Provides tools for mobile security testing on Android devices, including app lifecycle management via ADB, memory manipulation, and Java hooking to monitor method calls and bypass security features like SSL pinning.
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-mcp Search for 'crypto' classes in com.example.app and disable SSL pinning"
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.
MCP server for Frida-based mobile security testing. Exposes Frida functionality as MCP tools for AI-assisted security research.
Python 3.11+
Frida server running on target device
ADB access for Android devices
Rooted device (for most operations)
Related MCP server: Frida MCP Server
cd frida-mcp
uv pip install -e .Build the Frida agent (required):
cd agent
npm install
npm run buildclaude mcp add frida-mcp -- frida-mcpTool | Description |
| List all available Frida devices (USB, remote, local) |
| List running processes on a device |
| List installed applications on a device |
| Attach to app by bundle ID, name, or PID. Supports |
| Disconnect from the current session |
| Check if Frida session is still alive and healthy |
| List all active Frida sessions (multi-device support) |
| Switch to a different active session by ID |
Tool | Description |
| Get PID of a running app by package name |
| Launch app via ADB and return its PID |
| Force stop an app by package name |
| Force stop, launch fresh, and attach Frida in one step |
Tool | Description |
| List all loaded modules (libraries) in the process |
| List exports (functions) from a specific module |
| Search process memory for hex pattern or string |
| Read memory at a specific address |
| Write bytes to memory address (for patching) |
| Get base address of a module by name (partial match) |
Tool | Description |
| List loaded Java classes, optionally filtered |
| List methods of a Java class |
| Hook a Java method to monitor calls |
| Search for classes matching a pattern |
| Disable SSL certificate pinning |
| Get the current foreground activity |
| Dump all methods, fields, and constructors of a class |
| Search Java heap for live instances of a class |
Tool | Description |
| Install a persistent hook script that collects messages |
| Retrieve collected messages from persistent hooks |
| Clear the hook message buffer |
| Unload all persistent hook scripts |
| List all installed persistent hooks |
| Hook a native function by module+offset |
Tool | Description |
| List files in a directory on the device |
| Read a text file from the device |
| Download a file from device to local machine |
Tool | Description |
| Execute custom Frida JavaScript code |
| Run JavaScript within Java.performNow context |
1. list_devices → Find your device
2. connect target=com.example.app spawn=true → Attach to app
3. android_search_classes pattern=crypto → Find crypto classes
4. android_hook_method class_name=... method_name=... → Hook methods
5. get_hook_messages → See captured callsSELinux is automatically set to permissive mode when connecting (required for Frida injection on many devices)
The spawn=true option uses ADB-based launch which is more reliable than Frida's native spawn
Multi-session support allows attaching to multiple apps/devices simultaneously
Get the current foreground activity
Inputs: no named parameters.
Hook a Java method to monitor calls. Logs arguments, return values, and optionally backtraces.
Inputs: class_name (required), method_name (required), dump_args, dump_return, +1 more.
List loaded Java classes, optionally filtered by pattern
Inputs: pattern.
List methods of a Java class
Inputs: class_name (required).
Search for Java classes matching a pattern
Inputs: pattern (required).
Disable SSL certificate pinning to allow traffic interception
Inputs: no named parameters.
Clear the hook message buffer without retrieving
Inputs: no named parameters.
Connect to an app by bundle ID, name, or PID. Must call this before using other tools. When spawn=True, uses reliable adb-based launch instead of Frida spawn.
Inputs: target (required), device_id, spawn, timeout_ms.
Disconnect from the current session
Inputs: no named parameters.
Dump all methods, fields, and constructors of a Java class
Inputs: class_name (required).
Download a file from device to local machine
Inputs: remote_path (required), local_path (required).
List files in a directory on the device
Inputs: path.
Read a text file from the device
Inputs: path (required).
Get collected messages from persistent hooks installed via install_hook
Inputs: clear.
Get base address of a module by name (partial match). Returns {module: {name, base, size, path}}
Inputs: name (required).
Get PID of a running app by package name. Useful before attaching with connect().
Inputs: package (required), device_id.
Search Java heap for live instances of a class
Inputs: class_name (required), max_results.
Hook a native function by module+offset. Messages collected via get_hook_messages().
Inputs: module (required), offset (required), name.
Install a persistent hook script that stays active and collects messages. Use get_hook_messages to retrieve collected output.
Inputs: js_code (required), name.
Check if Frida session is still alive and healthy. Returns connection status, PID, and device info.
Inputs: no named parameters.
Launch an app via adb and return its PID (smart wait - polls for process).
Inputs: package (required), activity, device_id, timeout_ms.
List installed applications on a device
Inputs: device_id.
List all available Frida devices (USB, remote, local)
Inputs: no named parameters.
List all installed persistent hooks with their names and indices
Inputs: no named parameters.
List running processes on a device
Inputs: device_id.
List all active Frida sessions. Supports multi-device workflows.
Inputs: no named parameters.
List exports (functions) from a specific module
Inputs: module_name (required).
List all loaded modules (libraries) in the process
Inputs: no named parameters.
Read memory at a specific address
Inputs: address (required), size (required).
Search process memory for a pattern
Inputs: pattern (required), is_string.
Write bytes to memory address (for patching)
Inputs: address (required), hex_bytes (required).
Run arbitrary Java code within Java.performNow context. Has access to Java.use(), Java.choose(), send(), etc. Return value is the result of the last expression.
Inputs: code (required).
Execute custom Frida JavaScript code. Java bridge is auto-imported, so Java.perform(), Java.use(), etc. work directly.
Inputs: js_code (required).
Force stop app, launch fresh, and attach Frida. The reliable alternative to connect(spawn=True) which often times out.
Inputs: package (required), device_id, wait_ms.
Force stop an app by package name
Inputs: package (required), device_id.
Switch to a different active session by ID. Use list_sessions to see available sessions.
Inputs: session_id (required).
Unload all persistent hook scripts
Inputs: no named parameters.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 37 tools
Most tools have clearly distinct purposes (connecting, hooking, memory, files). However, some overlap exists: `get_pid` and `launch_app` both provide PIDs; `spawn_and_attach` duplicates `connect(spawn=True)`; `run_java` and `run_script` are similar. Descriptions help differentiate but confusion is possible.
Naming conventions are mixed: some tools use prefix_noun_verb (e.g., `android_get_current_activity`, `memory_read`), others use verb_noun (e.g., `list_apps`, `disconnect`), and some lack a clear pattern (e.g., `is_connected`). While prefixes like `android_`, `file_`, `memory_` provide consistency within groups, the overall pattern is not uniform.
37 tools is on the high side but still manageable. The server covers a broad domain (Frida-based mobile analysis) and each tool serves a specific purpose. However, some tools could be consolidated (e.g., multiple hooking variants, PID retrieval tools slightly redundant).
The tool set covers most core operations: connection/session management, file operations, memory manipulation, Java and native hooking, class introspection, and process control. Minor gaps exist (e.g., no direct tool for Java field modification or method call), but `run_java` provides a workaround. Overall, the surface is quite complete for its domain.
MCP server for static security analysis of Android source code
Mozark's MCP server for AI-powered app testing: device access, test automation, and QA insights.
Security scanner for MCP servers. Detect vulnerabilities, prompt injection, and tool poisoning.
MCP server for building and testing AI agents with multi-model experimentation and insights.