2026-08-27 · Guide
Grok Bot Weekly Allowance: How Overflow Billing Actually Works
Wednesday emptied the included pool and the bots kept running, so Friday opened on an invoice for work you thought grok bot weekly allowance had already covered.
That is overflow billing. It is not a crash, not a spend-cap slider you missed, and not a monthly line you can divide by four. It is the second phase of a two-phase meter: an included weekly pool, then on-demand work billed from model and token cost if the plan you hold actually allows it (Grok Bot FAQ).
This page is that model. It is not the quota exceeded banner page, which is the hour after the product stops you. It is not the bill-shape page, which is how frequency multiplies while you still design a roster. The size of grok bot weekly allowance is not published. Not in dollars, not in credits, not in runs. Confirm the live usage screen and the live invoice.
Draw grok bot weekly allowance as two phases, never as a monthly line item
People budget months because seats and cards bill months. Grok Bot does not meter that way. Eligible subscriptions include a weekly usage allowance. Usage past that allowance is billed on demand from model and token cost (Grok Bot FAQ).
Phase one is the included weekly pool. Phase two is overflow: work that continues after the pool hits zero is a purchase, if overflow is on. The week does not become a month because you think in months. The pool does not become a per-bot wallet because you named four bots.
There is no Grok Bot-specific spend cap, only the account-level On-demand monthly limit (teams and enterprises). Grok bot weekly allowance is not that ceiling. It is the prepaid-looking phase. Overflow is the slope after it. Pause is the stop you own. The monthly limit is the only dollar setting, and it fences the whole account, not a bot.
| What you pictured | What the docs describe | What that picture costs you |
|---|---|---|
| A monthly AI budget split across bots | A weekly included pool, then overflow | You discover the cut on a Wednesday, not on the first of the month |
| A published dollar size for a spreadsheet | No published dollar, credit, or run figure | Any quoted number is fan fiction. Read your own invoice |
| A Bot-specific spend cap you can raise | No Grok Bot-specific cap; one account-level On-demand monthly limit | Searching for a per-bot slider while clocks keep firing |
| Overflow as a hard stop | Overflow as on-demand continuation if the plan allows it | Friday invoice for Thursday work you thought was still included |
Which plan you hold does not print the missing number. Every paid Cursor plan (Pro, Pro+, Ultra) and every self-serve Cursor Teams seat include Grok Bot, an individual SuperGrok, SuperGrok Plus, SuperGrok Heavy or X Premium+ subscription can be linked, and a one-time trial exists (plans). Confirm names and prices on cursor.com/pricing and x.ai/pricing. Seat price is the door, not the allowance.
Spend phase one from one account pool that every named bot drinks
All bots on an account share one persistent cloud computer assigned to the user, not to a bot (computer and apps). Grok bot weekly allowance follows the same grain. The pool is an account fact. A screen is a work surface. A screen is not a second tank.
That is why a quiet briefing can look healthy on Tuesday and still be the victim of a five-minute inbox clock you forgot you armed. There is no audit view of Bot actions outside Enterprise, so Friday will not name which bot drank the week (teams and enterprises). The app keeps the 20 most recent records per routine. That window is a cache, not a bill (skills, routines, and automations).
Deleting a bot does not refill the pool, and does not remove shared-computer files or sessions. It deletes that bot's routines. Nothing is team-level. Do not delete as a substitute for pause, and do not create a second bot as a second meter. Shared computer security is the isolation page. Separate bots are not a billing boundary.
If you hold both a Cursor subscription and a SuperGrok subscription that include Grok Bot, the product uses whichever has more usage. That is still one working pool. Confirm that rule on the live FAQ. Do not invent a split.
Phase one ends when the included week is gone. You will not get a receipt that names Standup Scribe versus Lead Scout. You will get silence, a banner, or a continuation. Continuation is phase two. Silence and the banner live on the quota exceeded page. Continuation is why this page exists.
Treat phase two as on-demand overflow that does not pause itself
Phase two does not announce itself as a new product. If overflow is allowed, the same routines keep firing. The briefing still arrives on Thursday. That is the trap. Healthy output is how you miss the phase change.
On-demand means billed from model and token cost. It does not mean a cheaper model. Grok Bot has no model picker, for members or admins. You cannot limp through overflow on a discount SKU. Frequency, pause, and one manual run are the levers you have.
There is still no Grok Bot-specific spend cap; the account-level On-demand monthly limit is the only fence, and a bot already working can finish past it. Overflow is a slope. A standing five-minute routine in phase two is a standing purchase. An approval does not reverse work already completed (approvals, security and privacy). Approval rules and reversibility are not a billing brake.
The undocumented wish is that phase two waits for you. It does not. If you wanted a halt, you pause. iPhone can pause and resume. Editing and testing a routine still need the desktop app; the phone can now show run history and delete a routine (mobile). The invoice is the setting you can actually feel.
Ask the product, not a thread, whether overflow is even available on this plan
"On-demand" is the documented overflow shape. It is not a promise that every account, every trial, every region, and every week will keep serving work the moment the pool hits zero. Confirm overflow in the product. Confirm it on the vendor's current pricing page. Do not confirm it from a screenshot in a thread.
If overflow is allowed, emptying grok bot weekly allowance is a phase change, not a halt. Work you keep requesting will bill. A standing routine is a standing request. Pause is the purchase decision, because the product will not make it for you.
If overflow is not allowed, emptying the pool is a halt until the week returns. Retry will not accelerate the reset. Upgrading mid-week is not a documented refill of the current weekly pool, and this page will not invent a refill rule. If you are looking at a banner, leave this page and read Grok Bot quota exceeded. That page is the next hour. This page is the model you should have had on Monday.
Either branch still has no published allowance figure and no Bot-specific spend cap. Cursor Pro at twenty dollars a month is the cheapest paid door as of the 23 September 2026 check. That is a seat price, not the size of grok bot weekly allowance. Confirm live. The cheapest way into Grok Bot is the shopping page.
Follow one Wednesday empty through a Friday invoice you did not mean to sign
Here is the worked example this page exists for. It is a shape, not a measurement. Because the allowance size is unpublished, nobody can tell you which clock hour Wednesday died. The shape is what you can trust.
You wanted mail to feel live. Twice a day felt slow. You set Inbox Triage to every five minutes, labelled mail only, draft replies, never send, and left Chief of Staff Briefing on a weekday morning clock. That pair is reasonable as jobs. It is not reasonable as clocks sharing one unpublished weekly pool.
Every five minutes is 12 fires an hour, 288 in a day. Read that as a count. Do not read it as a price. There is no published price per run.
Monday looks like a win: labels and a briefing, still phase one. Wednesday the included pool is gone, overflow is on, nobody paused. Thursday the briefing still arrives. That "win" is phase two. Friday you open billing. The surprise is that you thought grok bot weekly allowance was still paying for Thursday.
| Clock | What the roster did | What you saw | What the meter was doing |
|---|---|---|---|
| Monday 07:00 | You enable five-minute inbox plus daily briefing and standup | Useful labels. A briefing you actually open | Phase one. Included weekly pool |
| Wednesday 11:00 | Included pool hits zero. Overflow allowed. Clocks still armed | No crisis. Maybe a banner you did not treat as billing | Phase two starts. You did not pause |
| Thursday 08:00 | Same clocks, still firing | Briefing arrives. Standup draft arrives. You feel ahead | Unattended on-demand. The healthy morning is the purchase |
| Friday 09:00 | You open the usage screen or the invoice | A line you did not mean to sign | Overflow billed. The week did not refill on Thursday |
The tell is not a dead briefing. That is the other article, where overflow is off. The tell here is a live briefing after the pool died. Twice a weekday for inbox is the clock that belongs in phase one. Grok Bot and Gmail is the draft-only page. Scheduling is how to pick that clock.
Keep the two-phase model off the quota banner page and off the bill-shape page
Three searches land on three different hours of work. Mixing them is how you rewrite a charter while a five-minute clock is still buying overflow, or how you hunt a price table while the Friday invoice is already the fact.
| Question in your head | The page that answers it | What that page will not do |
|---|---|---|
| How does grok bot weekly allowance meter, pool then overflow? | This page | It will not invent a dollar size, and it will not walk the banner tap-by-tap |
| The product said quota exceeded. What do I tap now? | Grok Bot quota exceeded | It will not explain a Friday invoice for work that kept succeeding |
| How does usage add up while I am still designing clocks? | Grok Bot cost | It will not tell you that Wednesday can stay healthy in phase two |
| The pool is on fire this hour. What do I pause and delete? | How to stop grok bot overspending | It will not retell the two-phase model from scratch |
| Which SKU is the cheapest eligible door? | Cheapest way into Grok Bot | It will not size the weekly pool. No page should |
No per-Bot spend cap is the page for writing ceilings into charters, because the product has only the account-level On-demand monthly limit.
Rank standing clocks by which phase they will occupy by midweek
You cannot know when phase one ends, because the pool size is unpublished. You can still rank jobs. Keep the briefing you open in phase one. Treat a five-minute inbox poll as the job that pushes everything else into phase two.
| Job | Default clock | If Wednesday comes early | Resume later as |
|---|---|---|---|
| Inbox Triage | Twice a weekday, never every five minutes | Pause the tight clock first. Process mail yourself, or one manual run | Twice a weekday. Event trigger if the product offers one, confirmed in the app |
| Chief of Staff Briefing | Once, before you start, weekdays | Protect this for phase one. Do not let the poller spend it | Same once-a-day clock. Never share a routine with a tight poll |
| Standup Scribe | Once per working day | Wait, or write the standup yourself for one day | Once per working day |
| Lead Scout | Daily or on an event, not a poll | Wait one day | Daily, coarsest window that still answers |
| Mail Cleanup Assistant | A coarse pass you chose on purpose | Wait. Cleanup is not why you opened the invoice | Coarse. Never five minutes |
| Churn Watch | Daily | Wait unless a named account is on fire, then look yourself | Daily, not a tight poll |
Would you buy this run on Thursday with a card in your hand? If no, it does not belong in phase two. If yes, buy it once, attended, then stop. Least privilege and the safety checklist are adjacent hours.
Paste a two-phase charter that forbids the bot from buying its own continuation
Pause is this hour. The charter is how the next Wednesday is quieter. Paste a block that treats an empty pool as a wall, and overflow as a human purchase. A bot that treats quota language or "try again" as a network blip will retry, then try another route, which is more work against an empty pool or more on-demand work you did not sign.
TWO-PHASE METER (paste into every charter)
Phase one is the included weekly pool on this account. Phase two is
on-demand overflow billed from model and token cost, only if the plan
allows it. You do not know the dollar size. Never guess one. Never
print one.
If a run fails because usage is exhausted, quota is exceeded, or the
product asks you to retry later, stop. Do not retry. Do not start an
alternate route to the same job. Do not treat this as a network fault.
Write one line: pool empty, paused, waiting for you.
Never initiate spend, a credit top-up, a paid upgrade, or a purchase
to "keep working." Never send, pay, or publish in order to catch up
while the pool is empty.
OVERFLOW IS A HUMAN PURCHASE
Do not resume yourself. Do not coarsen your own clock. Do not enable
a tighter schedule to finish faster. If the operator has not paused
every standing routine and then named one attended run, you wait.
INBOX CLOCK
Twice a weekday after any empty-pool event, never every five minutes.
The last block is the boundary. You decide overflow, in the product, as a person. The bot reports that the pool is empty, or that it is about to keep working, and it waits. Put the same clause on bots that were not the poller. The briefing bot is the one you will keep because it still "works" in phase two. Split anything that shares a routine with a tight clock. Inbox Triage does not also write the standup.
A routine assigns a workflow to one Bot. Max 50 routines per Bot. Do not pack inbox and briefing onto one card so one dropdown spends two jobs. Deleting a Bot deletes its routines. Pause first. Delete only the poller you have decided to kill, on desktop, after pause. How to stop grok bot overspending is that hour.
Diagnose a surprise invoice from the phase change, not from a missing slider
The invoice is a lagging indicator. The phase change already happened. Match what you see to one move. Do not start by searching for a per-bot spend cap. There is not one; the account-level On-demand monthly limit is the only dollar setting.
| What you see | What it usually is | The one move |
|---|---|---|
| Friday on-demand line, Thursday artifacts still arrived | Phase two ran unattended after Wednesday emptied the pool | Pause every routine now. Then coarsen the tightest clock |
| Banner plus a successful retry | Overflow is on, and the retry was a purchase | Stop. Success is not a refill of grok bot weekly allowance |
| Banner on every bot, nothing completing | Overflow off, or the product halted included work | Pause all. Wait for the week. Do not upgrade to guess a refill |
| You cannot tell which bot spent the week | No audit view of Bot actions outside Enterprise | Use your own pause list and clock list. The vendor will not name the card |
A successful tiny run is not proof the included pool refilled. It is proof the product will still serve work. Read the usage screen to see which phase that work sat in. Do not revoke Gmail as a substitute for pause. Cookies and sessions live across bots. Pause first. Then open billing.
Open the live invoice and fail the check if phase two is invisible
You need a check that can fail. "I waited a while and things still worked" is how Thursday fools you. Working is the phase-two symptom.
Before you resume anything, write four facts in a note you own: the time you paused, whether the product said overflow was available, which clocks were armed on Wednesday, and which one bot you are willing to test. Then open the live usage screen and the live invoice. Not a thread. Not a remembered number.
The check fails if overflow lines exist and you have no pause list, or if you are about to resume every five minutes because the briefing was useful. It passes only if you can say out loud: phase one ended on a day you can point at, you paused, you bought at most one attended run if overflow was on, and the five-minute inbox routine is still paused.
Then, if you test, run one bot once, manually, with a tiny job. No second bot. No clock. If that run fails with quota language, you are not in a refilled week. If it succeeds, look at the usage screen again. Success plus an on-demand bump means you just bought the test.
iPhone can pause and resume, and it can show run history. Resume nothing from the phone until you have desktop. Switching devices does not refill the week. There are desktop apps for macOS, Windows and Linux, and companion apps for iPhone (iOS 18 or later) and Android (9 or later); the iOS app also runs on iPad (FAQ). See supported platforms. The computer is a managed Linux VM, not a second pool.
Concede that overflow is the product working, then show why the invoice still shocks
The strongest case against treating Friday as a failure is simple. Overflow is documented. You wanted the Thursday briefing. You got it. The five-minute inbox bot labelled a lead at 11:05 that a twice-daily clock would have left until evening. On-demand from model and token cost is the price of that speed. If the plan allows it, the "surprise" is that you did not read the FAQ. The product worked.
That argument wins for a short, named window: a launch morning, an incident, you at the desk, overflow confirmed in the product, clock back to twice a day at lunch. One attended manual run is the clean version of the same idea. You are buying a run. You know you are buying it. Event triggers, where the product offers them, are cleaner than 288 fires on a dead Sunday. Confirm trigger types in the app.
It loses as a standing setting. You will not coarsen the clock at lunch. The weekly pool is unpublished, so you cannot prove a tight poll fits inside grok bot weekly allowance. You find out when Friday bills phase two, or when Chief of Staff Briefing goes quiet on the other branch. There is still no audit view outside Enterprise and no Bot-specific spend cap, so the slope's only product-side fence is the account-level On-demand monthly limit, which a run already working can pass.
For a genuine one-off, you at the desk, overflow confirmed, one manual run. Then stop. The shock is an unattended roster that kept purchasing after Wednesday while still looking like phase one.
Keep trials, invented credits, and mid-week upgrades outside the two-phase story
This model assumes a paid eligible plan, a weekly pool, and unattended routines. Mixing other meters in is how people invent numbers.
The one-time trial is limited usage, not grok bot weekly allowance with a published overflow slope. Do not invent a dollar figure for the credit. The trial page is the right place.
A plan change is not documented as a refill of the current weekly pool. Confirm in the product. If you were changing SKUs anyway, do it as a door decision on the cheapest-way page, not as an emergency top-up. No published page prints grok bot weekly allowance as a credit count. Do not paste a thread's number into a charter.
Overnight, pause every bot that is not the incident bot. A warm fleet is how phase two starts at 3am. A long attended sweep is a purchase you are present for. Do not convert it into a scheduled job so it can finish overnight. Teach-by-demonstration is work: up to ten minutes, draft skill, browser only, unavailable on iPhone. Do not start one to "see if the pool is back."
Whether Grok Bot is worth another week is a question for after you can draw the two phases: is Grok Bot worth it. Worth is not a published allowance. Worth is whether you will pause on Wednesday.
Make every overflow run a purchase you signed, never a roster that kept going
Everywhere else on this site a boundary is the action the bot never takes without a human. On grok bot weekly allowance, you need a boundary too, because the product will not cap you and phase two can look like success.
You never let a standing routine continue after you know the included pool is gone, unless you have named that continuation as a purchase. You never invent a dollar figure or a credit count. You never resume a five-minute inbox routine in the same week Wednesday emptied. You never let the bot initiate spend to keep working, and you never delete a bot as a substitute for pause. A named launch window you are watching is a purchase. An unattended poll is not.
The bot's line is the two-phase clause: stop, report, wait. Inbox Triage still never sends. Mail Cleanup Assistant still never permanently deletes. When the weekly pool returns, resume from the pause list, coarsest clock first. Confirm the live invoice before you call the new week phase one. If overflow lines are still growing, you are not in phase one, no matter how healthy the briefing looks.
Keep reading: Grok Bot Quota Exceeded: What That Message Actually Means, How to Stop a Grok Bot Roster From Burning the Weekly Pool, The Cheapest Way Into Grok Bot Without Restating Stale Prices.
Frequently Asked Questions
What is grok bot weekly allowance, in the product's own shape?
Grok bot weekly allowance is the included weekly usage pool on an eligible subscription. Work inside that pool is part of the plan. Work past it is billed on demand from model and token cost if the plan allows overflow. The pool is weekly, not monthly, and it is account-wide, not per bot. All named bots drink from it the way they share one computer. The product does not publish the size in dollars, credits, or runs. Confirm the live usage screen and the live invoice rather than a quoted figure from a thread.
Does grok bot weekly allowance come with a published dollar size?
No. The FAQ describes a weekly usage allowance and on-demand overflow. It does not print a number you can budget against. The enterprise page states there is no Grok Bot-specific spend cap, and it also does not print an allowance figure. A thread that names a dollar amount or a credit count is not a source. Seat prices such as Cursor Pro+ are door prices, not the pool. If you need a number, open your own usage screen and your own invoice. This page will not invent one.
If the weekly pool empties on Wednesday, do the bots stop?
Not necessarily. If overflow is allowed on the plan you hold, emptying the included pool is a phase change, not a halt. Standing routines keep firing. That work bills on demand from model and token cost. If overflow is not allowed, the product should stop serving included work until the weekly pool returns. Confirm which branch you are on in the product. Do not assume a banner appeared. A Thursday briefing that still arrived is often how people first miss the change. The Friday invoice is how they notice phase two.
Is grok bot weekly allowance the same thing as a spend cap?
No. A spend cap is a ceiling you set so spend cannot pass it. Grok Bot does not offer a Bot-specific spend cap. Grok bot weekly allowance is the included weekly pool. Overflow after that pool is a slope, not a cap. The closest thing to a cap is the account-level On-demand monthly limit, which fences overflow for the whole account, and a run already working can finish past it. Pause is the stop you own. A charter that forbids retries, alternate routes, and any purchase, top-up, or paid upgrade is the stop you write. Approvals do not refund work already completed. Confirm the live invoice. Do not hunt a per-bot slider.