How do I use Frida MCP Server?

1. Click on "Install Server". 2. Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state. 3. In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Frida MCP Server spawn notepad.exe and create an interactive session" 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.

Frida MCP Server

A MCP server that provides dynamic instrumentation capabilities through Frida. Built with TypeScript and optimized for Bun runtime.

Features

Dynamic Instrumentation: Hook functions, inspect memory, and modify behavior of running processes
Device Management: Support for local, USB, and remote Frida devices
Process Control: Spawn, attach, resume, and kill processes
Interactive Sessions: Create persistent sessions for complex instrumentation workflows
Script Execution: Execute JavaScript code with both one-shot and persistent modes
File Operations: Download files from instrumented processes and query module information

Prerequisites

Bun >= 1.0.0
Frida >= 17.0.0 (installed on target system)

Installation

No installation required! Use npx to run directly from GitHub (see Configuration below).

For local development:

# Clone the repository git clone https://github.com/nonsleepr/frida-mcp.ts cd frida-mcp.ts # Install dependencies bun install

Note: While this project is built with Bun, npx (from npm) is required to run from GitHub due to native dependency requirements. Use npx --yes github:nonsleepr/frida-mcp.ts to run the server.

Configuration

Environment Variables

FRIDA_DEFAULT_DEVICE: Default remote device connection string (e.g., "192.168.1.100:27042" or "192.168.1.100", port defaults to 27042)

Claude Desktop Configuration

Add to your Claude Desktop config file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{ "mcpServers": { "frida": { "command": "npx", "args": ["--yes", "github:nonsleepr/frida-mcp.ts"], "env": { "FRIDA_DEFAULT_DEVICE": "192.168.1.100:27042" } } } }

Roo Configuration

Add this to your project's .roo/mcp.json:

{ "mcpServers": { "frida": { "command": "npx", "args": ["--yes", "github:nonsleepr/frida-mcp.ts"], "env": { "FRIDA_DEFAULT_DEVICE": "10.254.1.69:27042" }, "disabled": false, "alwaysAllow": [] } } }

Note: Set FRIDA_DEFAULT_DEVICE to your Frida server's connection string (hostname:port or just hostname), or leave it empty to use local devices.

Available Tools

Process Management

Tool	Description	Parameters
`spawn_process`	Spawn a process with Frida attached in paused state. The process will be paused at startup. Use `resume_process()` after loading scripts to continue execution.	`program` (string) `args` (optional string[]) `device_id` (optional: "default" or connection string)
`resume_process`	Resume a spawned process.	`pid` (number) `device_id` (optional: "default" or connection string)
`kill_process`	Kill a process by PID.	`pid` (number) `device_id` (optional: "default" or connection string)

Interactive Sessions

Tool	Description	Parameters
`create_interactive_session`	Create an interactive session for dynamic instrumentation. Establishes a Frida session for injecting JavaScript, hooking functions, and monitoring the target process. The session persists until explicitly closed or the process terminates.	`process_id` (number) `device_id` (optional: "default" or connection string)
`execute_in_session`	Execute JavaScript code within an existing Frida session. All scripts are persistent and continue running. Use `wait` parameter to capture initial output before returning. Messages available via `frida://sessions/{session_id}/messages` resource.	`session_id` (string) `javascript_code` (string) `wait` (optional number, default: 0 - seconds to wait for initial output)
`load_script_file`	Load and execute a Frida JavaScript file into an existing session. All scripts are persistent and continue running. Use `wait` parameter to capture initial output before returning. Messages available via `frida://sessions/{session_id}/messages` resource.	`session_id` (string) `script_path` (string) `wait` (optional number, default: 0 - seconds to wait for initial output)

File Operations

Tool

Description

Parameters

download_file

Download a file from remote system using Frida instrumentation. Uses double backslashes for Windows paths. Attaches to specified PID or finds explorer.exe. Works best for files up to ~500MB with 60s timeout.

file_path (string)

output_path (string)

pid (optional number)

device_id (optional: "default" or connection string)

Available Resources

Resources provide real-time, read-only access to Frida state via URI.

Direct Resources

URI	Description
`frida://devices`	List all connected Frida devices
`frida://sessions`	List all active Frida sessions and their statuses
`frida://documentation`	Frida tips & non-obvious JavaScript API features

Resource Templates

URI Template	Description
`frida://devices/{device_id}`	Get detailed information about a specific device by ID
`frida://devices/{device_id}/processes`	List processes on a specific Frida device. Use "default", "local", "usb", or "remote" for automatic device selection, provide a specific device ID, or use a connection string (hostname:port or hostname).
`frida://devices/{device_id}/processes/by-name/{process_name}`	Find a process by name (case-insensitive partial match) on a specific device. Use "default" for configured remote device. Supports connection strings (hostname:port or hostname).
`frida://devices/{device_id}/processes/{pid}/module`	Get main module information for a process (path, base address, size). Use "default" for configured remote device. Supports connection strings (hostname:port or hostname).
`frida://sessions/{sessionId}/messages`	Retrieve messages from persistent scripts with default 100 message limit. Messages are consumed when retrieved.
`frida://sessions/{sessionId}/messages/{limit}`	Retrieve messages from persistent scripts with custom limit. Use `/last:N` (e.g., `/last:10` for last 10 messages) or `/all` for unlimited. Messages are consumed when retrieved.