A $60/Month VM With an LLM Agent Can Run Autonomous Security Scans

Wallarm CEO Ivan Novikov shared on X that a low-cost virtual machine running an unconstrained LLM agent can now perform continuous, autonomous cybersecurity work. The setup, costing around $60 per month, uses tools like Sourcegraph for code search and runs without human intervention for days at a time. Novikov's approach sidesteps the content restrictions found in public coding assistants like Claude Code, which he says either refuse security tasks or silently downgrade model capability mid-session. The agent handles patch hunting, runtime tracing, and iterative code analysis — tasks that traditionally required expensive analyst hours for verification. However, experts note the system carries real risks, including scanning code without proper authorization, and still requires human oversight before any findings are acted upon.