PingOne

AI Agents

Artificial intelligence (AI) agents are non-human identities that perform actions on behalf of users or systems. In PingOne, registering these agents as first-class OAuth 2.0 identities enables you to manage who owns them, how they authenticate, and which tools and APIs they can access.

Learn more about agents in What are AI agents?

Agent identity and lifecycle

You can onboard, enable, update, and disable agents in PingOne. Each agent has a unique identity with its own credentials and lifecycle. This enables centralized management of your agents, their owners, and their resource permissions across your customer and workforce use cases. To manage AI agent owners, you can grant the applicable users with the Application Owner role and assign the required AI agents. Learn more in Managing user roles.

To get started, add an AI agent.

Delegation, not impersonation

To securely act on a user’s behalf without impersonating them, agents use the OAuth 2.0 Token Exchange flow. The process works as follows:

  • The agent presents the authenticated user’s access token (the subject token) alongside its own client credentials (the actor token).

  • PingOne evaluates the agent’s identity, the user’s consent, and the requested scopes to issue a new, downscoped access token.

  • The delegation token contains an act (actor) claim, which clearly communicates to downstream resources that the agent is acting on behalf of a human user. The agent never sees the user’s credentials.

Learn more in Configuring OAuth 2.0 token exchange.

Least privilege permissions

The delegation token is short-lived and might require additional token exchanges as the agent performs new actions or accesses different resources. Audience restriction ensures the token is only valid at a specific resource server, preventing acceptance of unauthorized tokens. Resource and scope mappings enforce least-privilege access to backend APIs and MCP servers.

You can find an end to end use case for onboarding a digital assistant and enforcing access control in Securing AI agents with PingOne and PingGateway.

Human-in-the-loop (HITL) approvals

If an agent needs to perform a high-risk action, such as transferring funds, PingOne can enforce a human-in-the-loop (HITL) approval. Using flows like client initiated backchannel authentication (CIBA), the agent’s request is paused while the human user receives a push notification to explicitly approve or deny the action in real time.

Learn more about configuring a CIBA flow.

Attribution and auditing

Because agents operate with their own unique identities and utilize delegation tokens, downstream systems and gateways can uniquely identify both the acting agent and the originating user. This generates a centralized audit trail of all agent-initiated activity, ensuring full accountability, traceability, and compliance.

Learn more in AI agent logging and auditing.