Workflows / Mobile app
Build a mobile app with AI coding tools
Mobile adds things web does not: app store rules, device testing and slower release cycles. This workflow picks Expo so you can test on your phone early, and keeps the AI focused on one screen at a time.
The steps
Recommended tools link to their Agentdex pages. These are starting points, not rankings: pick what you already pay for.
Plan
Screens, navigation and data for v1, written down.
Use
Claude Code. Strong at turning a messy idea into a structured plan file you keep in the repo.Or try:
Cursor,
Gemini CLIHelp me plan a mobile app. Ask about the user and core job, then write PLAN.md: list of screens, navigation map, data model and what needs offline support.
Design
Rough screen layouts you can react to.
Use Figma. The standard place to sketch screens; your AI tool can work from exported frames.
Or try:
v0, ExcalidrawDescribe each screen in PLAN.md as a simple wireframe: header, main content, primary action. Keep tap targets at least 44px.
Build
A running Expo app with real navigation and state.
Use
Cursor. Multi-file edits and rules files help keep a React Native codebase consistent.Or try:
Windsurf,
Claude CodeUsing Expo and TypeScript, scaffold the screens in PLAN.md with Expo Router. One screen per commit. Explain any native module you add and why.
Test
The core flow works on a real device and on both platforms.
Use Expo. Expo Go lets you open the app on your own phone while you build.
Write a manual test script for the core flow on iOS and Android, plus unit tests for the data layer. Flag anything platform-specific.
Ship
Store-ready builds and listing copy.
Use
Claude Code. Good for drafting store listings and walking through release checklists.Or try:
GitHub Copilot,
Gemini CLIWrite the App Store and Play Store listing: name, subtitle, short and long description, and a privacy summary based on what the app actually collects.
Maintain
Crash reports in, small fixes out.
Use Sentry. Crash and error reporting with a free tier and a React Native SDK.
Or try:
GitHub Copilot, JulesHere is a crash report. Find the likely cause in the code, propose the smallest fix and a test that would have caught it.