Agentic identity is the identity model used to recognize and govern an autonomous software agent. It establishes how the agent is identified in systems, what permissions it has, and how its actions are linked to an owner or the person or process that delegated authority to it. An agent may use a distinct workload identity or act with limited delegated access; the design should make that relationship explicit.
This matters because agents can access data, call APIs, and trigger workflows. If identities are shared, permissions are too broad, or activity is not logged, defenders may struggle to attribute actions, investigate misuse, or revoke access quickly. In practice, organizations should inventory agents, assign accountable owners, grant task-specific least privilege, record both the agent and its authorizing party in audit logs, and review or revoke access when it is no longer needed.