PentAGI is a fully autonomous AI Agents system capable of performing complex penetration testing tasks in isolated Docker container environments. It features multi-agent collaboration, supports multiple LLM providers, integrates knowledge graphs for experience accumulation, includes 20+ professional security tools, and generates comprehensive security assessment reports.
Overview#
PentAGI (Penetration testing Artificial General Intelligence) is an enterprise-grade autonomous penetration testing platform that deeply integrates Large Language Models (LLM) with professional security tools, enabling end-to-end automation from target reconnaissance to vulnerability exploitation.
Core Problems Solved#
- High penetration testing labor costs: Reduces dependency on expensive security experts through AI agents
- Incomplete test coverage: AI agents can execute tests 24/7, covering more attack surfaces
- Difficulty in knowledge retention: Structured storage and reuse of testing experience through Graphiti knowledge graph + Neo4j
- Complex environment setup: Docker containerized deployment, ready to use out of the box
- Time-consuming report generation: Automatically generates detailed security reports with exploitation guides
Core Capabilities#
AI Agent Capabilities#
- Fully Autonomous Execution: AI-driven decision making, automatically planning and executing penetration testing steps
- Expert Team Collaboration: Delegation system supports specialized AI agents (research, development, infrastructure agents)
- Intelligent Memory System: Long-term storage of research results and successful methods, supporting reuse in future tasks
- Context Management: Chain summary intelligently manages LLM context windows, supporting long-cycle complex tasks
Security & Isolation#
- Secure Sandbox Isolation: All operations executed in isolated Docker container environments
- Intelligent Container Management: Automatically selects appropriate Docker images based on task type
- API Token Authentication: Bearer Token authentication system supporting REST and GraphQL API programmatic access
- Self-hosted Deployment: Complete control over deployment and data, no external data leakage risk
Knowledge & Intelligence Gathering#
- Knowledge Graph Integration: Graphiti-based knowledge graph using Neo4j for semantic relationship tracking and advanced context understanding
- Web Intelligence Gathering: Built-in browser fetches latest information from web sources
- External Search Engine Integration: Supports Tavily, Traversaal, Perplexity, DuckDuckGo, Google Custom Search, Searxng
Tools & Reporting#
- Professional Penetration Testing Toolkit: 20+ built-in professional security tools including nmap, metasploit, sqlmap, etc.
- Detailed Report Generation: Generates comprehensive vulnerability reports with exploitation guides
- Process Visualization: Modern UI for real-time test process monitoring
LLM Provider Support#
| Provider | Model Series | Special Capabilities |
|---|---|---|
| OpenAI | o-series (o1, o3, o4-mini), GPT-4.1 series | Reasoning model support |
| Anthropic | Claude 4 series, Claude 3.7 | Extended Thinking |
| Google AI (Gemini) | Gemini 2.5 series | Up to 2M token context |
| AWS Bedrock | Enterprise multi-model | Enterprise compliance support |
| Ollama | Local models | Zero-cost local inference |
| Deep Infra/OpenRouter/DeepSeek/Moonshot | Multi-model proxy | Cost optimization & flexibility |
Use Cases#
| Scenario | Description |
|---|---|
| Automated Penetration Testing | Regular internal security assessments, red team exercises |
| Vulnerability Discovery & Exploitation | Automatically identify and verify exploitability of target system vulnerabilities |
| Security Report Generation | Output industry-standard security assessment documents |
| CI/CD Security Integration | Integrate into DevSecOps pipelines via REST/GraphQL API |
| Security Capability Verification | Blue team adversarial tool for validating defense measures |
System Requirements#
- Docker and Docker Compose
- Minimum 2 vCPU (4+ recommended) | Minimum 4GB RAM (8+ recommended)
- 20GB available disk space (50+ SSD recommended)
Quick Start#
# Create installation directory
mkdir -p pentagi && cd pentagi
# Download installer
wget -O installer.zip https://pentagi.com/downloads/linux/amd64/installer-latest.zip
# Extract and run
unzip installer.zip
sudo ./installer
Access: https://localhost:8443
Default credentials: admin@pentagi.com / admin
API Support#
- GraphQL API:
/api/v1/graphql - REST API:
/api/v1/ - Swagger UI:
/api/v1/swagger/index.html - GraphQL Playground:
/api/v1/graphql/playground