Your First Real App with Claude Code (No GitHub Required)

I’m on the far right side of the “working with AI” spectrum - I’m an advocate and I enjoy working with it and I enjoy sharing the knowledge with others. The whole point of these computer things is that they are supposed to make our lives easier, anything that helps with that is a win in my book.

Have you found yourself thinking something similar to this:

“I want to make an app with Claude, but the thing it builds always gets redone, and I don’t want to sign up for GitHub or learn about repos and all that other stuff. I just want to see what this is about!”

That’s a completely fair thing to want to avoid. If you’re curious, follow along with this article and you’ll have an app running on your local machine by the end. If you can follow a recipe, you can build a real, working web app on your own computer in about ten minutes, most of which is spent watching things install. You don’t need GitHub, Git, AWS, Azure, Base44 or any of the low code platforms. You don’t need to know what those are yet, we’ll get there at the very end, and only if you want to.

For this guide I’m assuming you are using Claude/Claude Code. I recommend trying one month of the $20 subscription for experimenting! The base Claude you use in the web based chat or in the desktop chat window builds artifacts - quick previews that live only inside that conversation. Claude Code is the same Claude but it’s running with more control on your computer instead of in a chat window. Claude Code can create and modify real files on your computer that don’t vanish when you start a new conversation. Claude Code is what’s called a coding harness.

You can get it from claude.ai/code and there are two ways to use it: a friendly desktop app with buttons and a folder picker — no terminal, nothing to type at a command line — and a terminal version for when you want it. Everything in this guide works both ways, and I’ll show you both options throughout. If the word “terminal” makes you tense up, good news: you can do this whole thing without ever opening one.

Once you experience the power of working with an AI using a coding harness, you may never go back to your chat windows. The power and flexibility of them eliminate the need for most of the “now with AI” or “connects to Claude” type of products on the market. Like fishing & hunting equipment most of those products are built to appeal to the human, not the LLM.

Step 1 — Give the work a home

Artifacts in conversations are useful but temporary. There’s no permanent copy sitting anywhere — it’s a drawing on a whiteboard that gets redrawn each time. There’s an easy fix! Give the work a folder on your computer. A folder is durable, it lasts between conversations and it limits where Claude can work. When Claude Code works inside a folder, three good things happen:

• Your files are saved for real. They sit on your hard drive like any other document. Close everything, come back next week, they’re still there.
• Claude remembers the project. It keeps notes to itself in that folder — what the app is, decisions you’ve made, how things are wired together — so the next session picks up where you left off instead of starting from a blank page.
• You can pick up and iterate. Telling Claude “Change the button to green,” a month later, works — because there’s a there there.

Create a folder on your computer in whatever way you are comfortable. If you are new, I’ve included some instructions here that will get you started. There’s nothing special about it - just a folder (also called a directory or a dir) like any other.

The no-terminal way (recommended for your first one — Windows or Mac):

1. Make an empty folder the normal way you’d make any folder. On a Mac, open Finder; on Windows, open File Explorer. Right-click in a spot you’ll remember (your Desktop is fine) and choose New Folder. Name it something like my-first-app.
2. Open the Claude Code desktop app.
3. Click the Code tab, choose Local, then click Select folder and pick the folder you just made.
4. Start typing your request. That’s it — no commands, ever.

The terminal / command line way (if you’re already comfortable there): open Terminal on a Mac, or PowerShell on Windows, and type this — invent any name in place of my-first-app:

mkdir my-first-app
cd my-first-app
claude

mkdir makes the folder (“make directory”). cd navigates into it (“change directory”) like double clicking a folder in your GUI. claude starts Claude Code, now pointed at your brand-new, empty project home. These same three commands work on both Windows and Mac. (On Windows, I recommend using PowerShell rather than the older cmd Command Prompt these days).

Either path lands you in the same spot with Claude awake and pointed at an empty folder that’s yours. From here on, everything you build lands in that folder and stays there.

Step 2 — Pick a lane

A web app can be built in a lot of languages. For your first one, I’m giving you two options that will work for 99% of your needs. You don’t have to understand either yet, you just need to pick one and I’m sharing the details on each so you have the details if you are interested.

	Python + uv	Node.js (JavaScript / TypeScript)
Good for	Data, forms, quick tools, anything “office-y”	Anything web-first; interactive pages
The upside	uv is one small tool that erases all the old Python set up headaches	JavaScript is the language browsers already speak — one language front to back
The tradeoff	The browser doesn’t speak Python, so the server hands it the page	A few more moving parts under the hood (Claude wires them up for you)

If you’re not sure, pick Python + uv. It’s the easiest on-ramp, uv is an incredibly useful tool and Python is the lingua franca of the machine learning world. You can do amazing things with it pretty quickly using things like Jupyter notebooks and Claude Code! If you already live in a web-and-JavaScript world, Node will feel more natural. Either one is a fine first choice; you are not marrying it and as you’ll learn you can quickly try out other options. Heck, try both if you’re really curious and have Claude walk you through the differences.

A word on “three tiered applications”

The phrase three tiered app is a somewhat dated technical jargon term - it describes the type of “architecture” that is common to applications like we’re making.

The UI is what you see — the page, the form, the buttons. The logic is the rules running behind it, the application itself. The data is where things are stored - a small database, a file or the like. This used to take a team of specialists to set up in a data center and with modern tooling and Claude Code it’s now a prompt and a coffee.

Step 3 — The prompt

Now the fun part! Let’s make an app. I’ve provided two sample prompts for you to use depending on which lane you picked in Step 2. Notice they both ask for the same app; only the language differs. Change “summer party” to whatever you’d like.

If you chose Python + uv:

Make me a small web app in Python. Use uv to set up the project. It’s a single page where someone can enter their name and whether they’re coming to our summer party, and it saves each response. Below the form, show the current list of who’s coming so far. Store the data in a local SQLite file. When it’s ready, start it and give me the address to open in my browser. Explain each step in plain language as you go.

If you chose Node.js:

Make me a small web app with Node (JavaScript/TypeScript). It’s a single page where someone can enter their name and whether they’re coming to our summer party, and it saves each response. Below the form, show the current list of who’s coming so far. Store the data in a small local database file. When it’s ready, start the dev server and give me the address to open in my browser. Explain each step in plain language as you go.

That’s a starting prompt for a complete three-tier app: a form (UI), the save-and-list rules (logic), and a database file (data). You don’t have to get the wording perfect. Tell Claude your goal — “I want people to RSVP and see who’s coming” — and let it choose the details.

Once you see your app and interact with it in your browser, just tell Claude if anything is off or if you want changes. Claude is INTENSELY accommodating and can easily do major rewrites / refactors with ease.

Step 4 — What you’ll see, and what’s safe to say yes to

This part can be a bit scary at first so let’s take the mystery out of it so you know what you are looking at. Claude Code doesn’t run wild, by default it asks permission before it does things, and it shows you what it’s about to do. Note - you can work with Claude Code with the --dangerously-skip-permissions option. I recommend waiting until you have a few projects under your belt.

Nothing happens without your approval. Here’s what to expect once Claude gets started.

Things it will ask to install — all normal, all safe to approve:

• The language runtime — uv and Python, or Node. This is just installing the engine your app runs on, the same way you’d install any program.
• A few small libraries — ready-made building blocks (a web framework, the database bit). Every app uses these. Approving them is expected.
• Starting a local server — at the end it launches your app and gives you an address like http://localhost:8000. “localhost” means this computer, just for you — it’s not on the internet, no one else can see it. Open that link in your browser and there’s your app.

When to slow down and read before clicking yes:

• Anything that would delete files, or touch folders outside the project folder you made.
• Anything asking for a password, an API key, or your credit card — a first local app needs none of these.
• Anything that sends your files somewhere or “deploys” before you’ve decided you want that.

The good news is you’re never stuck guessing. If a prompt looks unfamiliar, the safest move is to ask Claude directly — “What will this command do, what are the risks and what factors should I be aware of to make a decision?” — and it’ll explain in plain terms before you decide. Saying “no” is always allowed; it’ll just propose another way.

When the server starts and you open that localhost address, you’ll see your app running — your form, and the list growing as you add names. That’s a real app, built and running on your machine. And because it’s all in that folder, it’s still there tomorrow. And besides your Claude Code subscription it’s all free and open source!

Step 5 — If or when you’re ready, put it online

Everything so far runs on your computer, for you. Organizing an event, trying to set a roster for your kid’s youth soccer or record race times on a single computer and this is all you need. But when you want to share it or if it becomes useful enough that you want a backup this part gives you a few options to explore.

• Stay local. Genuinely fine. Your app and its folder live on your machine as long as you want. Nothing forces you off that shore.
• Back it up with Git + GitHub. Now GitHub can come into play. In a nutshell, GitHub helps make your work un-lose-able. People use it for source code, research notes, writing. You can use it 100% offline at first, and Claude can use it to save snapshots and revert changes in your local folder. You can also utilize GitHub as a remote repository as a backup. Ask Claude Code to “set up Git and push this to a new private GitHub repo” and it will walk you through it.
• Publish a simple site with GitHub Pages. GitHub Pages supports hosting basic web pages (no complex APIs or backend databases, yet). A real web address and no server bill.
• Go full three-tier in the cloud. Want to host a full app “in the cloud”? Amazon Web Services (AWS), Azure and Google Cloud all have personal free tiers. Ask Claude Code to walk you through getting an account set up and setting up a $10 billing guard to start. All the providers offer command line tools that Claude Code can use to set up your account.

You start with a folder and a prompt and you can get a real working app. Then, only if and when you want, you can explore backups and hosting. Nobody has to sign up for anything to build the first one.

---

That’s it, that’s the big secret. A folder to give the work a home. One of two flexible “frameworks”, Python or Node and a prompt you can paste into Claude Code or any other coding assistant.

What itch do you want to scratch? Go make your folder and see what happens. If this guide helps you or if you want help with exploring hosting drop me a line, jdostal@witekdivers.com.

Happy hacking :)