mobile-next/mobile-mcp
Model Context Protocol Server for Mobile Automation and Scraping (iOS, Android, Emulators, Simulators and Real Devices)
TypeScript · ★ 7,634 · Fork 655 · 在 GitHub 查看
在手机上浏览源码、订阅每日 AI 趋势日报与项目架构深度解读:获取 CopoHub 客户端(iOS & HarmonyOS)→
GitHub Trending 上榜记录
mobile-next/mobile-mcp 已在 GitHub Trending 每日榜单上榜 2 天,最高排名第 14,首次上榜 2026-09-26,最近一次上榜 2026-09-27。
| 日期 | 排名 | Star | 当日增长 |
|---|---|---|---|
| 2026-09-27 | #14 | 7,634 | +573 |
| 2026-09-26 | #14 | 7,291 | +143 |
同期上榜的项目
- paperclipai/paperclip – The open-source app everyone uses to manage agents at work
- vectorize-io/hindsight – Hindsight: Agent Memory That Learns
- debpalash/VoiceStudio – VoiceStudio is the open-source, fully-local ElevenLabs alternative — voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages.
- rohitg00/ai-engineering-from-scratch – Learn AI Engineering! Learn it. Build it. Ship it for others.
- InfinityLoop1308/PipePipe – An open-source Android app to let you browse YouTube and other services freely.
- vercel-labs/scriptc – TypeScript-to-Native Compiler
mobile-next/mobile-mcp 项目深度剖析与商业评估报告
mobile-next/mobile-mcp 项目深度剖析与商业评估报告
1. 核心定位与商业落地价值
mobile-mcp 的根本价值在于将"移动端 UI 自动化"这一长期被 XCUITest / Espresso / Appium 三座大山割裂的领域,抽象为面向 LLM 的语义化工具层。传统移动自动化需要工程师掌握平台专有 DSL、元素定位策略(XCUIElementQuery、UiSelector)、驱动生命周期管理,而本项目通过 MCP(Model Context Protocol)协议把设备控制暴露为约 30 个平台无关的原子工具(mobileclickonscreenatcoordinates、mobilelistelementsonscreen、mobilelaunchapp 等),让 Agent 以自然语言驱动设备。
商业落地场景集中在四类:其一,跨平台 E2E 测试的 Agent 化重构——把脆弱的 XPath/UiAutomator 脚本替换为"目标驱动"的 Agent 循环,显著降低测试脚本的维护成本;其二,RPA 与数据录入——如 README 中"Pomodoro 应用注册 + 评分"这类多步跨 App 流程;其三,Agent 训练与数据采集——为 GUI Agent 提供真实设备交互环境;其四,CI/CD 中的真机回归,配合其 Cloud 产品实现设备池化。
护城河并非开源代码本身,而是三层结构:(1) 底层 mobilecli 作为"通用设备 CLI",把 adb/xcrun simctl/go-ios 等碎片化工具统一为 JSON-RPC,这是真正的技术资产;(2) Accessibility-first 策略——优先读取原生无障碍树而非截图,规避了视觉模型的 token 成本与延迟,这是与 Appium 系(依赖 WebDriver 协议)和纯 Vision 方案(如 Anthropic Computer Use)的关键差异;(3) Cloud 商业化闭环——开源 MCP 作为流量入口,通过 mobilelogintocloudprovider 等工具将用户导流至真机租赁,典型的 OSS-led SaaS 模式。628 forks / 7067 stars 的比率(约 11%)显示社区参与度健康,但需警惕其"开源引流"本质下核心能力(真机池、设备编排)的闭源倾向。
2. 核心架构设计与技术深度
从 README 描述与工具集推断,架构呈现清晰的三层解耦:
MCP Client (Claude/Cursor/Codex) │ stdio / Streamable HTTP ▼ ┌─────────────────────────────┐ │ MCP Server (TypeScript) │ ← 工具注册、参数校验、遥测 │ Tool Registry + Transport │ └──────────────┬──────────────┘ │ Robot 抽象层 ┌────────┼────────┐ ▼ ▼ ▼ iOS Robot Android Robot Cloud Robot │ │ │ mobilecli mobilecli HTTP │ │ simctl/ adb/ go-ios emulator
在 README 中展示上榜徽章
[](https://github.fq6825.top/repo/mobile-next/mobile-mcp)