adb-mcp and agent-droid-bridge

These two tools are complements because they both implement the MCP server protocol for controlling Android devices via ADB, with `srmorete/adb-mcp` providing a general TypeScript implementation and `Neverlow512/agent-droid-bridge` leveraging it to offer AI agents programmatic control.

adb-mcp
50
Established
agent-droid-bridge
42
Emerging
Maintenance 2/25
Adoption 7/25
Maturity 24/25
Community 17/25
Maintenance 13/25
Adoption 11/25
Maturity 18/25
Community 0/25
Stars: 37
Forks: 10
Downloads:
Commits (30d): 0
Language: JavaScript
License: MIT
Stars: 14
Forks:
Downloads: 652
Commits (30d): 0
Language: Python
License: MIT
Stale 6m
No risk flags

About adb-mcp

srmorete/adb-mcp

An MCP (Model Context Protocol) server for interacting with Android devices through ADB in TypeScript.

Exposes comprehensive Android device control through stdio-based MCP tools including app installation, logcat filtering, file transfer, shell execution, and UI inspection via screenshot/XML hierarchy dumps. Designed as a Claude Desktop integration that bridges AI models with local ADB commands, supporting Android 8.0+ devices and emulators with configurable ADB paths. Provides both high-level operations (APK deployment, permission management) and low-level access (custom shell commands, activity/package manager control).

About agent-droid-bridge

Neverlow512/agent-droid-bridge

Agent Droid Bridge gives AI agents programmatic control over Android devices and emulators via ADB, exposed as an MCP server.

Exposes 13 structured MCP tools for Android automation—UI inspection, screen capture, touch/text input, and arbitrary ADB commands—optimized for agent workflows with minimal context overhead. Built on stdio transport with shell injection protection via `shlex` parsing, it targets any MCP-compatible AI client and supports multi-device selection with auto-detection for single connections. Purpose-built tools return parsed, compact responses instead of raw XML hierarchies, enabling long-running automation tasks without bloating agent context.

Scores updated daily from GitHub, PyPI, and npm data. How scores work