ServiceNow (NYSE: NOW) has introduced AI Passport, a runtime authorization layer designed to give enterprises more direct control over what AI agents and Model Context Protocol servers can do inside multi-agent workflows. The capability targets a growing operational problem for IT and security teams: an agent may have access to data and tools needed for a task, but that access also needs to be defined, verified and limited while the work is taking place.
The company said AI Passport issues a cryptographically signed credential that an agent can consult at runtime to verify its authorized scope. When the agent operates within that scope, it can continue its assigned work. When it moves beyond it, the system is designed to apply graduated responses, including redirecting or restricting the agent, reconnecting it with the appropriate context or ending its activity through a kill switch.
That approach differs from relying solely on an audit trail after an action has happened. ServiceNow said the credential can be grounded in live enterprise context from its Configuration Management Database, governance, risk and compliance policies, and workflow approvals. For third-party agents, the company said this creates a permissions record that can be verified rather than accepting an agent’s own statement about what it is allowed to access.
The company is building the feature on open specifications so governance products, identity providers and enforcement points can issue and verify credentials for their own and third-party agents. It said the credentials can be checked at each stage of a multi-agent workflow and can support scoped delegation and precision revocation. Those functions could give teams a way to narrow permissions as work moves among agents rather than treating access as a static, broad grant.
AI Passport is part of a wider expansion of ServiceNow’s Super Intelligence Control Tower. The company said the platform will combine AI discovery, visibility, control, measurement and compliance on one data model. Planned capabilities include shadow-AI detection for unapproved agents on a network, access mapping based on actual entitlements, continuous observability of agent traces and behavior, and policy enforcement that can move from alerts to throttling or blocking.
ServiceNow also said its AI Cost Management capability, part of IT Asset Management, will bring model, embedded software-as-a-service AI, hyperscaler and data-center GPU spending into a single view. The company said the system is intended to connect that spending with consumption, quality and realized user value, and to alert teams to potential budget overruns. This connects agent governance to an operational question that often sits with separate teams: whether an AI workload remains within both its authorized scope and its budget.
For IT organizations, the practical value of the design depends on how consistently an enterprise can connect its identity, asset, workflow and policy data to the controls. The company said AI Passport, shadow-AI detection and AI Cost Management are expected in the fourth quarter of 2026, while its Operational Resilience offering for financial services is available now. Organizations moving agents from pilots into production may find that runtime authorization and revocation address a constraint that periodic access reviews and post-incident logs cannot resolve on their own: controlling actions while they are still in progress.