Why this matters:
Most organizations are deploying AI agents without a governed identity underneath them, without central registries, access policies, or ways to shut them down. Research shows that 69% of organizations report security concerns, including data leakage, over-privileged access, and unauthorized agent actions. These slow teams down, making security a blocker.
Development teams’ missions are two-pronged: scaling agent deployments and upholding enterprise-grade security requirements. Without the right identity foundation, teams begin custom coding for every secure hand-off between users and agents – slowing down the organization from shipping new work.
What we’re announcing:
Today, Okta and Amazon Bedrock are announcing a strategic technical collaboration to deliver the identity foundation for secure, production-ready AI agents. Okta for AI agents is a comprehensive platform to discover and register known and unknown AI agents, standardize agent access, and instantly revoke to mitigate the impact of rogue agent behavior. By combining AWS AI workloads with enterprise-grade identity security, customers can scale agents more efficiently and securely.
As one of the leading AI management platforms, Amazon Bedrock delivers the critical infrastructure to rapidly build and deploy agents capable of seamlessly acting across tools, data, and workflows. As an identity provider integrated with Amazon Bedrock, developers can leverage Okta for AI Agents to secure these agents from the first line of code with human-in-the-loop workflows and Universal Logout.
Secure your workforces’ AI agents
This integration enables developers to use out-of-the-box agents to connect Okta identities to Amazon Bedrock agents, enabling organizations to centralize identity policy while Amazon Bedrock AgentCore orchestrates secure agent actions across distributed environments. Key features include:
- Shadow AI Agent discovery
- AI agent registration in Universal directory
- API access management, with least-privilege enforcement
- Privileged credential management with automated rotation
- Governance for Agents as a Resource
- Agent Gateway with virtual MCP server support
- Universal Logout, an enterprise-wide kill switch for rogue agents
- System logs for full audit and SIEM integration
To learn more, visit our landing page to get started.