@saidattax we do have a Pro+ plan. And if your Pro plan lasts 2 weeks that might work for you.
Yes, but, the competitive landscape is also dynamic.
Cursor’s behaviour isn’t unreasonable, but it isn’t looking like something that is building customer loyalty.
I don’t want “unlimited access for free”, but when billing time comes round, I’ll review how Cursor stacks up to the competition, and see what I move to.
Personally, I’ve changed my workflow to use “Jules for a first draft, and fix problems that the CI picks up, then Cursor to debug”.
That’s reduced my Cursor call usage a lot… and as things stand, I’m likely to renew, but my “role in the local dev community” has changed from “raving Cursor fan” to “someone who uses their software.”
3 Likes
Gosh. Every time we can stop thinking about pricing and focus on code you shift things around.
I’m pro cursor, but this is very annoying!
3 Likes
If your company went all-in based on the pricing structure of a new product in a rapidly-evolving field… well, that’s on you guys for making risky moves 
if you cancel and go to a competitor as you say, then A) have fun with that inferior product, and B) get ready to do this all again when that company also stops offering unlimited usage.
There’s no way any company will provide unlimited AI use for very long. They burn money and subsidize memberships early term, but eventually they need to be profitable. So long as LLMs are so expensive to run, and Cursor has to pay for their API usage, you better believe that cost is gonna get passed on to us users.
what is unclear to me is the auto price specific to teams usage or everyone ?
an an ultra user are you saying to me tha auto is no longer unlimited and now just part of my usage costs like any other model i can directly use ?
so what is the point of auto from my perspective now ?
seems like no point and i still dont know what model auto is using at any point in time
That’s why I’ve never committed to yearly, and won’t even though I could, in theory, lock in Auto unlimited for a year IF THEY DON’T CHANGE the annual. But, they already have changed annual and team users in various ways, so we know that’s a line they can easily cross.
This is why I’ve hedged with being on top of other tools and options and intermixing them with Cursor from day one. The key is to not put all your eggs in the Cursor basket.
That said, I do wish they were public because a short play on LEAPS would be very lucrative.
I do agree with the argument that Auto has improved, and that charging for it allows them to improve it more. However, there is also a need for unlimited that they seem to of quit caring about. They could do 2 autos, unlimited and metered. But, clearly, they chose not to.
Fortunately, open source models are improving dramatically. The day is coming when it will make more sense to invest in local rigs to run AI, which you can easily do for about $3k and share with a team. Combined with some low cost API and unlimitted GPT teams, and you can have a robust productive development option.
1 Like
The pricing itself is fine; even $100 a day would be acceptable. However, frequently changing the price in a short period is a problem. It’s rare to come across someone with such muddled logic as yours.
Of course, I understand that Cursor has to pay for API usage and that offering unlimited usage for free is unrealistic. But didn’t the Cursor team previously say that auto would remain unlimited? Maybe I’m misremembering and it was only for the Pro plan, but that’s exactly the part I’m questioning.
What I can say for sure is that having unlimited (even if slower) requests was one of the big attractions of this product. Before cutting that off entirely, I think it would be better to at least leave auto available without limits. Tab completion, honestly, is too weak to be useful—it only works decently when you already have a clean, well-commented codebase. In reality, most engineers are stuck working with large, messy legacy systems, and in those cases, it just doesn’t cut it.
If Cursor only wants to cater to shiny new product builders and early adopters, then maybe this direction makes sense. But if the goal is to expand the market and grow revenue, removing unlimited usage entirely will make Cursor weaker compared to alternatives. When VS Code fully integrates Copilot, for example, Cursor could lose ground. For a startup product to really succeed, it needs to appeal to the late majority as well. And realistically, most engineers are battling messy codebases, while many CEOs are skeptical of AI tools and slow to approve them.
In those environments, even a lower-performance model with unlimited requests is a huge selling point. A rate-limited approach like ClaudeCode would make sense—or even a slightly higher price point would be fine.
For what it’s worth, my company has already adjusted our internal rules so that we can keep using Cursor even under the new usage-based pricing. I just hope many other companies will be able to do the same.
I’m on the Ultra plan—last month it cost me $350, and at some point I had no choice but to switch to another IDE for a few days [and it was awful, nothing compares to Cursor].
My mistake was that Sam said in an email that it’s really unlimited [when will you get rid of that terrible bot?], so I used Opus, which burned through my package quickly.
This month I’m trying to use only the relatively cheaper models, but I still see that it won’t last me until the end of the month. I don’t know what the solution is.
It’s not related to you—getting a $200 bonus is nice anyway, but I’m still waiting for an unlimited plan with OpenAI.
P.S.
Just so you understand how little time it lasted for me—last month I actually reached $900 worth of usage in Cursor, because I got a few bonuses plus free usage of GPT-5 and a lot of usage with AUTO.
We have some urgent questions regarding our subscription plan and upcoming billing cycle. As our next monthly billing cycle begins on August 24, we would appreciate your prompt response to the following points so we can make timely decisions:
1. If we switch to the annual plan now, will the change take effect immediately?
2. After switching to the annual plan for our 300+ seats, will we be required to pay the full annual amount upfront, or is monthly/quarterly billing available?
3. If we switch to the annual plan before August 24 but the payment fails, can we immediately contact you to revert to the monthly plan?
4. Since a new billing model will be implemented on September 15, if we switch from the monthly plan to the annual plan between August 24 and September 15, can we retain the current pricing model of USD 40 for 500 requests for one year?
5. If we switch to the annual plan, can you provide a contract?
Given the urgency of this matter and the upcoming changes, we would greatly appreciate it if you could reply as soon as possible (ASAP).
May I inquire if, upon subscribing to the annual Pro plan on September 14, 2025, my Auto feature will remain perpetually free until September 14, 2026?
Indeed. That is what they said, and I intend to follow suit.
Thanks cursor for upgrade auto mode.
the direction is becommig everyone has nvida desktop blackwell devices running local models with the occaissional fallback to our openai api accounts.
its inevitable and will be facilitated by an open source clone of cursor, using desktop technology especially if the average man gets priced out of cursor …
the Ai endgame itself is a threat to all software
1 Like
All options are on the table. Fortunately, open source models are improving rapidly, opening new doors.
The challenge will be enlisting multiple agents into development, which Cursor basically does for us, though tools like Roo Code do that, too. I think, though, once we take ownership of that, we’ll create new things that blow away Cursor, and the concept of the IDE being the center of the universe. We already see this beginning in CI pipelines.
In the end, we’ll all be better off, more efficient and productive. Cursor doesn’t really have much of a place in this future. AI native IDEs are a good transition, but just not the long game.
Cursor also burnt too much trust capital to benefit from loyalty during the paradigm shift.
3 Likes
its clear the recent 900 million financing round had a lot of strings attached …
My team members are each paying $40 monthly, yet we’re only getting $20 in value. Since the billing rules have been revised, the team plan ought to reflect this with a $40 credit as well.
how to know exactly how much bonus usage we will get?
Small question.
I now see my fast requests (requests, there is no slow anymore) burn at 15 - 30 for a prompt. GPT-5 is supposed to be dirt cheap compared to claude and yet it burns, burns, through CLI tool.
Since I’m now billed at ‘API prices’ :
1 - what is even the point using this cursor thing and not just use API keys ?
2 - Claude code with opus 4.1 eats less than GPT-5, how is that possible ?
3 - Since openAI released its CLI with a price to use their models, using cursor previous plan, what gives me the incentive to stay ? They give unlimited model use over time.
All in all even copilot has a hard cap on price and gives unlimited access.
Cursor team, what on earth makes you billing like this, while OpenAI kept your previous pricing strategy, anthropic did the same, even MICROSOFT.
Give one good reason to use your product instead of Codium with all of the above ?
You probably don’t need to use -fast.
You’re paying more to be jumped to the front of a queue that is likely not very long anyway.
I stopped using -fast after the trial finished. Since then I might wait a few seconds for a response to start, but the whole response takes a minute anyway so I don’t consider an extra few seconds anything to worry about.
See here
My main use for Cursor is access to multiple models and to be able to chop and change the model for each prompt. With a complex task I start with gpt5-high or similar heavy lifter. For subsequent smaller edits or tweaks I use gpt5-mini or -nano. That’s a real money saver.
1 Like