# numasec

An open-source AI security agent that runs in a terminal, following runbooks and switching between specialized agents while tracking findings, evidence and reports for authorized application-security and penetration-testing work.

numasec is a terminal-based AI security agent aimed at people who already work across a shell, browser, HTTP requests, scanners, advisories, notes and reports during security engagements. It uses tools already installed on the operator's machine, follows security runbooks, and switches between different specialized cyber agents while keeping the operation's context, findings and evidence together as the work progresses.

Rather than acting as a chatbot or a wrapper around an existing scanner, it gives the model a persistent security workspace: target and scope, available tools, runbook state, findings, evidence, a replay history and a report in progress, all tracked together so results can be turned into a report afterward. The authors describe it as strongest today for authorized application-security and penetration-testing workflows.

It installs with npm and includes a full terminal recording as a demonstration of its workflow.

Website: https://github.com/FrancescoStabile/numasec