My milage for kilocode
Tokens: Up 3.4m Down 45.7k
Cache: 3.2m
API Cost: $3.67
2 Likes
i didn’t know about kilocode, go switch right now !!!
Windsurf is already gaining traction thanks to Cursor’s dishonest practices. By deceiving their user base, not only did they lose trust, but they also fueled the growth of their competitors. It’s disappointing, when I started paying for Cursor almost a year ago, it was an incredible product. Now, I’ve canceled my subscription because I won’t support something built on false promises.
11 Likes
My 7 days of timeout on Cursor Discord will be gone in 12 hours but tbh at this point I don’t feel like discussing there anything.
Because it feels like the only meaningful topic is which IDE to use instead after the current subscription ends.
2 Likes
Yeah, also feel for Cat and Boris, they seem to be fine people. Also for Claude Code team, it seemed to have been perfect before.
2 Likes
When u made unlimited requests it was so promising but after usage limits appering my subscription is cancelled! I hope I’ll be able to return when someone Makes Cursor Great Again!
The character ‘Danperks’ has started deleting comments and closing threads because no one believes in his propaganda about fair tariffs. As he said, if you don’t like something, unsubscribe. He believes that our opinions are an insult to the ‘Cursor’ team, and if you don’t like something, you can leave.
3 Likes
While I fully understand that the 500/requests/month for $20 was a sweetheart deal, it unfortunately is never coming back, because the cost to the company exceeds the revenue for the deal. Yes it’s bait and switch, but I don’t blame them either. This company needs to make money to be sustainable in the long term, and giving away their Venture Capital is unsustainable.
Although I’ve managed to get back my 500 requests/month deal, I still cancelled my Cursor subscription. Sorry, but “Auto Mode” is unusable. I’m going to try other tools, and if Cursor gets its act together, maybe I’ll come back.
3 Likes
It’s more than “Auto Mode” is a waste of time; you don’t realize how much time it wastes because you can’t detect when it goes off the rails. You just stop and try again, and it’s all that loss of time, trying to tweak it to do what you want. (Edit: reworded)
The company needing to make money is not the main problem. The problem is in handling the situation.
The stages are sth like this:
- “We moved you to a better new plan without any kind of notice, for your own good.”
- “We changed our plan again and we’ll claim that’s what we meant from the start.”
- “If you don’t like our plan, just cancel and stop complaining.”
4 Likes
This is precisely it; the way they handled this was unethical and unprofessional. It screams absolute amateur hour. You come here, and then you get gaslit by the moderators.
Every other company in the world gives a warning to their customers about incoming changes, which don’t get made in the darkness of the night, subtly hoping the clients won’t notice. The fact that Dan is still perpetrating the dishonesty tells everyone the calibre of people employed at Anysphere.
And don’t worry Dan, I have cancelled and I’m sure hundreds if not thousands will follow as well.
2 Likes
Using Cursor Pro plan (20USD monthly) for 4 months and till this pricing update it was so far pretty good. I could have understand if you guys have just increased plan pricings and stick to 500 amount of requests or request based pricing or claude usage pricings would have stayed static..
Biggest issue right now is I don’t know how much work I will get done by how much token or money I spent is a big question mark. My monthly usage resetted at july 9 and I am out of requests for claude sonnet products. I have checked how many claude-sonnet(3.7 and 4 thinking non-max mode) requests I have made since my reset and there are 64 requests.. 20USD / 64 = averagely makes 0.31USD per request.
Some users say they spent monthly usage with even less amount of requests..
You are saying “we switched to usage based pricing” which I could have understanded if we knew how much we are going to spend for our tasks. With this new consumption/pricing method it may charge a dollar or even higher for one request?
For the first 3 months of my Cursor experience we used to have 500 requests + slow unlimited requests and now we don’t exactly know how much work will be done but for my case lets say 64 requests, so this shows insane amount of difference between this new pricing method and previous ones.
I would have accepted something like 250 claude sonnet 4(thinking) static request amount for something between 25-35$ and unlimited requests for free models such as gemini 2.5.
I will be following your pricing updates for couple of weeks with a hope for possible reconsideration of this vagueness. If Cursor team can’t come up with a understandable solution, I will cancel my subscription and look for another alternative aswell.
4 Likes
Indeed, the way this company makes money is really quite interesting. 
“We moved you to a better new plan without any kind of notice, for your own good.”
This company unilaterally changed the user’s plan without any prior notice, then framed it as a generous move done “for your own good” In reality, it’s a forced change disguised as kindness.
“We changed our plan again and we’ll claim that’s what we meant from the start.”
This company keeps changing its policies or terms, then pretends that was always the original intent. It’s a way of rewriting history to avoid accountability.
“If you don’t like our plan, just cancel and stop complaining.”
This company shows arrogance and indifference toward user feedback. Instead of listening or improving, they adopt a “take it or leave it” attitude. ------ For example, I purchased a one-year plan and couldn’t find any option to request a refund, even though I only used it for two months. Then it told me, “just cancel and stop complaining.”
1 Like
what about code quality? Cursor had tools that allowed it to spit out less ■■■■■■ code than other competitors. how does kilocode perform in this domain for you?
You’re talking directly to the models, so no middleman reprompting rubbish. Code quality is far better than cursor. If your prompts are good, that is.
1 Like
It looks like this discussion has been un-pinned from the main topics page. @danperks started the thread 4 days ago, and then wrote a long response to many questions 3 days ago (thank you) with no posts here since then. That’s excusable since it was over the weekend. But I’m hoping that Dan or someone from the Cursor team will show up to answer more of the many questions that have been asked, including my own.
I guess the burning question for me right now is, what is meant by “API prices” in the context of Cursor – Models & Pricing? I understand what API cost is, and there used to be a table at Cursor – Models showing exactly what the API cost was for each kind of token in each model, with a note that there was a 20% surcharge on top of the raw API prices, which I found reasonable. But now all the pricing information has been scrubbed from both the docs and the dashboard, which now only shows the number of tokens used, with no way to get from that to the price paid for the prompt. So instead of seeing that a Max request to o4 cost $0.03 or $0.12 or whatever, it now says that it cost 65,526 tokens, with no way to see how much those tokens cost. Is there still a 20% surcharge, or has that been increased?
1 Like
I will definitely not use Cursor with such pricing and approach. It is simply impossible! I am not married to a Rockefeller daughter, but that is not the point. I could pay $500 a month if I knew for sure that the error rate for tasks would be 5% out of 100%, and not 80% out of 30%. If I was sure that I would set a task and Cursor would accurately complete it, without canceling, going in circles (stealing my money), and later completely forgetting the task, asking “I see your project, what task needs to be completed?” - Then it would be possible to work and understand where the money is spent and a good volume of tasks will be completed! But now, when you sit on a task for 4 hours and Cursor with models just fools you and in the end breaks everything, you have to roll back Git and start over, and then you get a bill for $50 for “NOTHING” - I am not a supporter of giving away your money just because Cursor and their investors really need it!
I bought an annual account, I am now testing Cursor’s competitors and when I decide on an option, of course I will demand my money back! And the fact that they simply transferred me to a new tariff without my knowledge is generally beyond stinginess, I have not seen such a thing for a long time, unless in some lousy office sitting in a basement somewhere in Iraq!
Then, when there were problems with Claude 4, they erased user posts, including my post from this forum. To be honest, even then there was an alarm bell!
And yesterday and today I encountered this problem:
After yesterday’s criticism of their prices, I started getting this error on slow requests, which I have never seen before, that is, the error is now constant, it is impossible to use o4mini in slow mode. At the same time, my friend on his account on the same communication channel works fine! Is this revenge? I’ll take a screenshot of this post just in case, because they just deleted this error from the neighboring forum thread 
Well, this is complete ■■■■ of course. I wonder if their investors see what’s going on? Is this a company or “Dad opened an account and a company, mom is a secretary, son is a developer, daughter is a marketer, and grandma makes office lunches”?
Request ID: 88b0fb76-dff5-4393-8357-a1363a8d7cdf
{“error”:“ERROR_OPENAI”,“details”:{“title”:“Unable to reach the model provider”,“detail”:“We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.”,“additionalInfo”:{},“buttons”:},“isExpected”:false}
ConnectError: [unavailable] Error
at x9a.$endAiConnectTransportReportError (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:4757:223764)
at hir.S (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:492:17741)
at hir.Q (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:492:17519)
at hir.M (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:492:16607)
at hir.L (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:492:15708)
at Bwt.value (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:492:14500)
at ve.B (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:48:2398)
at ve.fire (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:48:2617)
at Git.fire (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:4744:10379)
at u.onmessage (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:6968:12271)
P.S.
To be fair, I want to note that I am sure that the Cursor developers themselves are great guys and they are now freaking out about what is happening, because a tyrant “marketer” has appeared in the company who is doing all this nonsense. And probably this topic was lobbied by them to show the true tyranny of everything that is happening!
1 Like
will give it a try thanks for the suggestion, one last question, do you recommend using kilocode with cursor or in vs code?
Using it for more than a week, on Sonnet 4 it works pretty well. I don’t see any issues, and I don’t see that Cursor somehow outperforms Copilot. In addition, Claude code is $20 with limits, but still you can code 2-3 hours before hitting the limit that clearly resets every 5 hours.