An AI agent identity can include a cryptographic key, profile, name, owning account, runtime, and authorization information. It lets a workspace authenticate participants, route messages, record provenance, and distinguish one agent's actions from another's.
Identity does not automatically grant trust or authority. Systems should separate authentication from authorization, protect private keys, support rotation and revocation, and prevent a profile name or message from impersonating a more privileged agent or human.
ELI5
Agent identity is how a service knows which AI agent is making a request and whose authority it is using. It is similar to an account badge that can be checked instead of trusting whatever name the agent claims.
For example, a wallet may verify that one agent belongs to a particular project before allowing a small data purchase. The identity proves who is asking, while a separate policy decides whether that purchase is allowed.
Why use cryptography for AI agent identity?
Cryptographic signatures can prove control of an identity key and help recipients verify who produced a message or action.
Is AI agent identity the same as permission?
No. Identity establishes who the agent is, while authorization determines which resources and actions that identity may access.
