@kevinn my usage has certainly dropped since this change was made since I have no idea how much I’m spending day to day and can’t risk a skyrocketing bill. I’m also looking into alternatives.
Here is another concrete example of why the current usage reporting is not just confusing—it is internally inconsistent.
I ran two composer-2.5-fast sessions that spawned four subagents using Grok 4.5, Claude Opus, and GPT-5.6 Sol. While those subagents were executing, they appeared in the Usage table as separate rows with their actual models.
As soon as the sessions completed, those rows disappeared. Their usage was rolled into two “Included” rows whose Model column says composer-2.5-fast: 6.5M and 5.1M tokens.
The screenshot shows the contradiction clearly. The “Group By: Model” chart attributes the 11.7M total across the actual models—including 5.1M Grok, 3.2M Opus, 1.7M Sol, and only 792.6K Composer—while the table below claims the same usage consists entirely of two composer-2.5-fast sessions.
So after completion, the column labelled “Model” no longer identifies the model that consumed the tokens. It identifies the parent session or orchestrator. That is materially different information.
This also makes it impossible to determine how much of my Ultra plan’s $400 API-model allowance was consumed by the frontier-model subagents versus the separate included Composer usage. Cursor’s own pricing policy says model usage is valued using the applicable pricing for each model, so preserving the actual model attribution is essential.
If Cursor is determined not to show dollar amounts, then at an absolute minimum:
- Keep completed subagent usage attributed to the model that actually ran it.
- Show the parent–subagent relationship instead of replacing every subagent with the parent model.
- Provide a per-model breakdown within each completed session.
- Show which included allowance or usage pool each entry consumed.
Rolling everything into a parent-session total may be useful as an additional view. Presenting it under a column called “Model” as composer-2.5-fast, however, is misleading and makes the token-only reporting impossible to audit.
It’s very clear to me that Cursor doesn’t want users to understand. Let’s stop trying to convince them with this argument. The only argument that will convince them is the loss of users when they compare Cursor to equally unclear tools like Codex and Claude Code.
Unfortunately, one of Cursor’s biggest selling points is gone.
I guess Cursor folks aren’t planning to bring back the $$ amount in the usage page, in particular the top line summary usage? Has anyone heard anything from them on this matter?
Still only seeing token usage only on the Chart that really matters - daily totals used to show $ costs - please revert to the previous layout showing actual costs
There is no way of gauging model performance without a real accounting of hard dollars consumed by each LLM on a daily basis. Now I am seeing ZERO metrics for all my usage.
The purposes behind the elimination of any cost accountability is purely nefarious. No one does this and at cursors scale, this practice is indefensible.
Please revert this change immediately. Even if it’s marked as Included, please display the USD cost. I’m fed up with the platform’s attitude of trying to hide Included costs. I’m starting to regret signing up for the annual plan, and I’ve already subscribed to Codex and Claude Code.
When I compared token counts between Claude Code’s $20 plan and Cursor’s $60 plan, Claude Code’s Opus 5 was able to process more tokens than Cursor’s Grok-4.5. Based on this result, Cursor’s plan doesn’t seem like a good deal.
This change is causing a lot of unnecessary confusion. Please revert it. Users are angry.
After these terrible changes, I won’t be renewing my Cursor subscription. ChatGPT Codex and Claude Code are better and cheaper anyway.
Their AI support agent, good old Sam, replies as follows: I don’t have a timeline or roadmap I can share on bringing dollar amounts back to the usage view. Token-based usage is the current standard there, and there’s no option to switch it. For dollar charge amounts, check the Billing tab: https://cursor.com/dashboard?tab=billing
I’ve logged your feedback about wanting granular dollar usage in the usage dashboard and will share it with the product team.
conclusion: looks like Cursor mgmt team will not revert this ridiculous change and bring back $$ usage any time soon, unless everyone threatens to get off of Cursor due to this issue. I’m in that camp now when renewal pops up!
@kevinn it’s time to respond to the community now. We deserve transparency and communication here, especially since many of us have been with Cursor since the very early days. This is not how you treat customers, it does not build brand loyalty.
I understand that Cursor is trying to make a profit, and while this may seem like an easy way to distance the actual spend from our understanding of it, it is going to drive Cursor’s move loyal advocates permanently away. This is something that must be addressed, it’s shocking that it’s taken this much time and waffling back and forth already, only to have you ghost the community.
My only ask is that Cursor stops actively being anti-consumer, and embrace the amazing community that has rallied behind you. Show us respect and have a little dignity here. If this is not coming back, just say that and explain why it isn’t so that we can understand Cursor’s point of view, because right now it’s pointing in a very dark direction.
I’ve not been an early user of Cursor, so can I ask if you noticed any noticeable trends/changes since the SpaceX buyout of Cursor was announced last month? I believe the deal is not complete yet, so just wondering…
As I’ve mentioned before in this thread, I don’t think Cursor will reverse its decision or that @kevinn will have good news for the community. This decision is deliberately aimed at maintaining a lack of transparency in order to maximize profits, and I’m very disappointed by that.
The only thing I can recommend is, since Cursor is now just as opaque as the other tools, try Codex and see if it’s right for you.