██╗  ██╗███████╗██╗  ██╗
  ██║  ██║██╔════╝╚██╗██╔╝
  ███████║███████╗ ╚███╔╝
  ██╔══██║██╔════╝ ██╔██╗
  ██║  ██║███████╗██╔╝ ██╗
  ╚═╝  ╚═╝╚══════╝ ╚═╝  ╚═╝

AI-Powered Pentesting Assistant for Terminal

About

Hex is a terminal-native AI pentesting assistant built with React + Ink. Chat with AI, execute security tools through natural language, create reusable attack workflows with skills, and automatically install missing tools — all without leaving your terminal. Supports 13 AI providers (5 local, 7 cloud, and 1 custom endpoint) and optional Docker isolation with Kali Linux.

Installation

npm install -g hex-ai
hex

The setup wizard configures your AI provider and execution mode on first run.

Recommended Local Model

For unrestricted pentesting assistance, we recommend running llama2-uncensored locally with Ollama:

ollama pull llama2-uncensored

Unlike cloud providers, local models have no content filters — ideal for security research and pentesting workflows where you need complete freedom in tool usage and analysis.

Development

git clone https://github.com/SK3CHI3/Hex-.git
cd Hex-
npm install
npm start

Documentation

Complete guides and references:

| Document | Description | |----------|-------------| | Features | Complete feature overview | | Setup Guide | Installation, configuration, and first run | | Quick Start | CLI commands and usage examples | | Multi-Provider AI | Configure 13 AI providers (OpenAI, Anthropic, Ollama, etc.) | | Tool Arsenal | 17 built-in tools + automatic tool installation | | Custom Tools | Install and use additional tools beyond the built-in set | | Architecture | How Hex works under the hood | | Local LLM | Running AI models locally with Ollama, LM Studio, etc. |

Tech Stack

• Runtime: Node.js 18+ with React + Ink for terminal UI
• UI: Component-based architecture with semantic color themes
• AI: 13 providers — OpenAI, Anthropic, Google, DeepSeek, Ollama, LM Studio, and more
• Execution: Direct (default) or Docker + Kali Linux
• Storage: Local JSON in ~/.hex/
• Skills: Reusable multi-step attack workflows

Key Features

• 17 Built-in Tools — nmap, sqlmap, hydra, hashcat, nikto, gobuster, and more
• Automatic Tool Installation — AI installs missing tools on-the-fly via install_tool
• Skills System — Create reusable attack workflows with /skill command
• Agentic Loop — Multi-step task execution with automatic tool chaining (up to 100 rounds)
• Web Search — DuckDuckGo integration for OSINT, CVE research, and documentation
• Multi-Provider AI — Switch between 13 providers with /provider command
• React + Ink UI — Fixed input at bottom, scrolling output, syntax highlighting
• Thinking Models — Support for models with reasoning capabilities
• Conversation Management — Persistent history with /resume and /history
• Docker Isolation — Optional Kali Linux container with 42+ pre-installed tools

Contributing

PRs welcome! See CONTRIBUTING.md for guidelines.

Hack Ethically · Learn Continuously · Share Knowledge

Report Bug · Request Feature · Security