How to build a mobile app with AI tools (UK guide)

Mobile apps used to be the hardest thing a solo builder could ship. Two platforms, two languages, two sets of review rules and a store politics problem on top. In 2026 that stack has quietly collapsed — you can now build a working iOS and Android app from a UK bedroom, using AI coding tools, in roughly the same time it takes to build a web app.

This guide shows you exactly how. No Swift. No Kotlin. No abstracted no-code that traps you in someone else's platform. Just Claude Code, Expo and React Native — with real prompts, realistic UK costs in GBP, and the mistakes worth avoiding.

Why build a mobile app in 2026?

The UK app market is still enormous. Statista estimates 50 million UK smartphone users spend an average of 3 hours 40 minutes per day in apps — longer than they spend on the open web. Certain verticals are app-first by default: fitness, finance, workouts, commuting, tickets, food delivery, dating. Trying to reach those audiences with a website is fighting gravity.

Yet most indie builders still skip mobile because the old tooling was painful. That hesitation is the opportunity.

The 2026 stack for a solo UK builder

You need five things to ship a real mobile app: a framework, an AI coding tool, a backend, a payments layer and the store accounts themselves.

Layer	Pick	Why
Framework	Expo (React Native)	One codebase, iOS and Android, managed native modules
AI coding tool	Claude Code	Long-context, agentic, handles multi-file changes
Backend	Supabase	Auth, Postgres, realtime, file storage, all in one
Payments	RevenueCat + Stripe	RevenueCat wraps Apple/Google IAP; Stripe for web fallback
Dev account	Apple + Google Play	£79/year and £20 one-off

This stack will take you from zero to live in the App Store. It also scales — Notion, Duolingo and Shopify all ship React Native in production.

Avoid any tutorial that tells you to start with native Swift or Kotlin unless you specifically need deep hardware access (ARKit, complex camera pipelines, some health sensors). For 95% of UK indie apps, React Native via Expo is the right default.

Step 1: Define the smallest useful version

Before writing a single prompt, write two paragraphs. What does the app do, and who is it for? Be specific enough that Claude Code can infer the screens.

A good example for a UK fitness app:

"Ride Log is a mobile app for UK road cyclists. Users log rides by tapping start and stop. Each ride shows distance (in miles), average speed and elevation gain. Users can browse past rides and see weekly totals. One paid tier at £3.99/month unlocks route maps and export to Strava."

That is 55 words and it defines every screen: sign-in, ride in progress, ride summary, history list, paywall. You now have enough for Claude Code to scaffold a project.

Step 2: Scaffold the app with Expo and Claude Code

Install Node 20 and the Expo CLI, then open your folder in Claude Code and run a scaffold prompt.

npx create-expo-app@latest ride-log --template tabs
cd ride-log

Now open Claude Code and paste:

"This is a fresh Expo project with React Native 0.76, Expo Router and TypeScript. Set up Supabase auth with email sign-in and passwordless magic links. Create a tabs layout with three screens: Ride (start/stop timer and live distance), History (list of past rides) and Account. Use Expo Location for GPS tracking. Keep the file structure clean and co-locate hooks with screens. Use the NativeWind setup for Tailwind-style classes."

Claude Code will install packages, create the auth flow, set up the tabs, wire up NativeWind and scaffold the GPS logic. Expect 15-25 files and a first run that boots in the Expo Go app on your phone.

If you get stuck, the single most useful prompt is "Walk me through the current file structure and explain what each file does, then tell me which file owns the bug I just described." Claude Code is much more accurate when you force it to orient itself first.

Step 3: Build on a real device from day one

The Expo Go app on your phone scans a QR code from the terminal and runs your app in seconds. Do not use the iOS simulator or Android emulator as your primary testing device. Both hide three bugs that kill early reviews:

1. GPS accuracy — simulators report fake, perfect GPS. Real phones drift.
2. Keyboard behaviour — simulators never show the broken modal that appears when the keyboard covers the submit button.
3. Battery drain — you cannot feel how much your app costs until you run it for 20 minutes on a real device.

Ship on a real device from the first session. You will catch issues two weeks earlier than you would otherwise.

Step 4: Wire up the backend with Supabase

Your Expo app needs a place to store rides and user accounts. Supabase gives you Postgres plus auth plus Row Level Security out of the box, and the free tier covers up to 50,000 monthly active users.

Paste this into Claude Code:

"Set up a Supabase project with tables: profiles (id, email, created_at), rides (id, user_id, distance_miles, duration_seconds, elevation_gain_m, started_at, gpx jsonb). Enable RLS so users can only read and write their own rides. Generate a migrations folder and a seed script I can run locally."

Claude Code writes the SQL, the RLS policies and a typed TypeScript client. You copy the Supabase URL and anon key into your .env and you are connected.

UK note: Supabase lets you pick a region when you create a project. Choose London (eu-west-2) or Frankfurt for the lowest latency to UK users. Both are GDPR-friendly defaults and reduce p95 request time by 80-120ms for UK traffic compared to the US-East default.

Step 5: Add payments with RevenueCat

Apple and Google take a 15-30% cut of in-app purchases. You still want to use their billing — they handle refunds, family sharing, tax and currency conversion. RevenueCat sits on top of both and gives you a unified SDK plus a web dashboard for £0 until you hit £2,500 MTR.

Prompt:

"Integrate RevenueCat with my Expo app. I want one subscription product called 'ride_pro_monthly' priced at £3.99/month. Gate the Map View screen behind an active entitlement. Show a paywall screen when a free user taps Map. Use the RevenueCat Paywalls UI if it is installed, otherwise build a simple one that shows features, price and a single Subscribe button."

Test on a real device with sandbox credentials. Do not ship without end-to-end testing a purchase — it is the single most common reason store reviewers reject an update.

Step 6: Polish and submit

Once the app runs end-to-end, you have three pieces of polish left.

1. Icon and splash — use a design tool to export a 1024x1024 PNG and a splash image. Expo handles the rest.
2. Store listing — four screenshots per platform, 80-character headline, 4000-character description. UK-specific: include "UK" in your app name if you want to rank in British searches.
3. Review notes — write two sentences explaining how a reviewer tests your paid tier without real money. Include sandbox credentials.

Submit to Apple via eas submit -p ios and to Google via eas submit -p android. Apple reviews usually come back in 24-72 hours. Google is often same-day.

What it actually costs

Here is the full Year 1 budget for a UK solo builder shipping a real mobile app.

Item	Cost
Apple Developer Program	£79/year
Google Play Developer	£20 one-off
Expo EAS (free tier)	£0
Supabase (free tier)	£0
Claude Code	£15-80/month depending on usage
RevenueCat	£0 until £2,500 MRR
Domain for marketing site	£10/year
Total Year 1 minimum	£109
Total with 6 months of Claude Code	£199-589

No server costs. No App Store Connect upload fees. No hidden middleware.

The three traps to avoid

Most UK indie mobile projects die for the same three reasons. None of them are technical.

1. Shipping without a paid tier. Reviewers are more sceptical of free apps. They assume the business model is data harvesting and scrutinise privacy disclosure harder. Even a £0.99 one-off unlock signals intent.
2. Skipping the physical device. GPS, keyboards and battery drain behave differently. You will find out from users — the worst possible time — instead of in week one.
3. Launching without a marketing site. App Stores are not search engines. Ship a simple site (Lovable takes 30 minutes) with a deep link, an explainer video and a privacy policy. Conversion triples when users can read about the app before downloading.

Where to go from here

If you have an app idea you could ship this month, the bottleneck is no longer knowledge — it is decision fatigue. Pick Expo, pick Claude Code, pick Supabase, pick RevenueCat. Run the scaffold prompt tonight and commit to a weekend.

If you do not have an app idea yet, browse our weekly reports for validated UK opportunities with keyword data, competitor analysis and builder prompts — many of them are mobile-first.

---

Get the weekly IdeaStack report — every Thursday we publish one deeply researched UK business opportunity with real keyword data, competitor analysis, SERP landscape, scoring frameworks and copy-paste builder prompts. Read the latest free report.