An Agent System of Record is a central registry for AI agents. It records which agents exist, who owns them, what systems they can reach, and what permissions or policy limits apply. In enterprise security, this acts like the control plane for agent identity: every agent should have a known owner, a defined purpose, and a scoped boundary of access.
This matters because AI agents can act repeatedly across applications, often with more autonomy than a normal chatbot. Without a system of record, teams may lose track of stale agents, overbroad privileges, or unauthorized connectors. Security teams use the registry to support least privilege, approvals, revocation, and audit trails. In attacks, a compromised or prompt-injected agent may try to abuse inherited access; in defense, the registry helps prove what the agent was allowed to do and shut it off quickly when behavior looks suspicious.