Agentic endpoint security is a runtime-focused security approach for endpoints that monitor and control AI agents, tools, plugins, extensions, and other dynamically loaded components while they are active. Instead of only scanning files or processes after they appear, it watches how trusted software behaves in the moment and enforces policy as components request access, load code, or interact with local resources.
This matters because modern endpoints, especially developer workstations, can execute more than ordinary applications. A malicious extension, compromised update channel, or over-privileged AI tool can read files, reach secrets, invoke commands, or make outbound connections inside a legitimate workflow. Agentic controls help defenders inspect those runtime trust decisions, limit risky actions, and detect unusual behavior that traditional endpoint tools may miss. In practice, it complements EDR by adding visibility into the software supply chain and into the behavior of AI-aware components on the device.