I used all my GrokBot inclusive tokens and switched to the API with $50 dollars thinking the 3 day wait I had would easily be enough - 10 mins later and just a few tasks and I checked the usage expecting a few cents… $27 
Is this thing running Fable extra high?
Suffice to say, I’m now on Cursor using the tokens I have there until I can bring GrokBot back online.
Aside from that - I am LOVING GrokBot… this thing could develop out to be a company in a box.
My GrokBot weekly usage just went from almost 60% to 1%.
Then, it went up to 3% with very little usage. Maybe they reset the pools to a lower weekly amount?
We did just do a reset!
We’re aware that in some cases, limits are depleting quickly. The team is working on this.
Yes, seeing that exactly. Went from newly reset 1% to 3% with a few turns.
Thank you so much for the reset while the team is working on this!
I am desperate to carry on using your awesome new product! Cursor Ultra @ $200/m is the max I can afford rn though. I’ve already consumed 100% of my weekly allowance today - after yesterday’s reset 
I really appreciate everything the team is doing to help! Thank you.
I’ve been absolutely loving Grok Bot so far and I can’t wait to find even more uses for my various projects, however I do agree that even on a Pro+ plan the usage limits are way too low. I hit 77% usage limit in just two days, which feels way too much. And I haven’t even used it that much!
Also, while I appreciate that the bots use models depending on the task in a seamless way, I would prefer to have control over it. For example, I noticed on my billing page that at one point a bot used Claude Sonnet 4.5… why? Why that and not Sonnet 5? Or even better, why not use a Cursor model since I have a much larger usage pool for that instead?
So being able to blacklist old models, for example, or force bots to use a specific one would be great!
Love the features and ease of use. But costs are abnormally high! If you compare the kind of usage you get on ChatGPT with a Pro account, it feels like there is something amiss (maybe routing related). Hoping that the team figures out what is happening and we get a lot more usage for the $. People are running Hermes and Chinese models doing similar stuff for peanuts.
Grok Bot is great, but it is expensive.
I took the Supergrok plus plan to try it out and exhausted the allowance within a day (I was hooked!). Activated the on-demand with 20$ which ran out in 3 hours with very few interactions. I saw later in usage that the free allowance was 200$.
even though the tool is great, committing to it as a daily too comes at a high price
Purchased the max plan yesterday and depleted it within a day. Never felt so disappointed in my life.
Guys… feeling the love for Grok Bot (GB) here and it’s the same theme where the issue lies.
Just now I’ve run out and I have a 4 day wait… It’s literally running my business dev side now and I need to get back on it, but the open API costs make Fable look cheap.
I’m on Ultra, and I will have to swap to my Cursor agents who are not as well suited to do a bunch of tasks and keep a running log of what they do so they can run a handover when GB is back online. It’s literally like they have gone on holiday! 
Seriously though, GB is the single best AI app I have ever used - not perfect, but the potential is mind bending. Please, please use me and the guys here as the test bed for adaptive token use on Ultra?
Cheers very muchly!
Jimbo
when will be the next reset? it is my Day 1 using Grokbot and it is already 0% in 2 hours 
While the concept of Grok Bot is great, at least for my use case ( developing a new platform ) is too expensive. It defaults to the most expensive model, I loaded up on credits when i first started, then after I went to 0% usage left in the first few hours, I then changed to a less expensive model, but even when doing that i again was met with 0% left the very next day and then hae to wait for a “reset” in another 6 days. This is not realistic for someone who wants to make progress with their software platform. So regardless of the model, and dumping nearly $200 into the platform …when I have exhaused the usage in less than 48 hours of using it, Its not going to be my go-to system regardless of how cool the setup is and ease of use is. I like the idea of it but its just ridiculously expensive. Maybe I’ll play around with it again someday another time (if the costs/expense comes down)…but for now…its a hard NO.
I have 13 bots. I don’t want to reduce the number of bots like yourselves suggested in another thread; I want 1 role per bot. I’m using about 5 or 6 bots actively.
Consumption is pathetic! 1 week’s usage on Cursor Ultra lasts about 8 hours. 2 weeks in so far. Highly likely that I won’t be renewing with such insanely high usage/such low included weekly usage - and no higher subscription available.
Support refuse to give me a straight price but the best I can calculate is that it would cost $438,000 to run this every day for a year, on the on-demand spending. That is $1,000 of usage included per week on Cursor Ultra, which I’m burning in 1 working day. So $1,000 * 365 days = $365,000. Plus 20% UK VAT = $438,000 
Just hire humans instead of using Grok Bot, then?!
I categorically refuse to enable on-demand spending, without a clear understanding of the pricing.
I already deleted all non-essential routines before starting week 2. It doesn’t help much.
Cursor needs to get a grip of their pricing! You are going to lose your customers just as fast as you win them. Just offer a higher subscription… why is $200/m the max? Super Grok Heavy @ $300/m doesn’t give any extra usage, from what I can gather.
Also, 16GB of RAM on the VM is not enough. We need to be able to upgrade the RAM.
GrokBots work like Cursor Multitask mode. Consumption is quadratic in the length of the conversation. Once a conversation gets long, make another instance of your bot. Don’t keep going in the same one. That will dramatically affect your costs. (This cost pattern is why multitask exists, so long running can split the cost across multiple bots. Each being a smaller run reduces cost. But on multi-hour scales, you should split the main thread too.)
Same here, Danny. On Ultra, I used 65% of my Grok Bot weekly allowance on day one and reached 99% by day three, even after cutting back. I love the tool, but on-demand isn’t affordable for me.
Have you found a way to make it last the week—shorter conversations, fewer routines, or delegating more to Cursor Agents?
For the Cursor team: in my own use of Codex Ultra with Astra, I see around 500–700 million tokens per day and my weekly allowance generally lasts the week. I know token accounting differs, but the practical gap is huge.
At this rate, Grok Bot is unsustainable for my serious, daily work on a $200/month subscription.
What workflow is Ultra actually intended to support, and how can we use it productively throughout the week without on-demand spending?
but this totally makes no sense, bots are meant to be durable, looks like a product bug 
@olegdater My usage has SIGNIFICANTLY improved! I haven’t changed anything, I’ve actually gone even harder this week! (Week 3 now for me)
I’ve now got 14 bots and I’m using about 8 of them actively. I’ve also started using /poteto-mode even MORE heavily! (From Lauren’s pstack)
My token usage has skyrocketed (but maybe that’s a good thing?) Use this tool to see yours: Tokens Analyser
What’s changed is that Cursor has fixed the usage. See Best way to export memory - #7 by DannyB
I’m now seeing 2 full days of work out of 1 week’s usage, instead of 1 day’s work from 1 week’s usage. And to be honest I have a queue of work to go through in my #completed-work Slack channel, on the off days 
(Everything gets tested and approved/rejected by my QA bot in a live environment in their browser, but I still gate the completed work to review each PR and test them myself)
…Although my “Other Models” monthly usage has hit 100% extremely rapidly this week (it was at about 25% used before this week, at 3 weeks in), then it ate it all very fast on Cloud Agents - I believe). My “Cursor Models” has barely been touched though, about 25% used total this month.
