AI tools for mobile apps.
MobAI helps AI agents work with real iOS and Android apps. Agents can see the screen, read the UI tree, tap, type, swipe, run flows, debug apps, and verify changes on real devices, emulators, and simulators.
The goal is simple: bring the AI coding loop to mobile development.
Desktop app that connects AI coding agents to mobile devices.
MobAI works with:
- iOS and Android
- physical devices
- Android emulators
- iOS simulators on macOS
- Claude Code, Cursor, Windsurf, Codex, and other AI tools
- MCP, HTTP API, local agent workflows, and Claude Code plugins
Use it to:
- test mobile apps with natural language
- let agents inspect and fix UI issues
- automate repetitive mobile workflows
- generate screenshots
- debug native iOS apps
- build reusable app skills
MCP server for MobAI.
Use MobAI from MCP-compatible tools like Claude Code, Cursor, Windsurf, Cline, and others.
A mobile harness inspired by browser-use/browser-harness.
It lets agents learn a workflow inside a mobile app, save it as a reusable skill, and run it again later.
In practice, it helps create a skill layer for mobile apps, even when the app does not provide a real API.
Build and debug iOS apps from Windows or Linux using GitHub Actions and a real iPhone.
Useful for Flutter and other mobile developers who do not want to switch machines just to build and install an iOS app.
Experimental Docker-based iOS build pipeline.
It uses your own Xcode.xip to extract the needed toolchain and SDK files. No Apple SDK redistribution.
A small bridge for sending context to terminal-based AI coding agents.
Works with tools like Claude Code, Codex, and Gemini CLI.
Chrome extension for selecting UI elements and sending rich context to AI tools.
It captures screenshots, bounds, HTML, CSS, and accessibility metadata.
Swift package for using a ChatGPT account to make API-style calls from iOS apps.
Includes OAuth flow, token storage, refresh logic, and a small client.
AI coding agents can write code, but mobile apps are still hard for them because they usually cannot see or use the app.
MobAI closes that loop.
The agent can:
- launch the app
- inspect the current screen
- interact with the UI
- reproduce bugs
- change the code
- verify the result on a real device or simulator
That makes agents much more useful for mobile development.
- Website: mobai.run
- GitHub: github.com/MobAI-App
- Contact: contact@mobai.run