The August 11 Grok Bot announcement describes several always-on agents working together on a shared cloud computer. Bots can message one another and coordinate in group chats, while users assign work from phone or desktop. The company’s examples span sales, operations and bug fixing; they are demonstrations of its intended workflows rather than independently measured productivity results.
The launch also describes teaching a recurring workflow by having a Bot follow along, then saving the process as a routine. It listed beta access for named SuperGrok and Cursor subscription tiers, with separate Bot usage and an enterprise waitlist. Those are the announcement’s historical terms, so prospective users should check current eligibility rather than assume the August list is unchanged.
Editorial take: Multiple agents are useful only if they reduce coordination work for the person using them. The interesting test is a task with distinct handoffs and one clear finished result. More chat participants alone are not evidence of better execution, and responsibility for approving consequential steps should remain easy to understand.