Decepticon - Vibe Hacking Agent

One-Minute Overview#

Decepticon is an AI-powered red teaming tool that uses multi-agent systems to automatically execute penetration testing tasks. Designed for security researchers and cybersecurity teams, it automates repetitive penetration work, allowing security experts to focus on strategic decision-making and threat defense. The core value is helping security teams proactively identify and defend against system vulnerabilities before attackers automate their attacks.

Quick Start#

Installation Difficulty: Medium - Requires Python environment and configuration of multiple API keys

# Clone repository
git clone https://github.com/PurpleCHOIms/Decepticon.git
cd Decepticon

# Create virtual environment and install dependencies
uv venv
uv pip install -e .

# Configure environment variables
cp .env.example .env

Is this suitable for my scenario?

• ✅ Security Research & Penetration Testing: Automatically performs reconnaissance, vulnerability discovery, initial access, and other red team tasks
• ✅ Cybersecurity Defense: Proactively identifies system weaknesses by simulating attacks
• ❌ Unauthorized System Testing: Strictly prohibited on systems without explicit authorization
• ❌ Fully Automated Security Solution: Requires human supervision and decision-making

Core Capabilities#

1. Red Team Agents - Automated Penetration Testing#

• Reconnaissance Agent: Network scanning, service enumeration, vulnerability discovery
• Initial Access Agent: Exploitation, credential attacks, system compromise
• Planned Privilege Escalation Agent: Rights elevation and lateral movement
• Planned Defense Evasion Agent: Anti-detection and stealth techniques
• Planned Persistence Agent: Maintaining access and backdoor installation
• Planned Execution Agent: Command execution and payload deployment Actual Value: Automates traditionally manual penetration testing steps, significantly improving security assessment efficiency

2. Multi-Agent System Architecture#

• Swarm Architecture: Direct peer-to-peer agent communication and collaboration
• Planned Supervisor Architecture: Centralized control with supervised workflows
• Planned Hybrid Architecture: Combined approach with both direct communication and centralized oversight
• Custom Architecture: Supports user-defined agent collaboration patterns Actual Value: Provides flexible collaboration models to adapt to security assessment scenarios of varying complexity

3. Replay Functionality#

• Execution results automatically saved in the logs/ folder
• JSON-formatted logs can be replayed via the Chat History button
• Export functionality enables community sharing Actual Value: Facilitates knowledge sharing and collaborative learning, helping users learn from others' testing cases

Technology Stack & Integration#

Development Language: Python Main Dependencies: LangChain, LangGraph (multi-agent system framework) Integration Method: API / MCP (Modular Command Protocol) tools

Ecosystem & Extensions#

• MCP Support: Tools can be loaded via the LangGraph MCP adapter, supporting both stdio and streamable_http transport protocols
• Custom Tools: Users can create custom MCP tool scripts in the src/tools/mcp/ directory
• Cloud Model Integration: Supports various cloud AI models including OpenAI and Anthropic
• Local Model Support: Compatible with Ollama locally deployed models

Maintenance Status#

• Development Activity: Actively developed with regular updates and an active community
• Recent Updates: Recent code commits and feature updates
• Community Response: Has a Discord community and encourages users to contribute test scenarios and improvement suggestions

Commercial & Licensing#

License: Apache 2.0

• ✅ Commercial Use: Permitted
• ✅ Modifications: Permitted
• ⚠️ Restrictions: Must obtain explicit authorization before use, not to be used on unauthorized systems

Documentation & Learning Resources#

• Documentation Quality: Basic
• Official Documentation: https://github.com/PurpleAILAB/Decepticon
• Example Code: Basic installation and usage examples provided