Mobile Device MCP
An MCP server that lets AI agents control iOS and Android devices (tap, scroll, type, take screenshots, read UI trees, and run code). Works with multiple devices at the same time.
README
Mobile Device MCP
An MCP server that lets AI agents control iOS and Android devices (tap, scroll, type, take screenshots, read UI trees, and run code). Works with multiple devices at the same time.
How It Works
Three-layer architecture:
- On-device servers — Lightweight HTTP servers running on each mobile device (UIAutomator on Android, XCUITest on iOS) that expose the accessibility tree and accept interaction commands.
- UI tree filter — Normalizes raw UI trees from both platforms into a unified flat element list.
- MCP server — The external interface. Handles device discovery, bootstrapping, port allocation, and proxies requests to on-device servers.
Devices are bootstrapped on first use — the server installs the driver app, allocates a port, starts the on-device server, and polls until it's healthy. After that, all tool calls are proxied over localhost HTTP with per-device bearer token auth.
Tools
| Tool | Description |
|---|---|
list_devices |
List available iOS and Android devices |
screenshot |
Capture the device screen (JPEG) |
uitree |
Get the UI element tree as a flat list, with optional search and limit |
tap |
Tap at screen coordinates |
double_tap |
Double-tap at screen coordinates |
long_press |
Long-press at screen coordinates (configurable duration) |
scroll |
Swipe from start to end coordinates |
type_text |
Type text into the focused element |
press_button |
Press a hardware/navigation button (home, back, enter, volumeUp/Down, dpadUp/Down/Left/Right/Center) |
launch_app |
Launch an app by bundle ID / package name |
terminate_app |
Force-stop an app |
list_apps |
List installed apps |
run_code |
Execute sandboxed JavaScript on-device (see run_code below) |
run_code
Agents can pass code that looks like UIAutomator or XCUITest, both being Javascript under the hood. The sandbox restricts (Android) potentially dangerous Java operations and only allows (iOS) some XCUITest-ish commands
- Android: Rhino engine with UIAutomator bindings —
uiDevice(click, swipe, find elements, press keys, read display info),By(selectors),Until(wait conditions),console.log() - iOS: JavaScriptCore with XCUITest bindings —
app(query elements, tap, type, swipe),springboard,device,openApp(bundleId),sleep(ms),console.log()
Both platforms automatically kill runaway scripts (infinite loops) and create a fresh sandbox per call.
Prerequisites
- Node.js 18+ (for running via
npx) - Android: Android SDK with
adbon PATH - iOS Simulator: Xcode with
xcrun,simctl - iOS Real Device: Xcode with
xcodebuild,devicectl, andiproxy(from libimobiledevice) - Building from source: Bun runtime, Gradle (Android), Xcode (iOS)
Installation
Claude Code
claude mcp add mobile-device-mcp -- npx -y @srmorete/mobile-device-mcp@latest
Or with custom ports:
claude mcp add mobile-device-mcp -e MDMS_PORT_ANDROID=20000 -e MDMS_PORT_IOS=21000 -- npx -y @srmorete/mobile-device-mcp@latest
Modifying .mcp.json (Cursor, Claude Desktop, etc)
{
"mcpServers": {
"mobile-device-mcp": {
"command": "npx",
"args": ["-y", "@srmorete/mobile-device-mcp@latest"],
"env": {
"MDMS_PORT_ANDROID": "18000", # optional
"MDMS_PORT_IOS": "19000" # optional
}
}
}
}
Building from Source
git clone <repo-url>
cd mobile-device-mcp
bun install
# Build drivers for both platforms and pack tarball
./scripts/build.sh
The build script compiles the on-device drivers (Android APKs via Gradle, iOS test bundle via xcodebuild), copies them to drivers/, and creates an npm tarball.
To run locally during development:
bun run start # Start the MCP server
bun test # Run the test suite
Configuration
| Environment Variable | Default | Description |
|---|---|---|
MDMS_PORT_ANDROID |
18000 | Base port for Android on-device servers |
MDMS_PORT_IOS |
19000 | Base port for iOS on-device servers |
Ports are assigned sequentially — first Android device gets 18000, second gets 18001, and so on. Same for iOS starting at 19000.
Acknowledgements
Mobile Device MCP server stands on the shoulders of giants such as mobile-mcp and Maestro. Used as inspiration but reframed the current approach to be multi-device and with seamless Native/WebView support (especially on Android).
License
MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.