PentestAI

Autonomous AI-Orchestrated Penetration Testing Platform

PentestAI is an AI-powered penetration testing automation platform that orchestrates security assessment workflows using LangGraph, LLMs, and industry-standard security tools.

Instead of executing individual tools manually, PentestAI coordinates reconnaissance, vulnerability scanning, AI analysis, human approval, optional exploitation, and automated reporting through an intelligent workflow.

---

Features

AI Orchestration

• LangGraph-based workflow engine
• AI reasoning for vulnerability analysis
• Intelligent decision making between scanning stages
• Supports local and cloud LLM providers

Automated Reconnaissance

• Subdomain Enumeration
• Passive Recon
• Active Recon
• Service Discovery

Vulnerability Assessment

• Port Scanning
• Web Discovery
• Vulnerability Scanning
• CVE Enrichment
• Risk Prioritization

Human-in-the-Loop Security

• Mandatory approval before exploitation
• Scope validation before every tool execution
• Consent confirmation
• Complete audit logging

Live Dashboard

• Real-time scan status
• Live command execution
• Streaming tool output
• AI reasoning visualization
• Scan history

Automated Reporting

• Markdown Report
• PDF Export
• Executive Summary
• Severity Ranking
• Remediation Recommendations

---

Architecture

                User

                  │

                  ▼

          React Dashboard

                  │

                  ▼

             FastAPI Backend

                  │

                  ▼

        LangGraph AI Orchestrator

                  │

      ┌───────────┼────────────┐
      │           │            │
      ▼           ▼            ▼

 Recon      Vulnerability    Reporting
 Engine        Analysis

      │
      ▼

 Human Approval

      │
      ▼

 Optional Exploitation

---

LangGraph Pipeline

Input Validation
        │
        ▼
Reconnaissance
        │
        ▼
Scanning
        │
        ▼
Vulnerability Analysis
        │
        ▼
Human Approval
        │
        ▼
(Optional)
Exploitation
        │
        ▼
Reporting

The workflow pauses before exploitation and requires explicit operator approval before continuing. :contentReference[oaicite:1]{index=1}

---

Integrated Security Tools

Recon

• Amass
• theHarvester
• Sublist3r

Scanning

• Nmap
• Masscan

Web Discovery

• ffuf
• Gobuster
• testssl.sh

Vulnerability Scanning

• Nuclei
• Nikto
• OpenVAS / Greenbone

Web Application Testing

• OWASP ZAP
• sqlmap
• Wapiti

Exploitation

• Metasploit Framework (Check Mode)

These tools are orchestrated through a unified interface rather than executed independently.

---

Technology Stack

Layer	Technology
AI Workflow	LangGraph
LLM	Ollama / Groq
Backend	FastAPI
Frontend	React + Vite
Queue	Celery
Cache	Redis
Database	SQLite
ORM	SQLAlchemy
Authentication	JWT
Reporting	Jinja2 + Markdown2 + WeasyPrint
Deployment	Docker Compose

---

Security Guardrails

PentestAI includes multiple safeguards to ensure responsible usage.

• Authorized scope validation
• Human approval checkpoint
• Consent confirmation
• Audit logging
• Timeout handling
• Rate limiting
• Uniform tool interface
• Retry protection

These controls help prevent unauthorized scanning and unsafe execution.

---

CVE Enrichment

PentestAI automatically enriches findings with:

• CVE Details
• CVSS Score
• NVD Description
• Severity Classification
• Cached Lookups

Supported through automatic extraction from tool output and metadata.

---

Multi-Target Parallel Scanning

Supports scanning multiple targets simultaneously.

Features include:

• Concurrent workers
• Live dashboard updates
• Status monitoring
• Scan history
• Queue management
• SQLite WAL optimization

---

Project Structure

PentestAI/

├── backend/
├── frontend/
├── agents/
├── tools/
├── reporting/
├── workers/
├── docker/
├── docs/
├── docker-compose.yml
└── README.md

---

Example Workflow

Create Scan
      │
      ▼

Validate Target

      │
      ▼

Recon

      │
      ▼

Scan

      │
      ▼

AI Analysis

      │
      ▼

Approval

      │
      ▼

(Optional)
Exploit

      │
      ▼

Generate Report

---

Target Users

• Penetration Testers
• Security Engineers
• Red Team Operators
• DevSecOps Teams
• Security Researchers
• Students learning offensive security

---

Roadmap

• Improved OpenVAS integration
• SQLMap JSON API support
• Better Gobuster parsing
• Alembic database migrations
• Advanced Metasploit integration
• Distributed rate limiting
• Enhanced AI planning

---

Legal Notice

PentestAI is intended only for authorized security testing.

Running security scans against systems without explicit permission may violate applicable laws.

Always ensure you have written authorization before performing any assessment.

---

License

This project is provided as a technical framework for authorized penetration testing and cybersecurity research.

Users are responsible for ensuring all usage complies with applicable laws and permissions.

---

Developed by IntelForge Spec