Bottom line: EdgeDefense MCP is a free, open-source MCP server that lets Claude (or Cursor, or any MCP client) answer plain-English questions about your home network — what's connected, what an unknown device probably is, and whether anything looks wrong. It runs entirely on your machine: no account, no cloud, no admin rights, nothing uploaded.
Two steps: install the server once, then add it to whichever client you use.
1. Install the server
Requires Python 3.10+.
pip install "edgedefense-core @ git+https://github.com/splashxmoon/edgemcp.git#subdirectory=core-engine" "edgedefense-mcp @ git+https://github.com/splashxmoon/edgemcp.git#subdirectory=mcp-server"2. Add it to your client
Cursor and VS Code install directly. Claude Code and Claude Desktop don't support one-click install yet, so those buttons copy what you need instead.
Paste one of these after installing.
Two more tools exist but are switched off until you explicitly enable them, since they need administrator access: deeper traffic analysis that watches network activity for a set number of seconds and always asks for consent first.
No outbound request, ever — not even analytics. Vendor lookups read a bundled local database.
Discovery and scoring run as a normal user. Deeper traffic analysis is opt-in, with a clear prompt first.
No tool can change anything on your network. This version looks — it doesn't act.
MIT licensed. Read exactly what it does before you run it on your home network.
EdgeDefense MCP covers zero-permission device discovery and identification — enough to answer "what's on my network and is it safe." The paid EdgeDefenseAI appliance goes further: it runs continuously, uses trained ML to catch behavior a one-off scan can't, and can actually isolate a compromised device. The MCP tool and the appliance share the same open-source discovery engine, but the ML detection pipeline behind the paid product isn't in the free repo — that's a standing decision, not a limitation waiting to be lifted.
See the source, star it, or read the full architecture notes on GitHub.
Join the Waitlist