/ Annuaire / Playground / jadx-ai-mcp
● Communauté zinja-coder ⚡ Instantané

jadx-ai-mcp

par zinja-coder · zinja-coder/jadx-ai-mcp

JADX plugin for AI-powered Android APK analysis — decompile, search, rename, debug, and trace xrefs with LLM assistance.

jadx-ai-mcp is a JADX decompiler plugin that integrates with MCP to enable AI-assisted Android APK reverse engineering. It provides 25+ tools for code analysis (class/method browsing, smali access), resource inspection (AndroidManifest, strings), refactoring (rename classes, methods, fields, variables), debugging (stack frames, threads, variables), and cross-reference tracking. Works with Claude Desktop, Cherry Studio, and other MCP clients.

Pourquoi l'utiliser

Fonctionnalités clés

Démo en direct

Aperçu en pratique

jadx-ai-mcp.replay ▶ prêt
0/0

Installer

Choisissez votre client

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "jadx-ai-mcp": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/zinja-coder/jadx-ai-mcp"
      ]
    }
  }
}

Ouvrez Claude Desktop → Settings → Developer → Edit Config. Redémarrez après avoir enregistré.

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "jadx-ai-mcp": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/zinja-coder/jadx-ai-mcp"
      ]
    }
  }
}

Cursor utilise le même schéma mcpServers que Claude Desktop. La config projet l'emporte sur la globale.

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "jadx-ai-mcp": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/zinja-coder/jadx-ai-mcp"
      ]
    }
  }
}

Cliquez sur l'icône MCP Servers dans la barre latérale Cline, puis "Edit Configuration".

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "jadx-ai-mcp": {
      "command": "TODO",
      "args": [
        "See README: https://github.com/zinja-coder/jadx-ai-mcp"
      ]
    }
  }
}

Même format que Claude Desktop. Redémarrez Windsurf pour appliquer.

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "jadx-ai-mcp",
      "command": "TODO",
      "args": [
        "See README: https://github.com/zinja-coder/jadx-ai-mcp"
      ]
    }
  ]
}

Continue utilise un tableau d'objets serveur plutôt qu'une map.

~/.config/zed/settings.json
{
  "context_servers": {
    "jadx-ai-mcp": {
      "command": {
        "path": "TODO",
        "args": [
          "See README: https://github.com/zinja-coder/jadx-ai-mcp"
        ]
      }
    }
  }
}

Ajoutez dans context_servers. Zed recharge à chaud à la sauvegarde.

claude mcp add jadx-ai-mcp -- TODO 'See README: https://github.com/zinja-coder/jadx-ai-mcp'

Une seule ligne. Vérifiez avec claude mcp list. Supprimez avec claude mcp remove.

Cas d'usage

Usages concrets : jadx-ai-mcp

How to analyze an Android APK with AI assistance using JADX

👤 Mobile security researchers analyzing APKs they have authorization to test ⏱ ~45 min intermediate

Quand l'utiliser : You have an APK to analyze and want AI to help understand its behavior.

Prérequis
  • JADX with jadx-ai-mcp plugin — jadx plugins --install 'github:zinja-coder:jadx-ai-mcp'
  • JADX MCP Server running — Set up the Python MCP server with uv
Déroulement
  1. Check the manifest
    Get the AndroidManifest.xml. What permissions does this app request and what components (activities, services, receivers) does it declare?✓ Copié
    → Parsed manifest with permission analysis
  2. Find interesting classes
    Search for classes related to network communication or encryption. List them with their methods.✓ Copié
    → Filtered class list with method signatures
  3. Analyze and rename
    Get the source code of the main network class. Rename obfuscated methods to descriptive names based on their behavior.✓ Copié
    → Decompiled code with meaningful names applied

Résultat : An annotated view of the APK with key functionality understood and obfuscation partially reversed.

Pièges
  • Heavily obfuscated apps resist decompilation — Use get_smali_of_class for bytecode-level analysis when decompilation fails
Combiner avec : filesystem

Debug an Android app with AI-assisted breakpoint analysis

👤 Android developers and security researchers debugging app behavior ⏱ ~30 min advanced

Quand l'utiliser : You need to understand runtime behavior that isn't clear from static analysis.

Prérequis
  • JADX configured for debugging — Connect JADX debugger to a running Android app or emulator
Déroulement
  1. Set breakpoints and inspect
    I'm debugging the login flow. Show me the current stack frames and local variables at the authentication method.✓ Copié
    → Stack trace with variable values at the breakpoint

Résultat : Runtime understanding of app behavior at critical code points.

Pièges
  • App detects debugger and exits — Look for anti-debugging checks in the code and bypass them before attaching

Combinaisons

Associez-le à d'autres MCPs pour un effet X10

jadx-ai-mcp + filesystem

Export decompiled and annotated code for documentation

Export the decompiled source of the authentication module and save it to ~/analysis/auth-module.java.✓ Copié

Outils

Ce que ce MCP expose

OutilEntréesQuand appelerCoût
fetch_current_class none Get the currently selected class source code 0
search_classes_by_keyword keyword: str Find classes matching a keyword 0
get_android_manifest none Retrieve the AndroidManifest.xml 0
rename_class old_name: str, new_name: str Rename an obfuscated class 0
xrefs_to_method class_name: str, method_name: str Find all references to a method 0
get_smali_of_class class_name: str Get Dalvik bytecode (smali) for a class 0
debug_get_stack_frames none Get current debugger stack frames 0

Coût et limites

Coût d'exécution

Quota d'API
N/A — fully local
Tokens par appel
200–2000 tokens per tool call
Monétaire
Free — both JADX and the plugin are open source
Astuce
Use search_classes_by_keyword to narrow targets before fetching full class sources.

Sécurité

Permissions, secrets, portée

Stockage des identifiants : N/A
Sortie de données : All analysis is local. No network calls from the MCP server.

Dépannage

Erreurs courantes et correctifs

Plugin not loading in JADX

Ensure JADX version is compatible. Try reinstalling with: jadx plugins --install 'github:zinja-coder:jadx-ai-mcp'

Vérifier : jadx plugins --list
MCP server connection failed

Ensure the JADX MCP Server is running and the port matches your client config.

Vérifier : Check the server terminal for startup messages
Class not found in search

Obfuscated names may not match your search. Try broader patterns or search by method signatures.

Vérifier : get_all_classes to see available class names

Alternatives

jadx-ai-mcp vs autres

AlternativeQuand l'utiliserCompromis
jadx-mcp-serverYou only need the MCP server component without the JADX pluginServer-only — still needs the jadx-ai-mcp plugin in JADX
GhidrAssistMCPYou want to analyze native binaries rather than Android Dalvik bytecodeNative binary focus vs Android-specific tools

Plus

Ressources

📖 Lire le README officiel sur GitHub

🐙 Voir les issues ouvertes

🔍 Parcourir les 400+ serveurs MCP et Skills