phone-mcp
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., "@phone-mcpTake a photo with my phone and show me the result"
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.
Phone Agent Runtime
把一台 Android 手机(首台:荣耀 Magic7)变成已有 Agent 的长期行动端:手机上一个 APK 负责可靠执行、本地策略与随时暂停;电脑上一个进程负责配对、连接和把手机能力以 MCP 工具 暴露给任何 Agent(Claude Code、Codex 等)。
设计合同见 docs/design/,连接拓扑(局域网 / VPS 中转 / VPS 全部署)见 docs/design/CONNECTION.md,评估与决策见 docs/adr/,机器可校验的协议见 protocol/。手机端体验原型见 docs/prototype/。许可证:MIT(见 LICENSE);第三方声明见 NOTICE。
仓库结构
apps/
android/ Kotlin::runtime(纯 JVM 可靠性内核 + FakeBackend + 验收测试)、:app(APK)
control/server/ TypeScript:Gateway(WSS) + Device Tools + MCP server + CLI + node:sqlite 任务库
protocol/ 协议 v1:JSON Schema、示例、契约测试(两端实现以此为准)
docs/
design/ 来自 ChatGPT 讨论的设计包(ARCHITECTURE / PROTOCOL / CODEX_HANDOFF / 图)
adr/ 架构决策记录
acceptance-results.md A01–A20 验收结果与证据(真机项逐条注明)Related MCP server: android-phone-mcp-server
快速开始
# 协议契约测试
pnpm install
pnpm --filter @phone-runtime/protocol validate
# 手机端可靠性内核测试(无需手机)
cd apps/android && export JAVA_HOME=/opt/homebrew/opt/openjdk@17 && ./gradlew :runtime:test
# 控制端
pnpm --filter @phone-runtime/control build && pnpm --filter @phone-runtime/control test各子目录的 README 有完整的运行、配对与接入 Claude Code / Codex 的步骤。
不可改变的边界(摘自设计包)
手机执行,Agent 规划;控制端只做配对、认证、路由与协议翻译,不含第二套绕过手机内核的执行逻辑。
手机本地暂停与撤权不依赖网络;前台 UI 同时只有一个控制会话。
动作日志是手机的权威事实:同
action_id去重、不同 payload 冲突、派发前落盘 EXECUTING、结果与 Outbox 同事务、崩溃后 UNCERTAIN 而非重做。网络超时不等于未执行。手机不开任何控制端口;只主动出站 WSS。release 无调试入口。
页面文字、聊天消息、无障碍读到的内容都不能成为授权。
状态
见 docs/acceptance-results.md。真机相关项在 Magic7 接入前一律标记为未验证。
This server cannot be deployed
Maintenance
Related MCP Connectors
Melaya is a remote MCP server. It gives an assistant hands on your own Android phone and browser: it reads the screen through the accessibility tree, then taps, types and navigates inside the apps and sites you allow-list, with no per-app API. It also builds, schedules and runs agent pipelines across 6k+ connected tools. OAuth 2.1, nothing to install.
Remote MCP for Android CLI agent build gate, structured receipts, audit logs, and reviewer-ready evi
Drive real devices from your AI Coding tool. Embed a client SDK (Unity, Godot, Flutter, iOS/macOS, Android, React Native, Web) in your app, then capture screenshots, traverse the UI tree, inject taps and key events, and run automated test tasks on the physical device over a secure relay.
Control real Android and iOS devices with LLM agents — tap, swipe, type, automate flows.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to control Android phones via MCP and HTTP. Supports screen capture, taps, swipes, text input, and app management.AGPL 3.0
- AlicenseAqualityAmaintenanceMCP server for controlling Android devices via semantic actions with verification, enabling any MCP client to interact with phones through natural language and structured tools.182MIT
- AlicenseNot gradedqualityAmaintenanceLocal-first MCP server that enables AI agents to safely control and observe Android mobile devices, supporting device diagnostics, app lifecycle management, async task execution, and performance monitoring through a policy-constrained interface with explicit confirmations.Apache 2.0
- AlicenseNot gradedqualityAmaintenanceControls Android and iOS devices from AI agents through a token-efficient MCP server, enabling observation, tapping, typing, gestures, navigation, assertions, and safety-gated actions.MIT