Product updates and improvements
A new Permissions tab lets you see and control which agent actions are set to "Allow always" vs. "Ask," and every organization now gets the platform's own Allow/Deny/Allow always confirmation prompt before an agent edits shared knowledge files.
A new image model built for crisp text, logos, diagrams, and transparent backgrounds is now selectable alongside the existing models.
A toggle on the Schedules page reveals recurring schedules that are currently paused, instead of having them silently disappear.
Two new frontier models — GPT-6 Astra and Claude Fable 5.1 — are now available for your agents to use, giving you more options for balancing quality, speed, and cost.
You can now reach support by email in addition to chat, so you're not stuck waiting in a chat window to get help.
Generated presentations can now pull in your own images directly, instead of relying only on stock or AI-generated visuals.
Generated and edited videos can now have captions burned directly into the footage, so they're viewable anywhere without relying on a separate subtitle track.
