Jadx MCP Server

🚀 JADX-AI-MCP

JADX-AI-MCP is an open-source tool that provides enhanced features for Android application decompilation and static analysis. By integrating Artificial Intelligence (AI) models with the Model Context Protocol (MCP), it makes reverse engineering more intelligent and efficient. This project aims to assist developers, security researchers, and academic personnel in conducting application analysis more easily.

🚀 Quick Start

1. Download and install JDK (version 8 or higher is recommended).
2. Configure environment variables to ensure that PATH includes the path to the jadx-ai-mcp tool.
3. Use the following command for basic decompilation:

jadx -d output_directory --mcp-enable <input APK file>

4. Enable AI-assisted features:

jadx -i --ai-enable <input APK file>

✨ Features

• AI-Enhanced Analysis: Integrates AI models to assist in code understanding, vulnerability detection, and code refactoring.
• Model Context Protocol (MCP): Supports communication with external model services via the MCP protocol for dynamic feature expansion.
• Multi-Platform Compatibility: Compatible with Windows, macOS, and Linux systems.

📦 Installation

1. Download the jadx-ai-mcp Toolkit

Download the latest version of the toolkit from GitHub. After extraction, you'll get jadx, mcp_server, and related documentation.

2. Installation and Configuration

Windows

# Add to environment variables
set PATH=%PATH%;C:\path\to\jadx-ai-mcp

macOS/Linux

# Add to the PATH environment variable
export PATH=$PATH:/path/to/jadx-ai-mcp

3. Start the jadx Service

Run the following command in the terminal to start the jadx analysis service:

jadx -d <directory> --mcp-enable

4. Configure MCP Plugins

Create a config.json file and add the required MCP plugin configurations:

{
    "plugins": [
        {
            "name": "security-check",
            "url": "http://localhost:8080"
        }
    ]
}

5. Use AI-Assisted Features

Run the following command to use the AI model for code analysis:

jadx -i --ai-enable <your APK file>

📚 Documentation

Main Functions

1. Enhanced Static Analysis

   • Code parsing and decompilation
   • Source code structure visualization
   • Type recovery and repair

2. Dynamic Plugin Support

   • Supports writing MCP plugins in multiple languages (e.g., Python, JavaScript)
   • Dynamically load/unload plugins with real-time effect

3. AI-Assisted Functions

   • Code understanding and comment generation
   • Vulnerability detection and repair suggestions
   • Automatic refactoring and optimization

4. Scalable Design

   • Supports multiple decompilation tools (e.g., jadx, smali)
   • Supports multi-platform APK analysis

Future Plans

• Support for More Languages: Add native support for languages such as Python and JavaScript.
• Hermes Decoding Integration: Enhance support for React Native applications.
• Expansion of the MCP Toolkit: Develop more useful MCP plugins.
• Local LLM Support: Achieve seamless integration with local AI models.

Contribution Guidelines

Contributions are welcome! Whether it's issue reporting, feature suggestions, or code submissions, we're eager to hear from you:

1. Submit Issues to report problems or propose suggestions.
2. Fork the repository, make code modifications, and submit a Pull Request.
3. Share your usage experiences or integration cases.

License

This project inherits the Apache 2.0 license, allowing users to freely use, modify, and distribute it. For detailed license information, please refer to the LICENSE file.

Important Note

Disclaimer

The JADX-AI-MCP tool is intended for educational, research, and legitimate security assessment purposes only. Users must ensure that their actions comply with relevant laws, regulations, and ethical standards. Any unauthorized use, such as unauthorized code analysis or infringement of intellectual property rights, is considered a violation, and the developers shall not be held responsible.

Please use this tool within the scope of legality and authorization and adhere to the corresponding code of conduct.

Acknowledgments

Thanks to @skylot for developing the original version of jadx. Thanks to the contributors in the open-source community and all those who support and participate in this project.

For more detailed information, please visit the GitHub repository or refer to the relevant documentation.