Android ADB Controller MCP Server
Overview
This MCP server enables AI agents to control Android devices via ADB commands. It provides tools for:
- Listing connected devices
- Executing screen taps
- (Extensible for more ADB commands)
Setup Guide
Prerequisites
- Android device with USB debugging enabled
- ADB installed and working (
adb devicesshould show your device) - Node.js v16+
Installation Steps
- Create MCP Server:
npx @modelcontextprotocol/create-server android-adb-controller
cd android-adb-controller
- Install Dependencies:
npm install adbkit @types/adbkit
- Configure MCP Settings:
Add to
mcp_settings.json:
{
"mcpServers": {
"android-adb": {
"command": "node",
"args": ["android-adb-controller/build/index.js"]
}
}
}
Implementation Details
Key Components
- ADB Client Initialization:
this.adbClient = adb.createClient();
- Command Execution:
private async executeADBCommand(deviceId: string, command: string) {
const device = this.adbClient.getDevice(deviceId);
return await device.shell(command);
}
- Tool Definitions:
list_devices: Lists connected Android devicestap_screen: Executes screen taps at specified coordinates
Troubleshooting Guide
Common Issues & Solutions
- ADB Command Not Found:
- Ensure ADB is installed and in system PATH
- Verify platform-tools are extracted correctly
- Device Not Detected:
- Check USB debugging is enabled
- Verify USB cable connection
- Run
adb devicesto confirm detection
- TypeScript Errors:
Solution: Added proper type definitions in
adbkit.d.ts:
declare module 'adbkit' {
interface Device {
id: string;
type: string;
}
// ... other type definitions
}
- MCP Server Connection Issues:
- Verify server is running (
node build/index.js) - Check MCP settings configuration
Extending Functionality
To add new commands:
- Define new tool in
ListToolsRequestSchema - Implement handler in
CallToolRequestSchema - Add corresponding ADB command execution
Example for text input:
{
name: "input_text",
description: "Input text on device",
inputSchema: {
type: "object",
properties: {
deviceId: { type: "string" },
text: { type: "string" }
}
}
}
Development Workflow
- Make code changes
- Rebuild:
npm run build
- Restart server:
node build/index.js
Architecture
MCP Server (Node.js) ↔ ADB Interface ↔ Android Device
↑
|
AI Control System
Future Improvements
- Add swipe gestures
- Implement screenshot capability
- Add device monitoring
- Support multiple simultaneous devices
Recommend MCP Servers 💡
playwright-mcp
A Model Context Protocol (MCP) server that leverages Playwright for browser automation, packaged as a Docker image.
mcp-server-docker
MCP server for Docker
@hiveflow/mcp-server
Official Model Context Protocol (MCP) server for HiveFlow, enabling AI assistants like Claude and Cursor to directly connect and interact with the HiveFlow automation platform for managing and executing workflows.
auto-mobile
Mobile automation suite of tools including an MCP and libraries for test authoring & execution
mcp-server-apache-airflow
A Model Context Protocol (MCP) server implementation for Apache Airflow, enabling seamless integration with MCP clients. This project provides a standardized way to interact with Apache Airflow through the Model Context Protocol.
mcp-server-macos-use
An MCP server in Swift that enables AI agents to control macOS applications using accessibility APIs, compatible with any model.