Cursor used to be my favorite AI IDE, but the recent pricing tactics have completely changed that.
Now, Cursor is simply not affordable for developers outside the U.S. or those not earning in dollars. What was once an inclusive, powerful tool is now behind a paywall that many can’t cross.
They forcefully migrated my Pro plan (which included 500 requests per month) to a new pricing model without consent.
They restricted access to high-quality models, pushing users toward low-quality ones that waste both time and effort.
When trying to switch to usage-based billing after hitting the rate limit, the charges feel unreasonably high, as if they’re treating us like gold miners.
Cursor has many bugs, especially in how it calculates tokens — I’ve noticed incorrect usage reporting and tokens being burned far more than expected. It feels like they’re draining tokens at will.
It’s the fourth month for me to use cursor, and this month in just ten days, token goes out. I can’t use it any more. auto mode is stupid useless.
so I canceled my subscription. searching for other solutions.
I also found this problem, and I can see some obvious abnormal token consumption on Token billing.
Some simple changes would cost millions or even tens of millions of tokens, which is something I can’t understand.
This is true in mode without Max turned on, so it’s hard to imagine if Max is turned on.
You are right. They force users to use AUTO, but AUTO mode is rubbish. Every now and then it just deletes a file for no reason, for no reason, completely changes functionality without my interaction, doesn’t remember what we were talking about a few minutes ago… it’s just completely unusable most of the time. And I know Cursor defends itself by saying that the prices are actually official API prices from AI companies like Anthropic and OpenAI, but I don’t believe they don’t have some kind of contract for cheaper API calls when they run a commercial service.
With the $50 plan, I’ve reached the limit in 2 weeks, I can’t do anything and especially NO SLOW MODE.
So that’s it, I’m cancelling the subscription. I’ve been working on a big project for several months and now I can’t move forward because of all these changes and the AI limitation on cursor.
I’ve had enough of all this scheming.
If you go with Windsurf, use Claude Code for Sonnet 4 access, and use Windsurf for o3 and o3 (high reasoning). Depending on your budget, I would also go with Gemini CLI for large content analysis and reasoning. Both o3 and o3 high reasoning currently do exceptionally well when it comes to troubleshooting, and Sonnet 4 excels at implementing fixes (for $20 you get credits that reset every 5 hours).
I switched to this setup and am currently paying $15 for Windsurf, $20 for Claude, and $15 for Gemini (I have a custom Gmail; if you get it through Workspace, it’s about $15/month on their plan, but if you buy it directly, it’s $20). I also tried Cursor for ~$60 and regretted that decision. Right now, I can’t even do anything—even though I have credits, it won’t edit, gets stuck, or takes hours to do something. I need to cancel the edit and resubmit, there are numerous tool call timeouts, and the performance on all agents (especially o3 and Sonnet 4) is way worse than the actual tools with CLI.
I strongly recommend this setup. I do development for ~12-18 hours a day and have barely run out of credits with these three tools, whereas with Cursor Pro+ I ran out of functional usage within a week and it became impossible to work with after they changed their plan.
I cancelled my ultra plan last night. I dont have time for this ridiculous pricing BS. Bad customer service (my emails were never responded to), ridiculous pricing (and constant changes), increasingly buggy cursor experience.
I find it very interesting that since I have cancelled my subscription my token usage goes over 50k a lot less than before… Almost like its done on purpose
If you are using Claude Code, it will reset every 5 hours with any subscription you have. Your credits are shared whether you use it in the browser or CLI. I am using it for coding, so I use it in my terminal, and it has nice caching that allows me to code and do builds for a longer time compared to doing it in the browser. In the browser, you will basically copy and paste and then open a new chat when you reach the chat limit. I hope this answered your question. I can give you more steps if you want for setup and usage. But you also have the option for API calls, which I don’t recommend as it is a bit expensive.
In case I confused you, to be clear: Claude Code is basically the same as using Claude in your browser, but you can use it in your terminal. For example, I can tell it to read my content, edit files, and so on. The beauty of it is that I can point to the rules I created for Cursor/Windsurf and tell it to follow and use them. It’s basically the same way you use an agent in Cursor or similar tools, but with a bit more manual work and the need to use the terminal for it.
Performance-wise, I found it is significantly better than anyone else who provides Sonnet 4 as an agent. It has its own pattern of thinking and a to-do list (although I like how Windsurf does their planning method more). The only downside is that on the $20 monthly plan, you can’t use Opus in the terminal as it’s a bit expensive, and on the browser it doesn’t last long either. To be honest, I never needed Opus given how well it solves problems right now.
But if you need Opus and have the budget, you can go for their $100 plan (Max 5x Pro) or their $200 plan (Max 20x Pro), as they give 5x and 20x the amount of credits respectively and include access to Opus on Claude Code.
