I’m burning through weekly usage far far far too easily doing coordinated dev work. Yes I’m already using Cursor Cloud Workers. Is there some sort of a plan for a larger plan? Like a Grok Bot specific, or a Cursor Ultra Ultra?
You may have to enable on-demand usage.
I’d consider it if I understood how much the current usage would be equivalent to at API costs. I’m not a fan of the opacity of what a weeks worth of included Grok bot usage is. If I did, at least I could gauge what kind of usage I’d be paying for in on-demand usage.
Hey @Zachariah, there’s no tier above Ultra for Grok Bot, and Cursor and SuperGrok pools don’t stack, so there’s no “Ultra Ultra” to buy. I’ve noted the demand for a bigger tier for the team.
On gauging cost: we don’t publish a dollar figure for the weekly included allowance, and there’s no fixed “$X = N requests,” since Grok Bot bills by the actual tokens and steps a task uses, not by messages. The real gauge is on-demand itself. Turn it on with a modest monthly cap at cursor.com/dashboard → Spending, and you’ll see your actual per-task cost in dollars in real time, with a ceiling so there are no surprises.
To stretch the weekly pool: start a fresh chat per task (the + at the top of the chat list), and keep delegating heavy execution to Cloud Workers, which bill against your Cursor plan usage rather than the Grok Bot weekly pool.
More: Grok Bot plans and billing. Other Ultra users comparing setups: making the weekly allowance last.
@Zachariah On-demand’s favourite pastime seems to be burning $10 notes. If you’re curious to see how long it lasts for you, enable on-demand and set a low limit like $25.
I have also asked for a Cursor Ultra Ultra plan:
Thank you for the reply Mohit.
A few thoughts that may be helpful for your team (at least from my users perspective) -
-
Some kind of metric of turn and total token usage would be super useful as the bots work, even while working using “weekly allowance”. That way future costs could be inferred. That of course would require the transparency of what weekly allowance actually is, but in this world of AI where there is a lot of distrust and often frustration about what is included in a plan’s usage (easily seen on X from all the recent and ongoing Astra usage chaos, and the grand Anthropic plan usage controversy), a little transparency would go a long long way with users, especially with agents, since agent trust is the very metric by which agents are and need to be measured by, as we as a society give so much over to them. I’d Love it if Grok Bot led the way with this.
-
As most every xAI live presentation has indicated, having long-running “specialists” have a lot of value for their specific tasks, as their context/memory remains valuable. My one project has 27 specialists in order to narrow their specific use to exactly what they do (to not pollute context), even down to “RepoMan” to work around the fact that Origin (xAI’s repo) won’t allow Grok Bot to merge with the API, so RepoMan does it with computer use. So, fresh chats are as fresh as they can be without loosing the benefit of context for each task.
-
I’m very happy with the product, and want to use it more - that’s a good thing. I just want to be able to understand, measure, and control my usage vs cost. It’s definitely in the budget to be able to spend more on Grok Bot, but I don’t like being in the dark about what that means and just experimenting. Grok Bot could be the very thing that keeps me engaged with xAI products (still having a lackluster time with Grok 4.7 compared to OpenAI models and harnesses), and a real Grok Bot usage plan and transparency about usage would go a very long way towards me staying (instead of, say, writing or modifying another agent harness to be able to use models I can understand and control usage with better).
Overall, cheers to the team for making a truly wonderful product.
~Z