a folder of them. The project that taught me the most was the one I stayed on after it stopped being fun, which is not the advice anyone wants and is the only part of this that actually changed my results.
Getting it to work, and getting paid for it
These are two different skills, and the second one is where most people who learn Claude Code stop.
You can build something genuinely useful in a weekend now. Finding the business that needs it, proving they need it, and putting a number on it is a separate job that no coding tool does for you.
If you want the whole method written out, I put one of them in a free playbook: find a company losing work to slow follow-up, prove it in 48 hours, build the fix, and charge for a 14 day pilot. It is the Quote Follow-Up Playbook, and it costs nothing.
Build something with it
Build an n8n lead response system that replies in seconds All the build notesQuestions people actually ask
Do I need to know how to code to use Claude Code?
No. It writes the code. What you need is a clear idea of what you want and the patience to read what came back. Knowing some code helps you spot a bad answer faster, and you pick that up by watching it work on your own project.
What does Claude Code cost?
The tool is free to install. You need a Claude subscription to use it, on the Pro, Max, Team or Enterprise plan, or a Claude Console account with credits. Prices change, so check the Claude pricing page rather than trusting a number on a blog post.
Is Claude Code the same as using Claude in a browser?
No, and the difference is the whole point. In a browser you paste code in and copy an answer out. Claude Code opens your files, changes them, runs the commands and reads the errors. You are handing over the work, not the question.
Does Claude Code work on Windows?
Yes. There is a PowerShell installer, and it also runs under WSL. Installing Git for Windows alongside it is worth doing, because it gives Claude Code a proper shell to work in.
Can Claude Code break my project?
It can change files you did not expect. Two habits fix it. Put the project in Git before you start, so any change can be undone in one command. And read the diff it shows you before approving, rather than approving on autopilot.
Where to go next
Install it tonight and give it one small job on a folder you do not care about. That is the whole first session. You will know within twenty minutes whether this is for you.
After that it stops being about the tool. It becomes about picking builds that finish and charging for them, which is what I teach inside my Skool community, The AI Builder's Lab. It is $9 a month. The chapters cover websites, AI video, real business playbooks and getting paid, and every one ends in what the thing you made is worth and who pays for it.