Hooman20

An AI network debugging agent for Linux, built on Meta’s Muse Spark. Debugging network problems usually means running many commands and reading output by hand. This agent does it automatically in four steps. (1) It learns the system by choosing commands from a strict read-only allowlist. (2) It writes its own Python test tool for the user’s request, for example a UDP sender and receiver to measure speed and packet loss. (3) The user approves the script, and it runs in an isolated Linux network namespace with a timeout. (4) The model reads the results and explains the root cause with evidence and a next step. To test it, I built a lab that injects real faults: packet loss, delay, firewall drops, MTU mismatch, and broken DNS. In my evaluation run, it found the right cause in 6 of 6 cases.