xAI has expanded the availability of Grok Bot, its always-on agent product, to SuperGrok Plus, Cursor Pro+ and all Cursor Teams plans. The company announced the change on 21 August, positioning the move as a step beyond the smaller beta audience that received the product at launch.

Grok Bot is designed as a persistent AI teammate rather than a single chat session. According to xAI, a Bot can work across tools and inboxes, continue when its user steps away and return for judgement calls or approvals. That framing matters because it puts the product closer to a delegated-workflow service: the user assigns a task, while the agent is expected to carry it through a series of actions.

Broader access after the beta launch

xAI launched Grok Bot in beta on 11 August for SuperGrok Heavy, Cursor Ultra and Cursor Teams Premium subscribers on desktop and iOS. The 21 August update broadens that access to a less restricted set of subscriptions. SuperGrok Plus users are included, as are Cursor Pro+ users and every Cursor Teams plan.

The announcement does not describe this as a general release. xAI still refers to the service as Grok Bot, and the earlier launch announcement describes it as beta. That distinction is useful for potential users: wider plan access does not by itself establish regional availability, feature parity across clients or suitability for every business workflow. Teams should check the current plan terms and controls before moving production work into the product.

For people already using Grok or Cursor, the expansion reduces a practical adoption barrier. The feature is bundled into plans many developers and teams may already hold, rather than being presented as a separate product to procure. It also connects xAI’s model offering to the day-to-day environment in which software and knowledge-work tasks are assigned and reviewed.

What an always-on agent changes

Conventional assistants respond inside a conversation and usually stop when the session ends. xAI describes Grok Bot as having its own computer and the ability to work in tools and apps on a user’s behalf. That is a more consequential operating model. It shifts the central question from whether a model can draft or answer to whether an organisation can safely scope, supervise and audit work that may run for longer than a single prompt.

The product’s approval model is therefore as important as its model capability. xAI says Bots should involve users when a decision requires their judgement. In practice, teams will want to define what counts as a decision, which connected accounts a Bot may access, and which actions must always remain human-controlled. These are particularly important boundaries for inboxes, customer systems, code repositories and any workflow that can create external commitments.

The Cursor availability is notable because it places the service near software-development work. A coding team could use a persistent agent for investigation, routine maintenance or preparation tasks, but should retain normal code review, testing and change-management safeguards. Expanded access is not a substitute for those controls; it makes it more important to establish them before agents become part of the everyday toolchain.

Availability and next steps

The official post names three new eligibility groups: SuperGrok Plus, Cursor Pro+ and Cursor Teams. It says Grok Bot is included with those plans, while the original beta launch identified desktop and iOS as the initial access surfaces. xAI’s announcement does not provide a full regional rollout matrix, so subscribers should use the product’s own plan and account pages to confirm whether access has reached their account.

This change also gives organisations a clearer signal about xAI’s product direction. Grok is not being positioned only as a model or chat interface; the company is investing in agents that can remain active, work across connected services and operate with human checkpoints. For AI buyers, the relevant comparison is increasingly between agent operating models, permissions and oversight rather than only benchmark results.

For existing eligible subscribers, the sensible first use is a bounded task with clear approval points: assemble research for review, prepare a draft from supplied materials, or investigate a reproducible engineering issue. Starting with contained work helps a team learn how the Bot handles context, hand-offs and exceptions before it is trusted with higher-impact processes.

xAI attributes the update to its official product newsroom. The announcement is an availability expansion rather than a new model release, but it materially changes who can use Grok Bot and where the service can fit in a team’s existing subscription stack.

How to evaluate the rollout

Subscribers assessing the expansion should separate access from operational readiness. The announcement confirms eligibility for the named plans, but it does not prescribe how a team should configure permissions, retain records of work or review outcomes. A useful pilot measures those practical questions alongside output quality: whether tasks are completed within scope, whether approvals arrive at the right point, and whether a colleague can understand the work performed. That evidence is more valuable than treating a bundled entitlement as automatic permission to automate an important process.