Observed arrival · 2026-08-28
Barrikade, Against Anonymous AI Agents
Barrikade is building security tools to discover AI agents across a company, identify their owners and permissions, and trace their actions.
Field notes
Lens is demonstrated through a command-line invocation that scans a repository path rather than asking visitors to trust an opaque risk score. The example checks 119 locations with 20 detectors, identifies deployment and API evidence, and explicitly declines to count AGENTS.md as an agent. Barrikade also states that its sample run collects no source bodies, prompts, credentials, or secrets; the surrounding platform is described as extending discovery into runtime.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue