JADX-AI-MCP Documentation¶

Welcome to the comprehensive documentation for JADX-AI-MCP, an AI-powered Android reverse engineering toolkit that bridges JADX decompiler with Large Language Models through the Model Context Protocol.

What is JADX-AI-MCP?¶

JADX-AI-MCP is a sophisticated reverse engineering solution consisting of two tightly integrated components:

1. JADX-AI-MCP Plugin (Java) - A JADX-GUI plugin that exposes the decompiler's functionality through a local HTTP API
2. JADX-MCP-Server (Python) - An MCP server that translates HTTP endpoints into MCP tools consumable by AI assistants

Together, these components enable real-time, AI-assisted Android application analysis, vulnerability detection, and code understanding.

Key Features¶

Advanced Analysis Capabilities¶

• Real-time Code Review: AI analyzes decompiled code as you navigate
• Cross-Reference Analysis: Track method/class/field usage across entire codebases
• Resource Extraction: Access manifests, strings, layouts, and resources
• Smali Analysis: Low-level bytecode inspection for deep analysis

Security & Reverse Engineering¶

• Vulnerability Detection: Automated SAST (Static Application Security Testing)
• Malware Analysis: Behavioral pattern detection and C2 identification
• Obfuscation Analysis: Systematic deobfuscation workflows
• Permission Auditing: Identify overprivileged and unused permissions

Development Assistance¶

• Intelligent Refactoring: AI-assisted renaming of obfuscated classes, methods, fields, variables, and packages
• Debug Integration: Runtime variable inspection and stack trace analysis
• Pagination Support: Efficient handling of large APKs (10,000+ classes)
• Multi-Client Support: Works with Claude, Cherry Studio, LM Studio, Codex
• Remote Access: Docker, WSL, and remote VM support via --host binding

Architecture Overview¶

graph TB
    subgraph "User Layer"
        A[User] -->|Natural Language| B[LLM Client]
    end

    subgraph "MCP Layer"
        B -->|MCP Protocol| C[MCP Server Python]
        C -->|Tool Invocation| D[FastMCP Framework]
    end

    subgraph "Integration Layer"
        D -->|HTTP Request| E[JADX Plugin Server]
        E -->|Javalin Routes| F[Route Handlers]
    end

    subgraph "Analysis Layer"
        F -->|JADX API| G[JADX Wrapper]
        G -->|Query| H[Decompiled Code]
        G -->|Query| I[Resources]
        G -->|Query| J[Debug Info]
    end

    H -->|Response| F
    I -->|Response| F
    J -->|Response| F

Network Architecture¶

There are two separate connections in the system:

┌─────────────┐    --host / --port     ┌──────────────────┐   --jadx-host / --jadx-port   ┌──────────────────┐
│  LLM Client │ ◄──────────────────►   │  jadx-mcp-server │ ──────────────────────────►   │  JADX-GUI Plugin │
│  (Claude,   │   Where the MCP server │                  │   Where the MCP server looks  │  (jadx-ai-mcp)   │
│   Codex..)  │   LISTENS for clients  │                  │   for the JADX plugin         │                  │
└─────────────┘                        └──────────────────┘                               └──────────────────┘

Quick Start¶

Prerequisites¶

Installation (5 minutes)¶

# Step 1: Install JADX Plugin
jadx plugins --install "github:zinja-coder:jadx-ai-mcp"

# Step 2: Download MCP Server
wget https://github.com/zinja-coder/jadx-ai-mcp/releases/latest/download/jadx-mcp-server.zip
unzip jadx-mcp-server.zip

# Step 3: Install UV
curl -LsSf https://astral.sh/uv/install.sh | sh

# Step 4: Run MCP Server
cd jadx-mcp-server
uv run jadx_mcp_server.py

# Step 5: Configure Claude Desktop
# Edit: ~/.config/Claude/claude_desktop_config.json

See Installation Guide for detailed instructions.

📚 Documentation Structure¶

For Users¶

• Installation Guide - Complete setup for all platforms
• User Guide - Workflows, prompts, and best practices
• Examples - Real-world analysis scenarios

For Developers¶

• API Reference - Complete tool documentation with examples
• Architecture - System design and component interaction
• Python Module Reference - Server-side code documentation
• Java Plugin Reference - Plugin internals and extension points

Support¶

• Troubleshooting - Common issues and solutions
• Contributing - Development guidelines

🔧 Tool Categories¶

Class Analysis Tools (10 tools)¶

Tools for analyzing decompiled Java classes, methods, and fields.

→ View Class Tools

Search Tools (3 tools)¶

Full-text search across classes, methods, and code with advanced scope filtering.

→ View Search Tools

Resource Tools (5 tools)¶

Access to AndroidManifest components, strings, layouts, and resources.

→ View Resource Tools

Cross-Reference Tools (3 tools)¶

Track usage of classes, methods, and fields across codebase.

→ View Xref Tools

Refactoring Tools (5 tools)¶

Rename classes, methods, fields, variables, and packages.

→ View Refactor Tools

Debug Tools (3 tools)¶

Runtime analysis during JADX debugging sessions.

→ View Debug Tools

🔐 Security & Privacy¶

• Secure Defaults: Server binds to 127.0.0.1 only — remote access must be explicitly enabled
• No Authentication: When binding to non-localhost with --host 0.0.0.0, traffic is unencrypted and unauthenticated — use only on trusted networks
• Proxy Isolation: Internal HTTP requests use trust_env=False to prevent proxy interception
• No Data Collection: No telemetry or usage tracking
• Stdio Safety: Banner and health check output goes to stderr to prevent JSON-RPC stream pollution
• Open Source: Fully auditable codebase

Supported Platforms¶

🤝 Community & Support¶

• GitHub Issues: Report bugs
• GitHub Discussions: Ask questions

📄 License¶

Apache License 2.0 - See LICENSE

Acknowledgments¶

• JADX Project: @skylot - Amazing decompiler
• Anthropic: Model Context Protocol development
• FastMCP: @jlowin - Python MCP framework

Next Steps¶

1. New Users: Start with Installation
2. Quick Demo: See Examples
3. API Exploration: Browse API Reference
4. Contributing: Read Contributing Guide

Current Version: 6.3.0 | Last Updated: March 2026