AI agents get first-class identities in Auth0; delegation chain auditable

We're excited to announce Agents as Principal is now in Early Access, giving AI agents a first-class identity in Auth0 - distinct from human users and clients. If you're running agents in production today, they're likely standing in as a shared client or user identity, which means you can't tell which agent acted, track it when it's acting for someone else, or separate its traffic from a human's. This release gives every agent its own identifier, credentials, and audit trail so you can!

• Know Exactly Which Agent Acted: Every agent gets a unique identity of its own, instead of hiding behind a shared client or a user account. You can create and manage agent identities directly, so "who did this" has a real answer.

• Track Delegation End to End: When an agent acts on a user's behalf, the full chain - who delegated, to which agent, through which app - is encoded in the token and auditable through every hop.

• Bring Your Existing Agents In: Already running agents outside Auth0 through your own registry, an MCP server, or elsewhere? Associate the identifier you already use internally with an Auth0 client, so you're not re-provisioning agents you've already built.

To enable the Agent as Principal Early Access release in your Auth0 tenant once available in your environment, please contact your Auth0 Account Team.

Learn more in the documentation.