# HexStrike AI

> Give an AI agent one interface to 150+ security tools

- Category: Agents & Automation / Security Agents
- Language: Python
- Licence: MIT
- Repository: https://github.com/0x4m4/hexstrike-ai
- Website: https://www.hexstrike.com/
- Canonical page: https://ai-tldr.dev/tools/hexstrike-ai/

## In short

HexStrike AI is an MCP server that sits between an AI client and the security tools already installed on your machine. It sits in the Security Agents category of the Agents & Automation tools and is open-source, written in Python and licensed MIT. To get started: clone the repository and create a virtualenv.

## Overview

HexStrike AI is an MCP server that sits between an AI client and the security tools already installed on your machine. Instead of teaching an agent to shell out to each scanner with its own flags, you point the agent at HexStrike and it gets one protocol-level interface covering more than 150 tools across reconnaissance, web testing, password attacks, binary analysis, cloud and container auditing, forensics and OSINT.

The project ships in two halves: `hexstrike_server.py`, a long-running server that owns the tool executions, and `hexstrike_mcp.py`, the MCP adapter an client such as Claude Desktop, Cursor, VS Code Copilot or Roo Code launches and talks to. The server does not bundle the tools themselves — Nmap, Gobuster, Ghidra and the rest are installed separately, and HexStrike wraps whatever it finds.

This is offensive-security tooling: it is built for penetration testing, bug-bounty work and security research on systems you are authorised to test. Treat the agent as you would any operator holding the same tools, and scope it accordingly.

## What it does

- One MCP surface over 150+ existing security tools, so an agent does not need a bespoke wrapper per scanner
- Network reconnaissance via Nmap, Rustscan, Masscan, Amass and Subfinder
- Web application testing via Gobuster, SQLMap, WPScan, Nuclei and Nikto
- Binary analysis and reverse engineering via Ghidra, Radare2, GDB and Binwalk
- Cloud and container auditing via Prowler, Trivy and Kube-Hunter
- Authentication testing (Hydra, Hashcat, John the Ripper) plus CTF, forensics and OSINT tooling
- Client-agnostic: works with any MCP-compatible assistant, including Claude, Cursor, VS Code Copilot and Roo Code

## Getting started

HexStrike runs as a local server that your MCP client connects to. The underlying security tools must already be installed and on your PATH — HexStrike orchestrates them, it does not ship them.

1. Clone the repository and create a virtualenv

   The README uses a project-local environment named hexstrike-env.

```bash
git clone https://github.com/0x4m4/hexstrike-ai.git
cd hexstrike-ai
python3 -m venv hexstrike-env
source hexstrike-env/bin/activate
```

2. Install the Python dependencies

```bash
pip3 install -r requirements.txt
```

3. Start the server

   Runs in the foreground and owns every tool execution. Add --debug for verbose logs or --port to move it off the default.

```bash
python3 hexstrike_server.py
# or: python3 hexstrike_server.py --port 8888 --debug
```

4. Point your MCP client at it

   For Claude Desktop, add the MCP adapter to claude_desktop_config.json and restart the app. The adapter connects out to the server you started above.

```json
{
  "mcpServers": {
    "hexstrike-ai": {
      "command": "python3",
      "args": [
        "/path/to/hexstrike-ai/hexstrike_mcp.py",
        "--server",
        "http://localhost:8888"
      ]
    }
  }
}
```

## When to use it

- Reach for it when you want an assistant to run an authorised penetration test end to end, chaining recon into scanning into exploitation attempts
- Reach for it for bug-bounty reconnaissance, where the value is breadth of tooling rather than one scanner
- Reach for it during CTFs, where the forensics, OSINT and binary-analysis wrappers cover most challenge categories
- Reach for it when you already have a hardened tool install and want an agent to use it, rather than a hosted scanner that sees your targets
