Android MCP

Android MCP

Enables full Android control from any AI agent via 7 MCP tools, including screen capture, touch interaction, app management, and system control.

Category
Visit Server

README

Android MCP

Full Android control from any AI agent — Claude, OpenCode, Windsurf, Cursor… 7 categorical MCP tools · 90fps live viewer · WiFi ADB · zero app to install


Quick demo

android_screen(action="screenshot")                          # PNG capture
android_screen(action="ocr")                                 # visible text
android_interact(action="tap", params={"x": 540, "y": 960}) # tap
android_interact(action="find", params={"text": "Send"})     # find + tap
android_system(action="battery")                             # battery info
android_screen(action="viewer")                              # 90fps interactive window on PC

Architecture

AI Agent (Claude / OpenCode / Windsurf / Cursor…)
        ↓  MCP Protocol (stdio)
    server.py          ← 7 categorical tools
        ↓
    device_manager.py  ← device selection, multi-device
        ↓
    backends/
        ├── adb_backend.py       ← PRIMARY  : uiautomator2 + direct ADB
        └── companion_backend.py ← FALLBACK : Flutter app WebSocket
        ↓
    N Android phones / emulators

ADB backend (primary) — works on any device with developer mode enabled. Nothing to install on the phone. Uses uiautomator2 + ADB commands.

Companion backend (fallback) — optional Flutter app when ADB is unavailable on the network.


Requirements

  • Python 3.10+
  • ADB in PATH (winget install Google.PlatformTools)
  • Android: Developer mode + USB debugging (or WiFi debugging)

Installation

git clone https://github.com/Steph-ux/android-mcp
cd android-mcp
pip install -r requirements.txt

# Initialize uiautomator2 (once per device)
python -m uiautomator2 init

Live viewer 90fps (optional)

winget install Genymobile.scrcpy
python viewer.py

MCP Configuration

# Generate the correct JSON for your AI client
python mcp_config.py --client claude     # Claude Desktop
python mcp_config.py --client opencode  # OpenCode
python mcp_config.py --client windsurf  # Windsurf
python mcp_config.py --client cursor    # Cursor
python mcp_config.py --write            # write directly to config files

Claude Desktop example (claude_desktop_config.json):

{
  "mcpServers": {
    "android-mcp": {
      "command": "C:/Python312/python.exe",
      "args": ["D:/path/to/android-mcp/server.py"],
      "type": "stdio"
    }
  }
}

WiFi connection without USB (Android 11+)

# On the phone: Settings → Developer options → Wireless debugging → Pair device
adb pair 192.168.1.42:38765    # enter the code shown on the phone
adb connect 192.168.1.42:5555

Full guide → WIFI_PAIRING.md


The 7 MCP tools

Call convention: android_xxx(action="...", params={...}, device_id="serial") device_id is always optional (uses the currently selected device).


android_device — Device management

Action Params Description
list All connected devices (USB, WiFi, emulators)
select serial Set default device
connect host, port WiFi ADB connection
disconnect Disconnect current WiFi device
info Model, OS, resolution, density
status Is device connected and ready?
setup Configure animations, stay-awake, ATX agent

android_screen — Capture & Stream

Action Params Description
screenshot PNG capture (bypasses FLAG_SECURE) → image
region x y width height Capture a specific area → image
size {width, height}
is_on Is the screen on?
wake Wake the screen
start_stream Start ADB stream (~16fps)
stop_stream Stop stream
live_frame Latest stream frame → image
ocr lang Extract visible text (Tesseract)
find_image template_b64 threshold Template matching (OpenCV)
viewer fps bitrate no_control Launch scrcpy 90fps window on PC

android_interact — Touch, Keyboard, UI

Action Params Description
tap x y Tap
double_tap x y Double tap
long_press x y duration_ms Long press
swipe x1 y1 x2 y2 duration_ms Swipe
drag x1 y1 x2 y2 duration_ms Drag & drop
pinch x y scale duration_ms Pinch zoom
multi_touch points Multi-finger gestures
type text Type text
clear Clear active field
submit text Type + Enter
key key System key (HOME, BACK, ENTER, VOLUME_UP…)
combo keys Key combination (keycodes)
hierarchy Full UI XML tree (uiautomator2)
find text partial_match Find element and tap it
wait text timeout partial_match Wait for element
scroll text direction max_swipes Scroll to element
assert text partial_match Assert text is visible

android_app — Applications

Action Params Description
launch package Launch an app
close package Force-stop an app
list include_system List installed apps
install apk_path Install APK from PC
uninstall package Uninstall app
current Foreground app package
url url Open a URL
intent action uri package extras Send Android intent
settings section Open system settings (main wifi bluetooth display…)

android_files — Files

Action Params Description
push local_path remote_path PC → phone
push_b64 remote_path data Base64 → phone
pull remote_path local_path Phone → PC
pull_b64 remote_path Phone → base64
list directory List directory contents

android_system — System & Network

Action Params Description
shell command ADB shell command
logs lines package Logcat
battery Battery level and charging state
clipboard_get Read clipboard
clipboard_set text Write to clipboard
volume level stream Set volume
rotation rotation Screen rotation (0-3)
wifi enabled WiFi on/off
bluetooth enabled Bluetooth on/off
mobile_data enabled Mobile data on/off
notifications Active notifications
gps lat lng Mock GPS location
sensors sensor Accelerometer, gyro, light…
wifi_list Available WiFi networks
wifi_connect ssid password Connect to WiFi
contacts search limit Read contacts
sms to message Send SMS (requires confirmation on phone)

android_automation — Batch & Macros

Action Params Description
batch actions stop_on_error Run N actions in one call
macro_start name Start recording a macro
macro_record action …params Add action to macro
macro_stop Save macro
macro_list List saved macros
macro_replay name delay_ms Replay a macro
macro_delete name Delete a macro

Live viewer — scrcpy 90fps

python viewer.py                  # auto-detect device, 90fps, interactive
python viewer.py --fps 60         # 60fps
python viewer.py --record         # record to .mp4
python viewer.py --record out.mp4 # named output file
python viewer.py --multi          # one viewer per connected device
python viewer.py --no-control     # read-only mode
Control Action
Left click Tap
Drag Swipe
Right click BACK
Scroll Scroll
Alt+H HOME
Alt+S Screenshot → PC clipboard
Alt+F Fullscreen
Alt+R Rotation

Tests

# Unit tests (no device required)
pytest tests/test_server_unit.py -v      # 94 tests

# Integration tests (ADB device required)
pytest tests/test_integration.py -v

Project structure

android-mcp/
├── server.py              # MCP server — 7 tools
├── viewer.py              # scrcpy 90fps interactive viewer
├── mcp_config.py          # JSON config generator
├── device_manager.py      # Multi-device management
├── relay.py               # ⚠️ LEGACY — companion fallback only
├── requirements.txt
├── pyproject.toml
├── WIFI_PAIRING.md
├── backends/
│   ├── adb_backend.py     # Primary backend (uiautomator2 + ADB)
│   └── companion_backend.py  # Fallback backend (Flutter app)
├── examples/
│   ├── agent_loop.py      # Autonomous automation loop
│   └── whatsapp_auto.py   # WhatsApp automation example
└── tests/
    ├── conftest.py
    ├── test_server_unit.py   # 94 unit tests
    └── test_integration.py   # Real device tests

Examples

# WhatsApp — send a message
await android_app("launch", {"package": "com.whatsapp"})
await android_interact("find", {"text": "Alice"})
await android_interact("type", {"text": "Hello!"})
await android_interact("key",  {"key": "ENTER"})

# Batch in a single call
await android_automation("batch", {"actions": [
    {"action": "key",  "key": "HOME"},
    {"action": "tap",  "x": 540, "y": 200},
    {"action": "type", "text": "search query"},
    {"action": "key",  "key": "ENTER"},
]})

# Macro: record + replay
await android_automation("macro_start",  {"name": "login"})
await android_automation("macro_record", {"action": "tap",  "x": 540, "y": 400})
await android_automation("macro_record", {"action": "type", "text": "password"})
await android_automation("macro_stop")
await android_automation("macro_replay", {"name": "login"})

See examples/whatsapp_auto.py and examples/agent_loop.py for full use cases.


License

MIT

Recommended Servers

playwright-mcp

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.

Official
Featured
TypeScript
Magic Component Platform (MCP)

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.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

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.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

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.

Official
Featured
TypeScript
Kagi MCP Server

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.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured
Exa Search

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.

Official
Featured