How happyhackingspace/mcp-hydra works
Project type and purpose
• This is a Dockerized, AI‑assistable pentesting microservice (“HydraMCP”) written in Python.
• It implements the Model Context Protocol (MCP) as a thin server layer (server.py + banner.sh) that dispatches AI‑driven requests to a suite of external security tools (nmap, sqlmap, etc.) wrapped by custom Python modules in the toolkit/ directory.
• Purpose: expose a standard JSON‑RPC–style interface so any AI assistant (Claude, Copilot, etc.) can invoke pentesting tools and ingest their output.
File structure
• Top‑level:
• Dockerfile + requirements.txt → container build & Python dependencies.
• server.py → MCP server entrypoint.
• banner.sh → optional startup banner.
• demos/ → usage exemplars.
• toolkit/ → a plugin directory where each .py calls a CLI tool (dnsrecon.py, holehe.py, nmap.py, ocr2text.py, sqlmap.py, sublist3r.py, wpscan.py, zmap.py).
• No web front end—communication is over STDIN/STDOUT or socket (MCP).
• No database—stateless command dispatch.
• Uses a plugin pattern: each toolkit/<tool>.py implements a common interface.
README for additional insights
• Confirms MCP protocol, Docker invocation, configuration JSON for AI clients.