Provides tools for dynamic analysis of Android applications, enabling AI agents to list installed apps, manage application lifecycles, monitor system logs via logcat, and inject Frida scripts for real-time instrumentation.
Integrates with Magisk-rooted devices and the zygisk-gadget module to facilitate deep process instrumentation and automated reverse engineering on Android.
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., "@Frida Agent MCP trace the Cipher.doFinal method in com.taobao.taobao to see the encryption parameters"
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.
AI 驱动的 Android 动态分析工具。通过 MCP (Model Context Protocol) 让 AI 直接操控 Frida，实现自动化逆向分析。
AI (Claude) → MCP Server → Frida CLI → zygisk-gadget (手机) → 结果返回Python >= 3.10
Android 手机已 root，安装 Magisk
安装 zygisk-gadget Magisk 模块（固定端口 14725）
ADB 已连接手机
Related MCP server: ya-frida-mcp
pip install .安装后运行一键配置（注册 MCP + 安装 Skills）：
frida-mcp-setup或手动配置，在 ~/.claude.json 中添加：
{
"mcpServers": {
"frida-agent": {
"command": "frida-mcp"
}
}
}工具 | 说明 |
| 建立 ADB 端口转发，连接 zygisk-gadget |
| 列出设备上的第三方应用 |
| 注入 Frida JS 脚本到前台应用（支持 |
| 杀死 → 重启 → 注入，一步完成（适用于需要启动时 hook 的场景） |
| 获取注入脚本的输出（支持分页、自动文件回退） |
| 查看 Android 日志，排查崩溃原因 |
| 启动指定应用 |
| 强制停止指定应用 |
| 重新建立连接（崩溃后恢复） |
| 断开当前注入会话 |
安装后可在 Claude Code 中使用：
/reverse-analyze — 逆向分析与动态 Hook，结合 IDA 静态分析 + Frida 动态验证
/trace-vm — VM 保护函数分析，指导人工 trace 并由 AI 分析结果
用户: hook 淘宝的 Cipher.doFinal 看看加密参数
AI 自动执行:
1. connect() — 建立连接
2. spawn_and_inject( — 启动并注入
"com.taobao.taobao",
"Java.perform(function() { ... })"
)
3. 提示用户触发操作 — "请点击登录按钮"
4. get_messages() — 获取 hook 结果
5. 分析并展示加密参数和返回值用户: /reverse-analyze 分析 libsgmain.so 中 0x1234 处的函数
AI 自动执行:
1. IDA MCP 反编译目标函数，分析参数和返回值类型
2. 生成 Frida Hook 脚本
3. spawn_and_inject 注入
4. 交叉验证静态分析和动态结果本工具使用 Frida CLI 而非 frida-python API，通过 frida -H 127.0.0.1:14725 -F -l script.js -o output.txt 注入脚本：
脚本中的 send() 调用自动转换为 console.log(JSON.stringify())
所有脚本自动包裹 try-catch 防止注入崩溃
输出通过 -o 标志写入文件，get_messages 时停止 frida 进程以刷新文件
src/frida_mcp/
├── server.py # MCP Server 入口，10 个工具注册
├── connection.py # Frida CLI 连接管理，进程生命周期
├── executor.py # 脚本预处理（send 转换、try-catch 包裹）
├── messages.py # 消息分页、大数据自动文件回退
├── logcat.py # Android logcat 封装
├── setup.py # 一键安装配置
└── skills/ # Claude Code 技能模板MIT
Connect to the Frida gadget on the device. Performs adb port forwarding (tcp:14725) and connects to the zygisk-gadget. Call this before any other operation.
Inputs: no named parameters. Returns: result
Detach from the current session and unload any injected scripts.
Inputs: no named parameters. Returns: result
Inject a Frida JavaScript script into the frontmost application. The script is automatically wrapped with try-catch for safety. Returns immediately after injection - use get_messages() to retrieve results after the user has triggered the target functionality.
Args: script: Frida JavaScript code to inject. Use send() to return data. script_file: Path to a .js file to inject. Use this for large scripts instead of script parameter.
Inputs: script, script_file. Returns: result
Get messages collected from the injected script's send() calls. Call this after the user has triggered the target functionality.
Args: limit: Maximum number of messages to return (default 50). offset: Skip first N messages for pagination. save_to_file: If True, save all messages to a JSON file and return the path.
Inputs: limit, offset, save_to_file.
Force stop an Android app by package name.
Args: package: The app package name (e.g. com.example.app).
Inputs: package (required). Returns: result
Launch an Android app by package name.
Args: package: The app package name (e.g. com.example.app).
Inputs: package (required). Returns: result
List running applications on the connected device. Returns application identifier, name, and PID.
Inputs: no named parameters. Returns: result
Get Android logcat output. Useful for debugging crashes after script injection.
Args: filter: Optional filter string (case-insensitive match on each line). lines: Number of recent lines to fetch (default 100). clear: If True, clear logcat buffer before fetching.
Inputs: filter, lines, clear. Returns: result
Reconnect to the gadget after an app crash. Detaches current session, re-establishes adb forwarding, and reconnects. After reconnect, restart the target app and call execute() again.
Inputs: no named parameters. Returns: result
Kill app, relaunch, connect, and inject script in one step. Use this when you need to hook early initialization (e.g. onCreate, static initializers). Equivalent to: kill_app -> launch_app -> connect -> execute.
Args: package: The app package name (e.g. com.taobao.taobao). script: Frida JavaScript code to inject at startup. Use send() to return data. script_file: Path to a .js file to inject. Use this for large scripts.
Inputs: package (required), script, script_file. Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 10 tools
Most tools have distinct purposes, but there is some potential overlap between 'execute' and 'spawn_and_inject', as both inject scripts, with the latter being a convenience wrapper. The descriptions clarify the differences, so confusion is minimal.
Tool names follow a consistent verb-based pattern (e.g., connect, detach, execute, get_messages), with minor deviations like 'logcat' (a noun) and 'spawn_and_inject' (compound verb). Overall, naming is readable and mostly uniform.
With 10 tools, the count is well-suited for the server's purpose of Android app instrumentation with Frida. It covers essential operations like connection management, script injection, app control, and debugging without being overwhelming.
The tool set provides comprehensive coverage for the domain, including session management (connect, detach, reconnect), script injection (execute, spawn_and_inject), result retrieval (get_messages), app control (kill_app, launch_app, list_apps), and debugging (logcat). No obvious gaps are present.
MCP server for static security analysis of Android source code
Give your AI agent a real cloud Android phone: push APKs, drive it, get screenshots and logs.
Let AI agents control Android and iOS devices on Mobilerun Cloud: screen, input, apps, workflows.
Cloud Android phones for AI agents: create a phone, read the screen, tap, type, install APKs, park.