I’m on the Teams plan, and there is no such information available. And, of course, I’m not going to ping an admin every day to check my usage. I’d rather find a less idiotic harness.
Where does the bug appear (feature/product)?
Somewhere else…
Describe the Bug
The dashboard export-usage-events-csv page no longer display the request cost.
cursor.com/api/dashboard/export-usage-events-csv
Steps to Reproduce
I just open the dashboard page.
Expected Behavior
It should display the cost per request as it did previously.
Operating System
Windows 10/11
Version Information
Cursor backoffice
Does this stop you from using Cursor
No - Cursor works, but with this issue
holy crap… i can’t believe that shady answer.. blows my mind, how this simple $$ display can’t be turned into an option… even Sam the cursor AI customer service gave the same reply as this guy… what is going on? On-demand as he claims shows $$ but it doesnt’!! it’s all token usage now! Sam even said it’s by design now that all usage is token based going forward for non-enterprise clients, so shady!!
Does AI also make product decisions for Cursor based on a simple “let’s hide the per usage $ for non-enterprise users”? This seems like a way to “hide” something.. once you lose trust, your product goes down the tube.. such a simple concept. Anyway, I’ve noticed so many unusual things around AI/LLM usage now, almost feels like this is a sign that “the” market has topped… Anthropic/OpenAI better IPO soon, it’s becoming obvious what’s going to happen next!!
Hello,
So it seems Cursor plan is to force customer on the Teams option to move to the enterprise option.
Well, we don’t pay to see pretty reports.
We pay to have access to the inference functionality.
The good old csv file ( that exported the cost ) was better than nothing.
I can not justify ( to myself, never mention to the management people) moving to the enterprise option just because Cursor wants to charge more money for what are basically reports.
I will recommend to the management team that they explore other options.
One thing is to pay more for inference.
Another thing it to pay more just to keep track of the costs.
How much more? Who knows, the Enterprise plan doesn’t even have a cost defined in the pricing page.
it’s so obvious they made this change: it’s harder to know the delta between token usage after each task, but $$ delta is glaringly obvious… recently, they’ve made changes to the UTC vs local timing as well, then started auto selecting sub-agents to an on demand model, when i had specifically selected included models (ie grok/cursor)… sam kept saying all these are standard features and won’t admit to being a bug!
they default the code review agent which spends on demand credits, which should obviously be an opt-in feature, you only realize it by viewing the $$ usage page, but now, hard to tell with the token based usage display… somehow only the privileged enterprise users can view $$ usage amount, such a privilege, wow!
last i checked, they won’t offer this feature.. i hope enough noise is thrown at them to make this option available at minimum!
This is simply unacceptable and cannot be justified.
Previously, Cursor removed the visibility of the cost of each individual request, and now the equivalent total API spend has also been hidden.
We are engineers; we understand the difference between the cost of our subscription and the equivalent API usage cost. These are two different metrics, and having visibility into both is essential for transparency and for understanding the actual value we are receiving.
Removing this information does not simplify the experience for technical users — it removes transparency. It also raises legitimate concerns about future changes to subscription pricing and usage policies, because users lose an important reference point for evaluating the real cost of the service.
A transparent usage page should show both the individual request cost and the equivalent API consumption. Hiding these figures does not benefit users; it only makes it harder for them to make informed decisions.
if i select subagent to default to grok, then you should just use grok, why would you auto-select based on what the AI parent model thinks? that’s just insane and makes no sense..this should be flagged as a bug! If you want subagents to be auto selected by parent model, then provide that option in the “Explore Subagent Model” drop down menu. It seems you have that option now “disabled”?
Hi Kevin,
I think the main concern is being missed here. Most users did not interpret the dollar amount as an additional charge. We used it as a normalized metric to compare the cost of different models and workflows.
A total token count cannot do that. GPT, Claude, Grok and Composer have different rates, and input, output, cache-read and cache-write tokens are priced differently. The same total number of tokens can therefore represent completely different costs.
The dollar-equivalent value helped us decide which model was appropriate for a task, understand how quickly we were consuming the included allowance, and estimate what the same workflow would cost under on-demand billing.
If the previous presentation was confusing, the solution was simply to label it clearly - for example, “Estimated API-equivalent usage” versus “Actual on-demand charge.” Removing the metric entirely does not reduce confusion; it removes useful cost transparency.
What Cursor has done may be legally permissible, but from a customer-trust and product perspective it is a remarkably poor decision. When a market leader chooses opacity over transparency, it creates exactly the opportunity competitors need.
It used to be simply called “Included,” and that was the perfect way to describe it.
Well, if this was even an intentional change … that is super bad for software that we are supposed to use in a business & enterprise context!!!
I cant even comprehend how bad this decision makes you look guys 
. Please reconsider a rollback. i need to manage a budget and me and my team members need to get a feeling how much they spend based on their model choice.
Not everyone in the team can be an admin, and the admin cannot monitor this on a request-by-request basis for the team members.
I am having the exact same issue. I always pull the reports on the first day of the month for the previous month to analyze usage and spend. This month, my report shows the on-demand token usage, but $0 in the spend column. This is very frustrating. Why would they remove the dollar amount for on-demand token usage?
We are on the Teams plan, and every month I pull the same data to analyze. This morning, August 01, 2026, I logged in to our Teams account and navigated to Dashboard > Usage, set my date range 07-01-2026-0731-2026, then exported the CSV. When I opened the CSV file, both On-Demand and Included are listed in the Kind column, but under the Cost column there are only zeros 0 for every On-Demand line.
I have even looked on the Members page, and yes, there is spend showing there, but for the current billing cycle ONLY (07/28/2026-08/28/2026). Where is the historical data? How can I access it? I need this data. This doesn’t look good for Cursor. It appears as though you all are trying to hide actual costs from your customers.
Echoing what others have already stated - I manage spend for an engineering team, and find this change to be a major step back in UX. We had just completed a change to better attribute spend between users and shared automations, and this really obscures the key data we want to track.
Hi all, @here. It was not our intent to remove dollar costs from the Usage CSV export. That was an inadvertent side effect of a related change. We are pushing an update to rectify this, and I’ll follow up here shortly when it’s live.
Appreciate your patience as we get this sorted.
Hi @Rachel_Gage We’re going to get this sorted out and I’ll post an update here once the change is live and validated. The removal of the cost data from the CSV was unintentional, and we’ll be restoring it asap.
When there are so many other options available (Claude Code/Codex), how could someone possibly think it’s a good idea to force-feed users a regression?
This is the nail in the coffin for my usage of Cursor. My entire team at work has already switched over to Claude Code/Codex, and I was the last holdout because I genuinely enjoyed the product and have all my keybindings in place. To me, this move indicates that you have people making product decisions who don’t understand what users want, or are more interested in delighting C-suite execs than the people who use the product.
Will this be available only to Enterprise subscribers? If so, what do you plan to do for individual professionals?
I’m on this page multiple times a day. The dollar view was how I managed spend in real time — not once at the end of the cycle on the Spending tab, but throughout the day while I’m actually working.
Tokens don’t replace that. Different models have different prices; a token count doesn’t tell me whether I just burned through a little or a lot. Aggregate on-demand totals also don’t replace it. I need the per-day dollar breakdown in the Usage UI so I can decide whether to keep going, switch models, or stop before I blow the budget.
Hiding cost data for self-serve / Teams while still billing on-demand in dollars makes the product harder to use responsibly. If the concern was people confusing “included usage value” with “amount charged,” label those clearly — Included vs On-demand — don’t remove the unit people actually budget in.
Please restore dollar amounts on the Usage page (even behind a Tokens / $ toggle). Without that visibility, day-to-day use of Cursor gets a lot harder.
This decision has already had an unintended consequence: it pushed me to test Codex for the first time.
I was genuinely impressed. On a $20/month plan, I used around 30% of my weekly quota and completed a substantial amount of real development work.
For comparison, today I used Cursor with GPT-5.6 for a fairly mechanical repository cleanup: reviewing Markdown documentation, removing obsolete files, splitting a large rules file into focused documents, and reorganizing the repository to provide agents with cleaner context and instructions. That task alone consumed roughly $30 worth of usage.
This is exactly why the dollar-equivalent metric mattered. It allowed users to compare the real cost of models and workflows instead of comparing meaningless raw token totals.
Removing cost visibility may discourage uncomfortable comparisons, but it also encourages customers to test competing products. In my case, that comparison has not worked in Cursor’s favour.