SV EN
Building with AI · Guide

How I built a real app without knowing how to code — step by step

Familjegaraget, my family-logistics app, started as an idea at the kitchen table and ended up as an app submitted to the App Store. Here's the whole journey — with every tool and service I used at each step — written for you: not a techie, but wondering whether you could do the same. (Spoiler: you can.)

Christian, Advanto July 2026 About a 12-minute read

I've worked on IT projects for over 20 years — but as a project manager, not a developer. I've never been able to code. And yet I've built Familjegaraget: a working family app with logins, a points system for chores, bookings for the family's stuff and real-time updates between everyone's phones. Built on evenings and weekends, published on the web and submitted to Apple's App Store.

This isn't an article about how amazing AI is. It's a manual. I'll walk through exactly which tools I used, in what order, what they cost and what you need to know (very little). If you can write an email and describe what you want, you've got the background you need.

~30 h
total build time, evenings and weekends
0
lines of code written by me
~SEK 1,500
in yearly costs for the services*

Step 1Talk the idea through with an AI assistant

It all started in an ordinary chat. I described our family problem: chores nobody wanted to do, things getting double-booked, endless nagging about screen time. I asked the AI to help me think: what should an app like this be able to do? What's the smallest possible version? What's hard and what's easy?

This step matters more than it sounds. Don't skip it. An hour of good conversation here saves ten hours later. Ask the AI to play devil's advocate: "What have I missed? What's going to get messy?"

Tools in this step
Claude (claude.ai)
AI assistant from Anthropic. Free to start with; the paid version costs about SEK 200/month and you'll want it once you're building for real. This is where I kicked the idea, the requirements list and the priorities around.

Step 2Get the tool where the app gets built

The next step is to download Cursor — a program that looks like a code editor (because it is one), but with one decisive difference: it has a built-in AI that writes the code for you. You type what you want, in plain English, in a chat box, and the AI creates and changes the files in your project.

Don't let the technical-looking interface put you off. You'll really only use two things: the chat box and the button that previews your app. The code scrolling past is something you never need to understand — but it's still fascinating to watch it take shape.

Tools in this step
Cursor (cursor.com)
AI code editor. About SEK 200/month. This is where all of Familjegaraget was built. Inside Cursor I picked Claude as the AI model — the same "brain" as in step 1, only now with its hands on the keyboard.
Tip: Paste your whole requirements list from step 1 as the first message in Cursor and write: "Build a first simple version of this. Explain what you're doing as if I'd never seen code." The AI then sets up the project itself, with the right technology.

Step 3Build the app — one feature at a time

Familjegaraget is built with something called Next.js — a framework for modern web apps. You don't need to understand it. It's enough to know that the AI chose it for me and that it's the same technology real companies use. Your job isn't to know the technology, but to be the client: describe, test, give feedback.

My way of working quickly became a rhythm: ask for one thing at a time ("now I want the kids to be able to tick off a chore and earn points"), look at the result in the preview, and describe what should change ("the button's too small, and the points should show straight away without reloading the page"). Just like giving feedback to a consultant — except the consultant answers in three seconds.

When something broke (and it does), I pasted the error message into the chat and wrote "this happened, fix it". That solved 95 percent of all problems. The remaining 5 percent got solved by asking the AI to "start over with a different solution to this problem".

Tools in this step
Cursor + Claude
Same as in step 2 — now in intensive dialogue. This is the step where most of my ~30 hours went.
Next.js
The technology the app is built in. Free and open source. You don't choose it yourself — the AI does — but it's good to recognise the name.

Step 4Give the app a memory: database and login

An app without a database forgets everything when you close it. For the family's chores, points and bookings to be saved — and for each family member to log in with their own account — I needed a database service. I used Supabase.

It sounds advanced but was surprisingly simple: create a free account on their website, copy two long keys (like passwords for the app) and paste them where the AI tells you to. Then the AI handles the rest: it creates the tables, wires up the login and even builds the real-time feature that means when a child ticks off a chore, the parent's screen updates instantly.

Tools in this step
Supabase (supabase.com)
Database, login and real-time updates in one. Free for small projects; Familjegaraget got by on the free tier for a long time.

Step 5Publish on the web — with your own address

As long as the app only exists on your computer, nobody else can use it. Publishing was handled by Vercel — a service that takes your code and puts it on the internet, automatically, every time you make a change. The AI in Cursor guided me through the whole setup.

I also bought the domain familjegaraget.se from Loopia, a Swedish domain provider, and pointed it at Vercel. That too with the AI as my guide: I wrote "I've bought a domain from Loopia, how do I connect it?" and got a step-by-step list to follow, with the exact settings.

Tools in this step
Vercel (vercel.com)
Hosting — where the app "lives" on the internet. Free for hobby projects.
Loopia (loopia.se)
Swedish domain registrar. A .se domain costs a few hundred kronor a year.

Step 6Teach the app to send email

When someone creates an account or forgets their password, the app needs to be able to send email. For that I used Resend — a service built precisely so apps can send email. I asked the AI to connect Resend to Supabase and write all the email templates in Swedish, in the app's visual style. One evening's work, done.

Tools in this step
Resend (resend.com)
Email sending for apps. A generous free tier that's plenty for a family app.

Step 7Test with real users — and fix the bugs

This step needs no new tool, but it's the most important of all: let real people use the app. We started with our own family, then pilot families. Every "this is odd" and "why did my booking disappear?" became an item on a fix list that I worked through in Cursor, exactly like step 3.

My experience after 20 years in IT projects holds here too: the bugs you find yourself are half of them. Users find the rest. Plan for it — that's normal, not a failure.


Step 8Turn the web app into an iPhone app

Now it got a notch more technical — but still perfectly doable with the AI as a guide. A web app can be packaged into a "real" app with a tool called Capacitor. Put simply, Capacitor builds a thin shell around your web app so it can be installed from the App Store like any other app.

To build iOS apps you need two more things: a Mac (Apple's own requirement — it won't work on Windows) and the program Xcode, which is free and downloaded from the Mac App Store. The AI wrote all the Capacitor settings for me; my job was to click the right buttons in Xcode according to the instructions and test the app — first in a simulated iPhone on the computer screen, then on my actual phone.

Tools in this step
Capacitor (capacitorjs.com)
Free tool that packages the web app into an iOS app (and Android, when the time comes).
Xcode
Apple's development program. Free, but needs a Mac.
Apple Developer Program
Membership required to publish in the App Store. USD 99/year — the only genuinely unavoidable cost in the whole journey.

Step 9Prepare everything Apple asks for

Apple doesn't let just anything into the App Store, and there's a checklist to get through. No single item is hard, but there are a lot of them. Here's what I did — with AI help on every point:

App icon. The icon is needed in a whole set of sizes. The AI generated the entire pack from a single original image.

Privacy policy and support page. Apple requires public web pages for both. The AI wrote them in Swedish, in the app's visual style, and they now live on familjegaraget.se.

App Store Connect. Apple's portal where the app's "store page" is filled in: description, keywords, screenshots, age rating and so-called Privacy Labels — a declaration of what data the app handles. I wrote the description texts together with the AI and filled in the declaration with the AI as a sounding board for every question.

Tools in this step
App Store Connect (appstoreconnect.apple.com)
Apple's portal for everything around publishing. Included in the developer membership.
Claude
Wrote the policy texts, the store description and helped me interpret Apple's requirements and guidelines.

Step 10Submit — and wait for Apple's review

The final step: in Xcode you choose Product → Archive, which packages the finished app, and then Distribute App, which uploads it to App Store Connect. There you pick your build, attach a demo account so Apple's reviewers can log in and test, write a few lines to the reviewer — and press the button.

Then you wait. Apple reviews every app by hand, usually within a few days. Sometimes the app goes through straight away, sometimes you get a note to fix and resubmit. That's part of the process, not a verdict on you.


What does it cost — and what does it really take?

Add it up: Claude and Cursor at about SEK 200/month each while you're actively building, Apple's developer membership at USD 99/year, a domain for a few hundred kronor a year. Supabase, Vercel, Resend and Capacitor get by on the free tiers for a project this size. All told, it lands around SEK 1,500 a year in fixed costs, plus the AI subscriptions for the months you build.

What it does take is something other than money and technical know-how: clarity. The ability to describe what you want, test carefully and give concrete feedback. Which are, perhaps unsurprisingly, the same abilities that make a good client on any IT project. The difference is that your "supplier" now answers in seconds, never gets annoyed at changes and costs a thousand kronor a month.

My best advice: Don't start by learning the tools. Start with a problem you genuinely have yourself — then you know exactly how the app should work, and you'll have the stamina to see it through. Familjegaraget exists because my own family needed it.