Trust3 AI announces AgentDOS, which it describes as an enterprise “control plane” aimed at improving visibility into how AI agents operate inside organizations. The company says many enterprises deploy AI agents with limited insight into what actions the agents take, what data they access, and how much compute or usage they consume. AgentDOS is presented as providing observability across agent activity and data access, along with real-time monitoring of token consumption. According to coverage of the announcement, the monitoring capabilities are intended to apply across platforms that enterprises use to build or run agents, including Databricks Agent Bricks and Microsoft Copilot Studio. The announcement frames the product as responding to a growing risk as AI adoption scales, including scenarios where autonomous agents operate beyond intended scope, reach regulated or sensitive data, or increase token usage without clear accountability. The stated purpose of AgentDOS is to help organizations track and control these aspects of agent behavior through a unified visibility layer.