Jaizz Noir AI-Powered Penetration Testing Suite

Multi-agent pentest automation with RAG-based CVE lookup, real-time streaming, and LLM-powered exploitation analysis.
🇺🇸 English · 🇧🇷 Português

---

Jaizz Noir is an open-source penetration testing platform that orchestrates 8 specialized AI agents across a 10-phase pipeline to automate reconnaissance, vulnerability detection, exploitation, and reporting.

Key differentiator: Unlike static scanners, Jaizz Noir uses agentic AI that selects tools, validates findings, chains vulnerabilities, and adapts its approach in real-time — mimicking the workflow of a senior penetration tester.

Quick Start

git clone https://github.com/yourusername/jaizz-noir.git
cd jaizz-noir
pip3 install -r requirements.txt
cp .env.example .env
# Edit .env with at least one API key
python3 wsgi.py

Open http://localhost:5000.

Architecture

Pipeline (10 phases)

Phase	Description
Recon	DNS resolution, domain/subdomain enumeration
Fingerprint	HTTP header analysis, technology detection (~20 signatures)
Web	Web scraping, robots.txt, security headers
Legacy Modules	OSINT/Infra/Web tools (Maryam, Naabu, Nuclei, etc.)
Orchestrator	AI-driven tool selection + execution + LLM analysis
ZAP MCP	OWASP ZAP spider + active scan
CVE	ChromaDB RAG + BNVD/NVD CVE lookup by tech version
Risk Score	Per-category risk scoring (headers, transport, CVE, etc.)
AI Analysis	Final LLM analysis with NIST CSF mapping
Report	HTML, JSON, Markdown report generation

8 AI Agents

Agent	Model	Provider	Role
Pentester	deepseek-v4-pro	free-gateway	Exploit execution
Searcher	deepseek-chat	free-gateway	OSINT/CVE research
Coder	deepseek-chat	free-gateway	PoC writing
Installer	llama-3.1-8b-instant	groq	Infrastructure setup
Enricher	llama-3.1-8b-instant	groq	MITRE ATT&CK mapping
Adviser	deepseek-v4-pro	free-gateway	Strategic guidance
Reflector	deepseek-chat	free-gateway	QA review
Planner	deepseek-v4-pro	free-gateway	Goal decomposition

8 LLM Providers

Groq · OpenAI · DeepSeek · NVIDIA NIM · Gemini · Free Gateway · Ollama · Fallback

Validation Pipeline

• GroundedPipeline — source-specific validation rules (SQLi, Nuclei, CVE IDs, etc.)
• Chain Table — 28 capability chains for vulnerability correlation
• 7-Question Gate — pre-submission validation with 18 never-submit patterns
• Scope Validator — target whitelist/blacklist, forbidden action blocking
• Execution Monitor — loop detection (same tool >5x), rate limiting, auto-intervention

Real-time Streaming

• WebSocket endpoint: /api/flows/<id>/ws
• Events: progress, phase_start, phase_end, llm_call, tool_call, finding, error, complete
• Langfuse integration for LLM cost tracking

Deploy

Render: Web Service → Python 3 → gunicorn wsgi:app Railway: Auto-detect Python project Docker: docker build -t jaizz-noir .

Documentation

• 🇺🇸 Full English README
• 🇧🇷 README completo em Português

Security Notice

For authorized security testing only. The scope validator restricts testing to approved targets. Maintainers assume no liability for misuse.