Pricing Megathread and Q&A

They can ■■■■ a ■■■■. Fucken scam artists

2 Likes

Prices seem to have doubled. But who knows, we are charged by the token now. The Agent might decide to completely re-write your code, decide to fix errors that were fixed an hour ago, create shell scripts you don’t want, forget how to apply edits and loop 5 times re-apply failing until it says “please do it manually”.

It’s turning into a crapshoot

1 Like

seems the AI support bot ran out of tokens too

9 Likes

probably the people at Cursor think they are dealing with housewives and not with programmers who know math well )

6 Likes

there is a strong suspicion that the cursor agent assigns a hidden prompt to Claude each time he completes a task, like "Make a diagram at the end, create a test script, “talk as much as possible at the end, tell your whole life story”

1 Like

At this point, I have no trust left. The system isn’t open source, and there’s no way to audit how agent instructions work. So how are we supposed to know it’s not being told to start breaking the code after a few prompts just to burn more tokens?

2 Likes


look at what these idiots are doing when people try to voice their opinions - https://x.com/mrpixl3/status/1944525618730971300

I’ve been using a Pro subscription since September 2024, mostly working with the Claude agents. Each month, I had about 50-100 fast requests left. But now, after two weeks of use, I got a notification that I reached my limit and have “spent” over $95. The funniest part is that the agent often rewrites the entire code and comments, just copying the existing one and recalculating it into tokens. I think it’s time to look for interesting alternatives, like Claude Code or something else. :thinking:

6 Likes

Robot AI did it, I didn’t like to publish it later! unsubscribe

While not immediate, [email protected] will eventually respond and they can reverse this. They put me back on the original 500 req plan, I presume all yearly subs will have the same option.

1 Like

The original 500 requests were very clear! They could be used up almost perfectly within a month, and even after using them up, there were unlimited slow requests, and the slow speed wasn’t even that slow. But now, I’ve only been using it for half a month, and it tells me I’ve used $111! It’s forcing me to switch to auto mode. When I ask what model auto mode uses, the LLM responds that it’s GPT 4.0.

Moreover, the quality of auto mode responses is really quite poor. I can accept unlimited slow requests, and I can also accept paying for additional quota, but why was there no notification to users when switching to the new billing mechanism? It’s both non-transparent and expensive!

3 Likes

Cursor thinks API-based pricing is the best approach. However, it should at least let me know how much quota I have left. I can accept higher prices, but not ambiguous behaviors. It’s so frustrating that I can only rely on the low-quality auto mode for the remaining time.

1 Like

I don’t object to price increases, but only on the condition that there’s a good user experience.

Starting from version 0.48, I had already guessed that Cursor’s pricing strategy would be adjusted, but I didn’t expect the final result to be so vague and abstract.

The key issue is that LLM responses are now becoming increasingly poor. In Auto mode, you have no idea which LLM is being used, and the results are often very unsatisfactory. In Claude 4 mode, the Rules settings are simply a joke - they’re not followed at all, and the responses are becoming as bad as the pricing. I’ve even discovered that it arbitrarily switches to Claude 3.7 or even Claude 3.5 models for responses, which is completely fraudulent.

Just 4 days after this renewal, I received a message saying “You are projected to reach your usage limits by 7/18/2025 based on your current usage. Consider switching to Auto for more requests, asking for smaller changes, or enabling pay-as-you-go to avoid interruptions until your cycle resets on 8/10/2025.”

Poor responses + vague pricing + the massive token consumption caused by poor responses make this unacceptable to me, and I’m considering canceling my subscription.

Let me know if you find an acceptable alternative to Cursor and the Agent mode which is awesome when Sonnet 4.0 (Or Geminie 2.5 pro ) is available

2 Likes

Cancelled my subscription as well.
After contemplating over what’s been going on with cursor policy and community response to it’s actions it is clear that most are sharing the same frustration.
Personally I barely could spend 500 requests per month, in average it was about 300 requests if I recall correctly, may be even less than that. But when I was switched to new token based pricing model, I burned through my plan in 5 days! Well, indeed they prompt you to switch to $60 plan which is basically the same as previous $20 plan, but that’s not how market raises price for the same product, not in 1 month at least.
Will give competitors a shot, pity that poor decision making force people to have a look for alternatives.

4 Likes

I would have thought that with Cursor’s reputation currently on fire, they would be putting “please stop talking” into their prompts.

It seems the subsidised days are over but many many people will just spend $100 on CC instead of Cursor purely out of spite.

The first post in this thread is by danperks Community Developer

[this is] a place for our team to help clarify and answer your questions.

That was 4 days and 150 replies ago.

Not one of those replies is from team Cursor.

That’s appalling. One of the secrets of customer relations is customers become more loyal if they have problems with your service and you solve them. “Oh, those guys do care, I’m going to keep giving them a chance.”

Here’s what they should be saying

“OK users, things are getting more expensive and we can’t control that. Here’s some techniques and rules on how to minimize your costs”

They seem to think it will all just blow over “we’ll get some churn but the more people leave, the more money we save, the VCs will keep the lights on”.

When I paid for a year, I thought my risk was a better service coming along and me being stuck with an inferior product but I was “well, I get value and that will be the same regardless”. Well you’ve made me feel a fool and I won’t forget that come re-subscription day. They have a chance for redemption but loyalty is for dogs.

3 Likes

I’m unfortunately in that boat too, not even reaching 400 requests but unwilling to have someone rate limiting on a productive day, they’re making bad decisions one after another, I still hope they turn back and enable the old plan, there’s also an explanation for this change given directly by Cursor CTO: https://youtu.be/4jDQi9P9UIw?t=519
“Are you finding is that your biggest challenges, working with third party inference, wether it’s anthropic, or openai, gpt, versus having full control of your own microservices?” CTO: “oh my god”

3 Likes

I will also cancell my subscription as well. I have no trust left, one more good company that become a scam. Just got Windsurf, and seems promising, plus you deal with more transparent people. Instead of: “Sorry for our new pricing plan and our lack of transparency, but that’s it, now deal with that”.

Hey @T1000, no one’s complaining, am I right? :laughing:?

Cursor are hoping for a buy out. Their product is basically dead in the water, Anysphere is a sinking ship at this point, this happens when people that make business decisions don’t know how to actually run a business. I feel sorry for the guys from Anthropic that decided to join the company at this point. :laughing:

3 Likes