Yeah, I agree. I have the CC pro, and it’s being very cool and enough for me. Reset the token limit every 5 hours, so you can really use it all day. And I’m using it with Kiro, what’s being amazing for planing and so on…
Windsurd we have it at work, and it’s also nice and easy. Like the UI a lot. So I’m missing to try Trae. But one thing is for sure, one of these (CC + Kiro, Trae, Windusrf, even copilot) will replace my cursor sub.
Ok, me with my stupid bragging about CC and perfect 10…I should have known that would invoke the mother of all jinxes. So here we go for the other side:
Does CC make mistakes, even when you’re under the impression you’ve build a solid guardrails package. Yes. You’ll have to be meticulous, patient and, above all, test the ■■■■ out of it or Claude WILL try to ignore all of it and go full Bruce-Lee. I think by now it’s a well known side-effect of Claude 4 that it’s persona is quite “Lets go go go” and this you will have to neutralize or you’ll end up with sessions that end in "I ■■■■■■ up, ignored all your rules, sorry to have wasted your time.
So, recently Antropic introduced these sub-agents and I’m now discussing with Claude how to utilize those as “ah ah ah!” triggers to prevent it from going completely off the rails and remind it of writings tests, keep docs and folders and naming conventions, don’t convolute core-files, take max code lines into account, work modular and do regular, useful git commits.
Will that work? No idea but will update when done.
I switched back to request based pricing a month ago, not sure if you still can do that. If you can though, just do that, it’s way better 
Hello, @Chessy,
I’ve read your posts, and I’m very curious about one thing.
Could you please tell me what you think is the best alternative to Cursor?
Before using Cursor, I used to work with Windsurf in VS Code.
After switching to Cursor, I found it to be a smart and impressive development editor.
However, the recent updates have been disappointing, and now I’m looking for an alternative.
I’d really appreciate hearing your thoughts.
Thanks!
That’s incredibly difficult to answer. I’ll try to break it down a bit:
-What do you need it for?
-How much of it do you need?
-How important is money to you?
Truth is, right now there are already quite a few alternatives to Cursor, but is there a clear winner? No. (and please never believe any of these YouTube “testers” who run 2 or 3 things and then claim “this is my new go-to IDE! -And now pls subscribe to my -OMG Vibe Coding is 1mln times better with my program-”)
Some tools will also connect much better with you, and some will put you off.
A tool can perform quite well, but if you simply dislike its interface,
that can be quite a hurdle.
The best advice I can give you is to test some and see how they make you feel.
For example, Claude Code can be seamlessly integrated into Cursor. That should give you at least some basic framework to judge whether you like its performance or not.
Last but not least, please do note that this entire landscape is continuously on the move. What works for you one week can fail miserably the next. And these cheeky updates Cursor has been pulling lately are unfortunately not unique to Cursor. For example, Claude-Code is probably going to get monthly limits soon.
That’s still a rumour, but given recent developments on IDE devs now having hooked their share of the pond, it wouldn’t surprise me one bit.
Thanks for your answer.
I prefer VSCode similar tools and also, can pay about $20 per month like Github Copilot or Windsurf.
Anyway, I like previous cursor’s codebase indexing and analyze functionality, embedded chatting and so on.
Thanks again.
As someone who moved away because of an unbearable bug, but come back occasionally to check if it’s fixed so I can reconsider subscribing, your sarcasm is such an testimony of my decision to leave.
Everyone here is getting increasingly fumed and frustrated as far as I can tell and I don’t really see such reply trying to steer into distracting users from the discussed problem anywhere remotely close to proper moderating. I don’t mean be harsh but remember you are not just a reddit mod or a github repo maintainer who are volunteering. People are paying for a service and expects what’s promised.
I personally moved to Roo (agent) + Continue (autocomplete & chat) Extensions with self-host models for easy tasks and OpenRouter if I need corps models.
The only Cursor does better is the autocomplete (which is now broken in jupyternotebook). Roo has quite good agent functionalities but tends to burn through tokens quickly if you don’t modify the agent’s context compression provider.
Just wanted to add that if you install Gemini CLI and you activate it via API key, it comes with quite a comfortable free tier. Running out? Switch API key xD.
Basically free Gemini Pro 2.5 for as much as you want to use it. Which, given its current state probably won’t be THAT much but hey, you can still use it as a handy assistant to save tokens for your main.
Agree with you on this. The current pricing model for Cursor is just not sustainable for many of us.
I’m on the Pro annual plan, and it’s become way too expensive. I used to get by just fine with the 500 requests a month, but now, I burn through them in about 10 days. That leaves me stuck with the “auto” mode for the rest of the month, which, as you pointed out, is incredibly frustrating to use.
I feel your pain with the auto mode’s reliability. I was trying to do something as simple as adding a dark theme to a web app I’m building, and it completely botched the job. No matter how many times I tweaked my prompts, it either did nothing or claimed the task was done without making any real changes. I ended up having to write the code myself, which defeats the whole purpose of using an AI assistant.
It’s a real shame because I genuinely liked Cursor, but this new direction isn’t working for me. I’m actively looking for alternatives at this point. Since I’m on the annual plan, I’ve been wondering the same thing about a refund.
You’re right, it’s not just about explaining the costs; it’s about making the tool feel fair and usable for the developers who rely on it. Hopefully, the Cursor team takes this feedback to heart.