Mobile Next

by mobile-next

Community Mobile Development 6k likes

A platform-agnostic MCP server for automating and testing native iOS and Android apps: tapping, swiping, typing, and reading the screen on simulators, emulators, or real devices through one interface.

Run with npx

npx @mobilenext/mobile-mcp@latest

Source: https://github.com/mobile-next/mobile-mcp

About

Mobile-MCP gives an agent the same automation surface for iOS and Android without needing separate platform-specific knowledge for each. It reads what's actually on screen primarily through native accessibility trees, which is fast and lightweight, and falls back to screenshot-based coordinate analysis when accessibility data isn't available for a given app, so it can still operate on apps that don't expose proper accessibility labeling.

The tool surface covers the full loop an automated test needs: discovering available simulators/emulators/devices, managing app lifecycle (install, launch, terminate, uninstall), listing on-screen UI elements with coordinates, tapping/double-tapping/long-pressing/swiping, entering text, pressing hardware buttons, taking screenshots or recording screen video, checking orientation, and pulling device crash logs. It's designed to be called by general-purpose coding agents, including Cursor, Claude, GitHub Copilot, Windsurf, and Gemini, rather than shipping its own dedicated test-runner UI.

Key features

  • Single interface for both iOS and Android across simulators, emulators, and real devices
  • Native accessibility-tree reading with screenshot-based fallback for apps lacking accessibility data
  • Full app lifecycle control: install, launch, terminate, uninstall
  • Gesture support (tap, double-tap, long-press, swipe) plus text entry and hardware button presses
  • Screen recording and screenshot capture, plus device crash log retrieval
  • Works with any MCP-compatible coding agent, not a proprietary client

Use cases

  • Automating an end-to-end mobile test suite across both iOS and Android from one agent-driven script
  • Letting an agent reproduce and screen-record a reported bug on a real device or simulator
  • Scraping or verifying UI state in a native app where no API exists to check directly
  • Pulling crash logs automatically after an agent-triggered test run fails

Available tools

mobile_list_available_devices

Discovers connected simulators, emulators, and physical devices.

mobile_launch_app

Launches an installed app on the target device.

mobile_click_on_screen_at_coordinates

Taps the screen at given coordinates.

mobile_swipe_on_screen

Performs a directional swipe gesture.

mobile_list_elements_on_screen

Enumerates visible UI elements with their coordinates.

mobile_list_crashes

Retrieves crash logs recorded on the device.

Frequently asked questions

Does it work without accessibility labels on the app?

Yes. When accessibility data isn't exposed, it falls back to screenshot-based analysis and coordinate taps instead.

Do I need separate setups for iOS and Android?

No. The same MCP server and tool set work across both platforms and across simulators, emulators, and real devices.