Grok Bot Usage Limits: Weekly Allowance, On-Demand Spending and Account Linking
Understand Grok Bot weekly allowance, monthly on-demand limits, account linking and trial rules, with hypothetical calculations that keep billing units separate.
Official-documentation guide, checked October 9, 2026. This article explains billing mechanics and gives hypothetical accounting examples. It does not report a measured task bill or quote a current subscription price.
Grok Bot’s included weekly usage and its monthly on-demand spending limit are different controls. Exhausting the weekly allowance does not mean a monthly subscription has expired. Enabling on-demand can allow additional work with extra charges, but the monthly limit is not described as an absolute stop in the middle of a running task. Start with those distinctions before changing a subscription or restarting a failed job. Official plans and billing.
A useful diagnosis separates four questions: does this account have access, how much included usage remains, is extra consumption enabled, and what limit applies to that consumption? A product screen saying work is paused cannot answer all four by itself. Authentication failures, missing application permissions and computer errors also have different causes; buying usage will not fix them.
Access entitlement is not a usage balance
The official plans page lists paid Cursor individual plans, Teams and enterprise arrangements, alongside access obtained by linking eligible Grok or X subscriptions. It distinguishes qualifying subscriptions from excluded ones, including SuperGrok Lite. Business access also depends on the applicable account or assigned seat. Treat the current official list and your account display as the sources for eligibility rather than assuming that any product with “Grok” in its name provides the same grant.
An entitlement answers whether you can use the product through that route. A usage balance answers how much included work remains. The account may be eligible but have exhausted its current allowance. Conversely, an existing paid model API key is not proof that a separate agent product is entitled. The service you paid for and the service reporting a limit need to be identified precisely.
Before taking action, record the account shown in Grok Bot and the billing account associated with it. For a team, identify whether the relevant access comes from an organizational seat or a personal subscription. Keep account identifiers in a private support record; they do not belong in a public comparison screenshot.
What resets weekly, and what is controlled monthly?
| Item | What it tells you | What it does not tell you |
|---|---|---|
| Included weekly usage | Remaining included capacity within the product’s current weekly window | A guaranteed number of messages or completed jobs |
| On-demand setting | Whether work beyond the included allowance can consume additional paid usage | Whether every account or administrator permits the same configuration |
| Monthly on-demand limit | The configured boundary for additional consumption over that monthly period | An absolute guarantee that a running task stops exactly at that number |
| Subscription status | Whether a paid plan is active | That its current included allowance is unused |
| Task outcome | Whether the requested work produced an accepted deliverable | A direct conversion between one task and a fixed percentage of allowance |
Read the actual reset information in your account instead of assigning a universal weekday or local midnight. Time zones can make a date look different between a screenshot and a support response. Save the displayed time and its zone when you diagnose a reset. If the interface does not expose a particular detail, mark it unknown rather than inventing an exact schedule.
The plans documentation describes trial-credit consumption using steps and tokens, rather than a simple number of user messages. This does not establish a universal conversion for every billing arrangement. Two prompts of similar length can initiate very different amounts of browsing, tool use and revision. A “one-message task” is therefore not a meaningful common billing unit.
What happens when the included allowance is exhausted?
First check whether the message actually names exhausted usage. If it asks for a login or reports an unreachable computer, follow that problem’s own diagnostic path. Restarting a task without understanding the reason can repeat work and make the usage record harder to interpret.
If included usage is exhausted and on-demand is off, the documentation describes work stopping until available capacity returns or the relevant account arrangement changes. If you prefer not to incur extra charges, keeping on-demand off is a deliberate consumption choice, not a malfunction to be corrected.
If you choose to enable extra consumption, inspect the account’s current billing details and administrator rules before resuming. The official page says on-demand consumption is billed through Cursor and counts toward its monthly limit. Do not assume that the billing destination is identical to the subscription provider through which you obtained access.
The documentation also warns that a running task can finish beyond the configured monthly limit. This matters when a task can perform many steps before ending. Do not promise that setting a small limit guarantees a strictly identical final bill. For a hard organizational spending requirement, first verify what control the account actually supports; a soft boundary cannot be converted into a hard one by stronger wording in a prompt.
Why another subscription may not add the capacity you expect
The official linking rules distinguish granting Grok Bot access from changing a Cursor subscription. Linking an eligible subscription does not simply upgrade the Cursor plan. Existing subscriptions can continue billing separately, and the grants are not described as additive just because you own more than one eligible subscription. Account linking and plans.
Check the account association before completing it. The documentation says linking is permanent and cannot freely be moved or undone. Do not experiment with a secondary account merely to see whether the button works. An incorrectly linked identity is not the same as a temporarily stale usage display.
For an already linked SuperGrok subscription upgraded to Plus or Heavy, the page notes that the grant may take up to 24 hours to refresh. If you have just made that specific change, preserve the receipt and time, verify that you are looking at the correct account, and follow the documented refresh window. Repeatedly buying or switching plans is not a sound way to diagnose delayed entitlement propagation.
This does not mean every missing grant should be ignored for a day. A plainly incorrect account, an ineligible plan or a documented error still needs attention. The distinction is between a change waiting to propagate and a change that never granted the expected entitlement.
A trial window is not unlimited trial work
The official page describes a trial usage allocation and a seven-day window. These are two separate conditions. A trial can have time remaining while its allocated usage is already exhausted. Calling it “seven days of unlimited use” would misrepresent that structure.
Check the terms for the specific route you used. The documentation distinguishes the product’s own trial, which does not automatically convert into a paid plan, from other introductory offers such as an Apple offer that may have renewal terms. Do not generalize one route’s cancellation behavior to every subscription sold through another platform.
For a useful trial, choose a small complete task and verify its output. Spending the entire allocation on an open-ended prompt can teach you less than completing a bounded document workflow. Save the original instructions and output so you can decide whether the result would have been valuable under normal paid conditions.
Calculate task cost without mixing units
The following example is hypothetical bookkeeping, not a Grok Bot invoice or pricing claim. Suppose the extra-charge display moves from 2.40 units of a displayed currency to 2.85 while one isolated task runs. The observed difference is 0.45 in that currency. That is an incremental charge observation only if no other work consumed the same account balance during the interval.
If other Bots, scheduled jobs or users are active, the account-level difference cannot automatically be attributed to this one task. Record that confounder or use a task-level breakdown when the product actually provides one. Never manufacture a task bill by dividing shared account consumption equally without labeling the allocation assumption.
If the interface shows only a usage percentage, subtracting two readings gives a percentage-point change, not a currency amount. A move from 80% remaining to 74% remaining is a six-percentage-point reduction. Without a documented conversion, it is not six dollars, six credits or six completed tasks.
For subscription allocation, choose and disclose your own accounting rule. If a hypothetical monthly subscription cost is S and you allocate it across N accepted deliverables, S/N is an accounting allocation. It is not the provider’s per-task rate. Failed outputs and unused capacity also affect what this number means.
A practical record can use this structure:
Task identifier:
Start and finish time, with time zone:
Account and applicable plan, recorded privately:
Included usage before / after, in displayed units:
On-demand charge before / after, currency if shown:
Other work active during the interval:
Output files actually delivered:
Corrections required before acceptance:
Unknown or unavailable billing fields:
Use null or “unknown” for unavailable values. Zero should mean an observed zero, not missing access to a billing page. Keep historical screenshots or exports with their timestamps; a new plan price cannot retroactively explain an old charge without the old billing conditions.
Control the work as well as the spending setting
A narrower assignment is easier to inspect and account for. Specify sources, deliverables and stop conditions. For research, cap the scope of the question and require the Bot to report inaccessible sources rather than repeatedly trying indefinitely. For content work, ask for a draft and a list of unresolved facts before authorizing publication or a large second pass.
This is workflow guidance, not a guarantee of a particular charge. A prompt cannot override the product’s billing system. Its value is making the requested work finite and reducing unnecessary ambiguity about whether another revision is authorized.
Review recurring work separately. An old routine can continue consuming capacity while you are evaluating a new task. Inventory enabled routines and inspect their recent runs before attributing all usage to the latest conversation. If you pause a routine, record the change; deleting a routine is a different action and should not be used merely to make the usage list look tidy. Routines documentation.
Diagnose the symptom before choosing a remedy
| What you see | What to verify | Appropriate direction |
|---|---|---|
| Access missing after login | Correct identity, plan and linked account | Entitlement or sign-in diagnosis |
| Weekly allowance exhausted | Current weekly window and on-demand state | Wait for the shown reset or deliberately review extra-consumption options |
| Monthly limit warning | Current limit, shared usage and any running task | Inspect spending controls; do not assume a precise mid-task stop |
| Linked SuperGrok upgrade to Plus/Heavy not reflected | Correct account and elapsed documented refresh period | Preserve evidence and follow the refresh/support path |
| Trial stopped before day seven | Remaining trial allocation | Distinguish elapsed time from consumed capacity |
| Browser or connector asks for login | Authentication state of that exact route | Complete the appropriate authorization, not another subscription purchase |
| Bot says computer unreachable | Cloud-computer state | Follow computer troubleshooting rather than quota workarounds |
The official troubleshooting page separates account, task and computer failures. For unresolved billing questions, provide the plan route, relevant times, exact message and permitted billing record through the official support path. Do not post passwords, payment details or full private task histories in public.
Continue with the related Grok Bot guides
Frequently Asked Questions
- Is the monthly on-demand limit an absolute spending cap?
- The official documentation warns that an in-progress task can finish beyond it. Do not present it as a guaranteed exact stop during execution.
- Do two eligible subscriptions double the included grant?
- Do not assume so. The documented account-linking and grant rules do not make every eligible subscription additive. Check the specific account arrangement before purchasing.
- Does one prompt consume one fixed unit?
- No fixed per-prompt conversion is established here. The work can involve different steps and token consumption, so message count alone is not a useful task-cost measure.
- Is a seven-day trial unlimited for seven days?
- No. The official description includes an allocation as well as a time window. Either can determine whether further trial work is available.


