Identity access management company Okta Inc. today announced a new framework designed to help organizations secure the growing use of artificial intelligence agents inside enterprise environments, alongside providing details of a forthcoming platform called Okta for AI Agents that aims to implement the model in practice.

The new framework, described as a blueprint for the secure agentic enterprise, is designed to help organizations answer three foundational questions as AI agents become more widely deployed: where agents are running, what systems they can connect to and what actions they can perform.

“AI agents are evolving faster than any software before them, making traditional security models obsolete,” explains Ric Smith, president of products and technology at Okta. “With this new blueprint, Okta is establishing the industry standard for the secure agentic enterprise.”

With the blueprint, organizations first need to identify and inventory every agent operating in their environment, including unofficial or shadow agents created by employees using external tools. To support that goal, Okta is expanding its Universal Directory so that AI agents can be registered as nonhuman identities with defined ownership and lifecycle management.

The framework emphasizes centralized control over the resources that agents can access through features such as an agent gateway and application programming interface access management.

The blueprint also focuses on governing what agents are actually allowed to do once connected to enterprise systems. Okta’s platform logs agent activity, including tool usage and authorization decisions, to allow organizations to monitor behavior and detect anomalies.

If an agent deviates from its intended task or accesses sensitive data unexpectedly, organizations can immediately revoke its permissions using a universal logout mechanism designed to act as a centralized kill switch.

Okta for AI Agents, which will become generally available on April 30, uses the framework announced today to provide the operational capabilities needed to enforce controls. The platform will allow organizations to discover and register AI agents, manage their access to enterprise systems and revoke that access if an agent behaves unexpectedly. The platform is designed to integrate with existing enterprise agent development and orchestration platforms.

Okta for AI Agents will also include tools designed to detect unauthorized shadow agents created by employees connecting third-party AI tools to enterprise systems. Using the service, security teams will be able to identify those agents, evaluate their permissions and apply baseline governance controls.

To support Okta for AI Agents, Okta is also expanding its Okta Integration Network, which currently includes more than 8,000 integrations, to support AI agent platforms such as those from Boomi Inc. and DataRobot In. The integrations will allow organizations to register AI agents as governed identities, assign human ownership and apply centralized access policies.

“Securing the agentic enterprise will require industry-wide collaboration,” said Carl Siva, chief information security officer at Boomi. “By combining Boomi’s expertise in agentic connectivity and modern integration with Okta’s identity leadership, we are delivering a unified security and governance layer that helps organizations harden their security posture while maintaining auditable visibility into every agent’s actions.”

Okta argues that treating AI agents as first-class identities will become essential as enterprises increasingly deploy autonomous software capable of executing workflows, accessing sensitive data and interacting with multiple applications without direct human oversight.

Image: Okta