g4sk0/mergen-mcp ? reverse-engineered prompt

Build me an AI powered red team MCP server in Python that lets an MCP compatible agent like Claude or Cursor run penetration testing workflows on a target I own. It should be able to take a domain, IP, URL, or CIDR, figure out what kind of target it is, choose the right security tools, chain them together, and keep adapting as results come back. I want a simple backend, a local dashboard, saved session history in SQLite, and clean reports I can export as HTML, JSON, or CSV. Please include a plugin based design so new tools can be discovered easily, plus some built in attack chains for recon, web testing, vulnerability scanning, and basic post exploitation analysis. Make it work over MCP with a local proxy, and add setup files or scripts so it is easy to install and run on Kali or Ubuntu. If you need to look up current docs online, go ahead.