Use Vault with agentic AI

Vault Enterprise 2.0.3+ provides native support for AI agents using agent registry governance and OAuth resource server configuration for JWT validation:

• Agent registry - Enroll, govern, and audit agentic identities.
• OAuth resource server configuration — Validate OAuth 2.0 tokens issued by trusted authorization servers so clients can authorize access to Vault directly.

Together, these capabilities let an approved agent present an OAuth 2.0 token to Vault without a separate login step. Vault validates the token, resolves the client to a Vault Identity entity, checks the entity's Agent Registry record, and enforces authorization constraints for agentic workflows.

The Agent Registry and OAuth resource server configuration each contribute their own audit metadata. Refer to the OAuth JWT metadata reference for a specification of the detailed audit metadata. When a client authenticates with an OAuth JWT, Vault does not create a stored token with an operator-defined metadata map. Instead, Vault derives token metadata from claims from the JWT each time the client presents the credential. The resulting token exists only for the lifetime of the request and Vault does not persist it to storage, so every request re-evaluates the claims in the presented JWT.

How Vault supports native AI agents

Native AI agent support combines identity, credential validation, and governance:

1. Vault validates an OAuth JWT using an OAuth resource server configuration profile.
2. Vault resolves the authenticated client to a Vault Identity entity.
3. Vault checks whether the relevant entity has an Agent Registry record.
4. Vault evaluates the entity's baseline policies and, for delegated on-behalf-of workflows, the agent's authorization ceiling.

Vault Identity answers the question "who is this client?" The Agent Registry answers the question "is this client an approved agent, and what governance constraints apply to it?" OAuth resource server configuration answers the question "which external OAuth JWT issuers does Vault trust, and how should Vault validate their tokens?"