Okta launches inline AI agent runtime gateway

Okta has launched an AI agent runtime gateway that positions the company inline with AI agent activity, allowing it to govern what autonomous agents do after gaining access, not just who gets in. The gateway monitors both the prompts entering a model and the actions it proposes, enabling real-time authorization decisions. Okta also acquired Permiso Security to add threat detection capabilities. The company plans to add intent-based authorization in 2027, using agent context and permitted tools to shape access decisions, though scoping permissions without blocking legitimate work remains an active research challenge.