AI Agent Identity

AI Agent Identity is the digital identity of an AI agent that accesses applications, interfaces and data independently or on behalf of a human and performs actions in the process. It makes it traceable which agent is acting, on whose behalf, with which rights and for what purpose. The term agentic identity is used synonymously.

AI agents differ from classic service accounts: they make their own decisions, dynamically chain tools and data sources, and often act as delegates for a user. The central requirement is therefore the separation between the rights of the principal and those of the agent. An agent should receive only narrowly limited, purpose-bound rights and not simply inherit the user's privileges. Technically, this frequently relies on OAuth-based delegation with short-lived tokens and clearly limited permission scopes; the Model Context Protocol (MCP), through which agents connect tools, also relies on OAuth for authorization.

Added to this are accountability and traceability: every action of an agent should be logged and attributable to a responsible person or team. AI agent identity is a sub-area of Non-Human Identity (NHI), transfers the Least Privilege principle to autonomous systems and is a growing field of application of Identity Security.