Pricing changes should be explicitly announced within the Cursor IDE.
If I weren’t so obsessively about monitoring my Usage page, I would have never even known about the switch to unlimited. And if I had stepped away from my computer just a moment sooner, I would have missed the fact that I’m being charged per use again.
Yeah, this is pretty bad. I paid for the whole year impression that I’m getting unlimited agent requests. Now suddenly I’m saying I’ve used 90% of usage. What kind of a scam they are trying to run here.
I had enough of the old Pro-tariff for 2 weeks. I’m afraid to ask how it is possible to spend an annual package of the new tariff in just a few days 
I was using it for like six months. I was paying every month. When I saw this unlimited agent request, and there is a discount if you pay annually so I thought, why not pay for it. And now I can’t use unlimited.
My only issue is when I use the old rates, I get charged 0.04 and 0.08 (3.7 Thinking) etc. which is great or fine for what I work on etc. but when I use the new rates I get charged $1 on dollars and it’s insane how expensive it can get. I don’t understand it at all lol.
The fact I have to find a way to “cheat” the new system just to save burning my entire wallet in a single prompt is pretty ridiculous, any clarification or transparency would be great for the small minded people like me.
Funny how that works, isn’t it? The pricing model feels a bit… “fragile”-like it’s almost designed to trip over itself just enough to rack up extra usage. Agents looping, timing out, or misunderstanding prompts… but hey, I’m sure that’s just a coincidence. We’re probably just imagining it, right?
Assuming we’re not all complete idiots, it does start to feel like the system benefits more from mistakes than accuracy.
It’s not all Cursor’s fault. Sometimes the problem is the layer between the chair and the monitor.
That’s a smug way of saying “user error."
Sure, sometimes it’s user error-but when even basic prompts cause loops or timeouts, and we have zero way to track or manage usage, blaming “the layer between the chair and the monitor” just sounds lazy.
If the system burns through your $20 without warning and gives you no breakdown while charging $0.71 a pop, that’s not a user issue. That’s a design flaw disguised as a feature.
I haven’t run into anything like “loops or timeouts,” even though I’m connected through a VPN, which can be a bit unstable. When there is a connection issue on my end, my profile just shows “Errored, Not Charged,” so even if the problem is on my side, Crusor doesn’t bill me for it.
As for “misunderstanding prompts” — yeah, that’s just a normal part of prompt engineering. You can even go to Google AI Studio and see how just tweaking a single word can completely change the output. It’s totally normal for me to look at the process, realize I left out something important or phrased it wrong, and then go back to edit and rerun the prompt.
And sure, there are still tasks that even Gemini can’t solve… but that’s fine? We’re not in the AGI era yet.