xAI is no longer asking subscribers to opt into an autonomous agent. Grok Bot, the multi-app agent framework it launched in beta on August 11, now ships bundled into five paid tiers: SuperGrok Plus, SuperGrok Heavy, Cursor Pro+, Cursor Ultra, and Cursor Teams (Standard and Premium), according to xAI’s own news page.
That distinction matters more than a feature-list addition usually would. A subscriber who paid for Cursor Pro+ to get a coding assistant did not necessarily choose to run an agent that operates a cloud computer, signs into their apps, and works unsupervised across their inbox and browser. Bundling changes the buyer. The person who now has access to a Bot is the seat-holder on a Cursor Teams plan, not the person who evaluated whether their organization wanted one.
xAI describes Grok Bot as a persistent cloud “colleague” with its own always-on machine, browser and terminal access, and the ability to log into a user’s actual software rather than draft inside a chat window. Users can run several Bots at once, group them into a shared chat so they hand off work to each other, and teach a Bot a routine once so it repeats the job unattended afterward.
Minimal supervision is xAI’s own framing, and the company’s example roles show it is not a fixed setting but something that varies by job. A Sales Prospector Bot researches accounts and drafts outreach, but xAI says it leaves every send for a human to approve. A Digital Declutterer audits email, Drive files, and paid subscriptions continuously, and only unsubscribes or deletes if told to. A Website Builder Bot, by contrast, is trusted to buy the needed domain itself and push the finished site live without a review step, handing back a working URL rather than a draft awaiting sign-off. An Inbox Manager clears routine email and surfaces only what needs a person. A Refunds Manager finds a claim, files it, and returns with the resolved outcome already recovered.
That range of default autonomy is the part worth tracking. Some roles keep a human in the approval loop for the action that matters most (sending, spending, deleting); others complete the consequential step (buying a domain, issuing a refund) before a person sees it. xAI’s announcement does not specify whether admins on Cursor Teams plans can restrict which Bot roles are available to seat-holders, or whether the approval-required behavior is configurable versus hardcoded per role.
The release announcement does not include independent usage data, adoption numbers, or benchmark comparisons against other agent frameworks such as Anthropic’s computer-use tools or OpenAI’s Operator. It also leaves enterprise rollout details thin: xAI says larger teams can join a waitlist, without stating pricing, seat limits, or a timeline for general availability at that tier.
For any team on a Cursor Teams or SuperGrok Plus plan, the practical question this week is not whether Grok Bot works. It is whether someone with security or IT authority has audited which of the nine published Bot roles are live for their seat-holders by default, and which of those roles can act (deploy, refund, delete) before a human reviews the output.
Per xAI’s news page announcement (“Grok Bot is now included with more plans”), published following the Grok Bot beta’s August 11 launch; xAI’s post does not carry its own dateline.
