Okta Agent SSO Replaces Static Keys—but It Does Not Govern Every Agent

Okta made Agent SSO generally available on August 24, 2026, bringing identity-controlled access for compatible AI agents into its core SSO offering. XAA-supported agents can be registered in Universal Directory and use identity-governed, short-lived tokens instead of hardcoded credentials or broad OAuth authorizations.

The August 24 release does not put every autonomous system under Okta’s control. TNGlobal’s independent account confirms that Agent SSO is generally available at no additional cost to customers on core SSO plans, but effective coverage still depends on the agent and connected service supporting the relevant identity and authorization signals.

Agent SSO covers known, compatible identities

Agent SSO applies Okta’s existing identity policies to an agent that participates in a supported connection. Administrators can register that agent as an identity, associate it with the user on whose behalf it acts and control its access to a sanctioned resource without distributing a permanent API key.

The replacement promised in the title is therefore specific: static credentials are replaced inside compatible Agent SSO and XAA flows. Agent SSO cannot unilaterally change the authentication method used by an agent, application, API or MCP server that does not support the required flow. Unsupported connections may still need another vendor integration, a different control or legacy credentials.

Okta lists out-of-the-box ecosystem support involving Anthropic’s Claude, Archestra.AI, Asana, Atlassian, Canva, Datadog, Figma, Glean, Granola, Linear, MintMCP, Notion, Serval, Slack and Supabase. This is not an exhaustive compatibility matrix: inclusion in the ecosystem does not establish that every product edition, agent-to-resource pairing or requesting and resource role is available in the same deployment state.

Cross App Access governs the connection

Cross App Access, or XAA, is the protocol layer behind Agent SSO. It extends OAuth and has been incorporated into the Model Context Protocol as its Enterprise-Managed Authorization extension. In a supported flow, identity and enterprise policy follow an agent’s request across applications, allowing access to be tied to the agent, the represented user and the destination resource.

The resulting token is short-lived and scoped to the authorized connection. That narrows standing credential exposure and allows policy to be applied centrally, but it does not make the identity provider a supervisor of every operation performed after access is granted. A valid token establishes what the recognized agent may reach under the defined scope; it does not by itself judge every prompt, plan, tool call or business decision made within that scope.

Compatibility is required at the relevant connection points. An agent may support XAA while a desired resource does not, or a resource may be ready while the requesting tool lacks the necessary implementation. Agent SSO’s availability therefore does not mean that every possible enterprise agent-to-application connection became usable on August 24.

The three scopes are related but not interchangeable

The product boundary is clearest when identity, connection authorization and broader governance are compared directly:

• Agent SSO: included at no additional cost for customers on core Okta SSO plans. It registers known, XAA-supported agents as identities and applies centralized access policy to compatible connections.
• Cross App Access: provides the OAuth-based delegation behavior used to replace hardcoded credentials and broad authorization grants with scoped, short-lived tokens. It governs supported agent-to-app and app-to-app connections, not arbitrary agent activity.
• Okta for AI Agents: is the separate product for broader discovery, agent lifecycle governance, access reviews, telemetry and controls beyond the initial connection. The cited launch materials do not publish a universal price for that broader product.

The visibility difference is material. Agent SSO makes an agent visible after that supported agent is registered or connected; it is not a general inventory of every autonomous process in the environment. Discovery of unknown or unmanaged agents belongs to the broader governance layer.

The same limit applies to shutdown coverage. Okta’s current kill-switch documentation describes a manual Okta for AI Agents procedure that disables the agent record, linked application and authorization server to prevent new tokens. Existing tokens remain valid until they expire unless separately revoked, and automatic behavioral triggers are not currently available.

General availability does not equal universal integration

The settled part of the announcement is the product status: Agent SSO has been generally available since August 24, 2026, and qualifying core SSO customers do not pay an additional charge for it. The unresolved deployment question is which requesting agents, resource applications, protocols and connection roles are supported in a particular Okta environment.

Okta’s public material names integrations but does not provide one exhaustive matrix covering every agent-to-resource pair, product tier and rollout state. Customers therefore cannot infer complete support from a vendor name alone. They must distinguish an out-of-the-box integration from broader ecosystem participation and verify whether the relevant product can act as the requesting agent, the resource application or both.

Agent SSO is consequently a bundled identity and authorization foundation for known, compatible agents—not a replacement for the entire Okta for AI Agents product. Broader discovery, continuing lifecycle governance, runtime controls and agent-level deactivation remain separate, while unsupported connections remain outside the new SSO path.

Also read:

• Mobile Banking Puts Your Bank in Your Pocket—But Not Every App Is a Bank
• Lovable Raises $400M—Exportable Code Does Not Remove Every Lock-In
• Your UK Employer Can Monitor You—But “Available” Does Not Mean Lawful