Developers and business users are rapidly adopting agents like Claude Code, Cursor, and Claude Cowork to handle sensitive source code, internal data, and SaaS applications. Because these tools often inherit the permissions of the user who installed them, they can execute complex, multi-step actions without manual oversight. Noma’s platform now integrates with existing EDR and MDM systems to provide a live inventory of these local assets, enforcing access controls and runtime protections that were previously limited to cloud environments.

The new capability functions by linking identity, permissions, and behavior into a single control plane. By utilizing its AI Detection and Response (AI-DR) layer, the platform monitors agent activity for anomalies, such as prompt injection, unauthorized data exfiltration, or deviations from established intent. According to Niv Braun, CEO and co-founder of Noma Security, these agents often possess some of the company's most privileged identities, making it critical for security teams to maintain visibility even as these tools proliferate beyond centralized servers. The system allows organizations to apply granular policies—ranging from simple monitoring to blocking specific tool calls—across a wide array of agent architectures.