Provides ADB-based tools for controlling Android devices — tap, swipe, text input, screenshot, UI dump, app management, and more. Enables AI agents to interact with Android devices programmatically.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@android-mcp take a screenshot of my device"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Android device automation MCP server for Hermes Agent. Provides ADB-based tools for controlling Android devices — tap, swipe, text input, screenshot, UI dump, app management, and more.
Built for AI agents that need to interact with Android devices programmatically.
Device Management — list, connect, disconnect devices; check connection state
Gestures — tap, swipe, long press, pinch, multi-touch
Input — text typing, key events, clipboard
Screenshots — capture device screen as PNG
UI Analysis — dump view hierarchy via uiautomator
App Control — launch, stop, list installed apps
Password Manager — store and retrieve app credentials locally
System Actions — unlock, lock, volume, power, airplane mode
Related MCP server: autoglm-mcp-server
# Clone the repository
git clone git@github.com:AsmanovLev/android-mcp.git
cd android-mcp
# Install dependencies
npm install
# Build
npm run buildAdd to your ~/.hermes/config.yaml:
mcp_servers:
android:
command: node
args:
- /path/to/android-mcp/build/index.jsThen restart Hermes Agent. The tools will be available with the mcp_android_* prefix.
{
"tool": "mcp_android_list_devices"
}{
"tool": "mcp_android_screenshot",
"arguments": { "save_to": "/tmp/screen.png" }
}{
"tool": "mcp_android_tap",
"arguments": { "x": 500, "y": 800 }
}{
"tool": "mcp_android_unlock",
"arguments": { "pin": "1234" }
}Tool | Description |
| List connected ADB devices |
| Capture device screen |
| Dump UI hierarchy via uiautomator |
| Tap at coordinates |
| Swipe from (x1,y1) to (x2,y2) |
| Type text via ADB |
| Send key event |
| Wake + swipe up + enter PIN |
| Launch app by package name |
| Force stop app |
| List installed packages |
| Store/retrieve app credentials |
# Watch mode
npm run dev
# Run tests
npm testDeepagent — AI agent by Hermes Agent
MIT
Download an APK from F-Droid by package name.
Inputs: package_name (required), output_dir.
Search for Android apps on F-Droid by query.
Inputs: query (required), limit.
Start an Android activity by intent (package/class, action, data URI, extras).
Inputs: action, extras, serial, package, +2 more.
Manage Android apps via ADB: install APK, start app by package, force-stop, uninstall, or list installed packages.
Inputs: action (required), file, serial, package, +1 more.
Press a hardware/system button on the Android device: back, home, recent apps, menu, volume up/down, power, camera, or call.
Inputs: button (required), serial, return_screen.
List connected Android devices and optionally select one as the active device for all mobile tools.
Inputs: select.
Execute gestures on the Android device: tap, swipe, long-press, drag, or multi-step gesture sequences.
Inputs: action (required), x, y, x2, +6 more.
Execute a multi-step gesture sequence (array of steps). Each step can be tap, swipe, fastswipe, tap-hold, drag, press, move, release, hold, swipe-hold-start, swipe-hold-end, or swipe-hold-startend.
Inputs: steps (required), serial, return_screen.
Execute a named gesture macro (e.g., 'unlock', 'home', 'back', 'recents', 'notification_panel', 'quick_settings').
Inputs: macro (required), serial, return_screen.
Control device power state: turn screen on/off, check if awake, or reboot.
Inputs: action (required), serial.
Restart the ADB server (kill + start). Useful when devices are not detected or ADB is stuck.
Inputs: no named parameters.
Capture a screenshot from the Android device. Returns the screenshot file path and UI hierarchy text.
Inputs: serial, return_screen.
Get device state: battery level, screen orientation, display info, foreground app, and network status.
Inputs: serial.
Type text into the currently focused input field on the Android device.
Inputs: text (required), serial, clear_first, return_screen.
Unlock the Android device screen. Wakes the screen, swipes up, and enters a PIN/password if provided.
Inputs: pin, serial.
Wait for a UI element to appear or disappear on the screen, with timeout.
Inputs: text, serial, timeout, className, +3 more.
Manage stored passwords for auto-login on Android apps. Store, retrieve, list, or delete credentials.
Inputs: action (required), app, password, username.
Download an APK from RuStore by package name.
Inputs: package_name (required), output_dir.
Search for Android apps on RuStore by query.
Inputs: query (required), limit.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 19 tools
Each tool targets a distinct function: gestures, buttons, app management, F-Droid/RuStore search/download, password management, device state, etc. Even gesture tools are differentiated by single, combo, and macro operations.
Naming is inconsistent: some tools have 'mobile_' prefix while others (fdroid_download, rustore_search) do not. Verb-noun combinations vary (e.g., 'mobile_power', 'mobile_activity_manager', 'mobile_device_list'), and some names are vague.
19 tools is on the higher side but justified given the breadth of Android device control. A few tools (e.g., three gesture tools) could potentially be consolidated, but overall count is reasonable.
Covers core device interaction, app lifecycle, and app store search/download. Missing advanced debugging (logcat), file operations, or notification management, but sufficient for common automation tasks.
Give your AI agent a real cloud Android phone: push APKs, drive it, get screenshots and logs.
Cloud Android phones for AI agents: create a phone, read the screen, tap, type, install APKs, park.
Let AI agents control Android and iOS devices on Mobilerun Cloud: screen, input, apps, workflows.
Disposable cloud Android emulators for coding agents: run an APK or PR build, tap, type, screenshot.