Grok Bot pricing: what we found behind the subscription

On this page

Grok Bot's access to X gives it a reason to sit on our desk. Much of the agent market announces itself there: founders post demos, customers report problems and developers argue about what actually shipped. An assistant that can follow those conversations and carry the research into a working document has an obvious use in our kind of work.

That was our reason to investigate the price. We wanted to know what the subscription buys, how much work comes with it and what happens when that work runs out. We followed the product documentation into Cursor's billing rules, compared the linked SuperGrok instructions and checked the trial and cancellation terms. We found a credible research proposition wrapped in a considerably less straightforward purchasing arrangement.

Our view is that X access is a concrete advantage for Grok Bot, but the advertised subscription price doesn't tell us the cost of keeping that research running. The documents establish access and charging rules. They don't establish a price for a finished briefing. This investigation concerns those documents; we haven't run a paid Bot workload or observed an Australian checkout.

Why we think the X connection matters

The specific capability is stronger than a vague claim that Grok knows social media. xAI's computer and apps guide says Bots can search public posts, inspect mentions and follow trends without connecting an X account. Private material such as DMs, bookmarks and the home timeline requires its X plugin and a sign-in.

For us, the attraction is a research workflow: follow a release through its announcement, developer discussion and subsequent corrections, then retain the useful material for the next investigation. That is a proposed use, not a task we've benchmarked. Public conversation is noisy evidence. Access makes reporting easier; it doesn't turn a founder's claim into a verified result.

We checked the apparent shortcut too. Tagging @bot on X can send a post into an existing Bot's working conversation, but the tagging documentation currently excludes accounts based in Australia. Public X reading and that shortcut have different availability. We wouldn't dismiss the research proposition because the shortcut is unavailable here, or advertise the shortcut as part of an Australian workflow.

The persistent computer adds to the case. Bots retain files and signed-in sessions, and one can continue work another saved. They share that computer and its credentials. We see useful continuity for research here, while different Bot names don't create separate account permissions. xAI documents that arrangement.

The subscription gets us in; it doesn't price the result

There is no separate Bot subscription on top of an eligible plan. xAI's overview lists paid individual Cursor plans, Cursor Teams and linked individual SuperGrok subscriptions as access routes.

Eligible Cursor plan	Published monthly base price
Pro	US$20
Pro+	US$60
Ultra	US$200
Teams Standard	US$40 per user
Teams Premium	US$120 per user

These figures come from Cursor's plan documentation. Its terms specify US dollars unless otherwise provided, and its pricing page excludes applicable tax. Extra usage sits above the base price.

The alternative route is less complete in the public material we retrieved. xAI's pricing page displays SuperGrok at $30/month and Plus at $100/month without a currency code in that text. Heavy is listed without a retrieved price. We haven't converted those amounts into an Australian quote or filled the gap with a guessed figure.

An existing subscriber can investigate Bot without another base subscription. That makes the access model appealing to us. It becomes harder to assess when someone pays primarily for the agent. The Bot plans guide orders allowances by tier without translating them into completed research jobs. We can verify the list price. We cannot derive a dependable cost per briefing from it.

We found that two subscriptions don't buy two combined allowances

The linking guide supplies a consequential detail: linking a qualifying Grok/X subscription doesn't cancel either subscription or add their Bot grants together. Individual SuperGrok, Plus, Heavy and X Premium+ qualify; Lite doesn't. The link is permanent and non-transferable.

We then asked which grant applies to an account holding both subscriptions. This is where the documents stopped giving us a consistent answer.

xAI's FAQ says the larger allowance applies. Cursor's billing guide says a Cursor downgrade changes its weekly grant and a SuperGrok link doesn't replace that change. We couldn't reconcile those statements into a reliable switching rule. They don't prove an account receives the wrong allowance; they leave the effect of a plan change unresolved.

This weakens the price comparison. Two plans can remain active and paid separately, under instructions that don't agree. We think the displayed grant and the consequence of changing tiers should be explicit in one authoritative account view. That is our product judgement arising from the discrepancy, not a claim about an interface we've tested.

Weekly usage and the monthly bill measure different things

Included Bot usage resets weekly. Paid continuation is measured against a monthly extra-spending limit. The settings documentation describes separate displays for those meters.

That distinction changed how we read the allowance. An illustrative request to summarise a thread and a request to investigate its claims and produce a sourced report can each begin with one message. They create different amounts of model work. The FAQ describes extra usage through model and token costs, rather than a fixed price for a job.

Cursor's overage policy says individuals opt into on-demand usage, while Teams has it enabled by default. When the weekly grant runs out, the Bot-specific rule stops work if extras are off. Enabled extras count towards the monthly limit.

Mobile adds another payment relationship. A store can bill the subscription while a separate card on Cursor's web dashboard pays for extras, according to its mobile usage instructions. Grok's consumer Extra Usage Credits are described separately in the Grok FAQ. We found no basis for treating those balances as interchangeable.

We think metered continuation is defensible for a useful job that needs more work. What remains missing from the public price story is the relationship between consumption and a satisfactory result. A longer investigation can cost more because it found something valuable, or because it went round in circles. The meter cannot distinguish them.

A spending limit can still let the current Bot finish

We checked the stopping rule because the work can run unattended. The Bot billing guide explicitly lets an active Bot finish past the cap, then stops further on-demand work.

There is a reasonable argument for completing a useful run rather than cutting it off mid-task. It also means that setting cannot be described as an exact final-invoice guarantee. Cursor's general spending documentation describes delayed enforcement and temporary credits for limited overage; increasing a limit in the same cycle can make credited usage billable. The Bot page doesn't settle how that credit applies to its completion exception.

Our preference would be a clear distinction between permission to continue a job and permission to increase its cost. The published exception makes that worth asking about. We haven't tested enforcement or received a vendor clarification.

Even “trial” covers two different offers

The trial pages gave us another reason to follow the billing trail beyond the headline.

Offer we checked	Documented ending
Web or desktop Bot usage-credit trial	Seven-day window, possibly exhausted sooner; never converts to paid. Cancelling removes remaining credit immediately.
Apple introductory subscription offer	Converts to a paid store subscription unless cancelled first.

The Bot plans page and mobile purchase guide document those paths. The Bot page doesn't offer an Android in-app trial. Existing eligible desktop access carries to mobile under the same account.

Paid web cancellation is another route: Cursor's instructions retain access through the paid period, while its terms require cancellation at least 24 hours before renewal. Apple handles its own subscriptions. A linked SuperGrok subscription remains managed through its original purchase channel.

We checked refunds separately. Cursor's policy generally conditions individual-plan review on a charge within 14 days and no subscription use in that period. Consumed on-demand usage is non-refundable under that policy, with statutory rights unaffected. Cancellation, unused trial credit and refund eligibility are different questions.

Our verdict: a clear research advantage, an incomplete cost story

We came away more interested in Grok Bot's research role than its tier ladder. Public X access, persistent files and work that passes between Bots form a coherent proposition for following a fast-moving field. That is why we think it deserves attention alongside the other personal agents we investigated.

We are less persuaded by the idea that the subscription explains what the service costs. The non-stacking grants, unresolved mixed-plan rule and active-run exception materially change that explanation. We want vendors selling ongoing agent work to make those conditions as legible as the base price.

The next evidence we need is a completed, sourced research job with its consumption and review effort recorded. That would let us judge whether Grok's access advantage becomes useful work at a worthwhile cost. This investigation establishes why we would run that test, and what we would need to observe. It doesn't pretend we've already done it.

Primary sources checked 9 October 2026. This is aecon's documentary analysis, with no paid workload, Australian checkout or vendor interview. Cover: AI-generated conceptual artwork. About Agentic Economy.