Back to Home
double16 icon

shyhurricane

by double16

Overview

An MCP server to assist AI in offensive security testing, addressing challenges with LLMs executing noisy/long-running commands, inefficient website enumeration, and prolonged port scans, by providing consistent tools and indexing content.

Installation

Run Command
docker compose up -d

Environment Variables

  • OLLAMA_MODEL
  • OLLAMA_HOST
  • GEMINI_API_KEY
  • GEMINI_MODEL
  • OPENAI_MODEL
  • OPENAI_API_KEY
  • BEDROCK_MODEL
  • AWS_ACCESS_KEY_ID
  • AWS_SECRET_ACCESS_KEY
  • OAST_PROVIDER
  • WEBHOOK_API_KEY
  • INTERACT_SERVER
  • INTERACT_TOKEN
  • DISABLE_ELICITATION
  • LOW_POWER
  • OPEN_WORLD
  • ASSISTANT_TOOLS

Security Notes

This server is designed for offensive security testing, inherently involving high-risk operations. It grants significant privileges to the AI agent, including arbitrary Unix command execution via a Docker-in-Docker setup, which runs with `--privileged` flags and can build/manage other containers. This means a compromised or misbehaving AI could fully compromise the host's Docker daemon. The `DISABLE_ELICITATION: True` setting (default in `docker-compose.dev.yml`) allows the AI to execute commands without explicit user confirmation, further increasing risk. Logging of HTTP requests/responses by `mitm_to_katana.py` and the use of OAST providers, while standard in security testing, could be vectors for sensitive data exposure or exfiltration if not properly isolated.

Similar Servers

Stats

Interest Score33
Security Score3
Cost ClassHigh
Avg Tokens20000
Stars3
Forks1
Last Update2025-12-06

Tags

offensive securityAILLMpenetration testingvulnerability analysisweb security