I have reported 36 bugs over the last year, pouring hundreds of hours into detailed bug reports and help for the Cursor team, to help fix and improve their product.
In almost every case, NOTHING EVER GETS FIXED - basically every problem I’ve reported, still exists right now.
Steps to Reproduce
Report a bug. Watch it never get fixed.
Expected Behavior
Someone should fix something here!
Operating System
Windows 10/11
Version Information
All versions, all platforms, all time.
Additional Information
This is a desperate post of of extreme frustration, and a warning to the hundreds of other victims who post in this forum expecting that their problems might get solved, when it seem in reality, there is no connection between here and anyone in the Cursor dev team.
I AM THE MESSENGER
Do not attack me or fight or make excuses or otherwise exact revenge because I have posted an honest, year-long, truthful account of what has actually happened, to me personally, over the last year.
An excellent response that I dream about, would be for someone to fix a few bugs, and then thank me for taking the time to report them in the first place!
hi @Chris_Drake and thank you for your feedback. I appreciate that you report bugs in the Forum though it’s not accurate that we do not fix them.
Over last months my colleagues and I have on behalf of Cursor users in the Forum reported issues to our engineers but also updated users on progress of bug fixes.
As @liquefy mentions we do fix bugs and have shipped bug fixes in every single version of Cursor, as well as in between major releases if not daily then every few days.
Please keep reporting bugs so we can reproduce and fix them.
Hi @condor - Your title - “Community Developer” - what’s that? Is there some way we can help you guys and submit bugfixes? I’m so desperate for many of these things, that I’ve taken to fixing them myself locally, which is tricky since you ship updates so fast and I have to keep re-doing it!
With respect (genuine), I think your push-back here is inappropriate. The crazy number of things I’ve reported, which are still broken today, and the crazy number of other things I see other people report, which are still broken, so vastly outweigh the (almost-certain co-incidence!) that some things people report do happen to start working in future releases, that I think it’s borderline disrespectful to disagree with my statement and attempt to defend Cursor devs. By sheer probability, cursor fixing or changing things, with no visibility to this forum, will from time-to-time surface as what looks like a bug got fixed, but that’s not the same thing as them fixing a bug that got reported!!
This may be true, however, but “shipping bug fixes” is not the same as “fixing bugs reported in this forum”.
Now go and look in your system for any tickets that match any of those topics: I promise you they will not be there in anyplace where any of us can expect for those things to be fixed!
I AM THE MESSENGER
Please do not flame me for trying to help.
FYI - I’ve joined the Cursor Community program, and I teach Cursor to meetup groups monthly, and I am, by far, my areas most vocal cursor-evangelist. I’m giving you guys a hard time, because your infrastructure sitting behind this duct-tape and band-aids front-end is actually crazy-special, and still worth using, warts and all!! (hoping to cure those warts!!!)
hi @Chris_Drake, thank you for the detailed reply and for all the effort you put into reporting issues and advocating for Cursor in your community. Your feedback and bug reports are genuinely valuable to us.
I understand that having issues and missing links between Forum reports and visible fixes is very frustrating. The more specific and clearly reproducible reports we get, the better we can prioritize and route them internally. If you have workarounds or have identified likely root causes in the latest Cursor IDE/CLI versions, sharing those details in the relevant threads will help both us and other users.
Some issues are already known on our side but do not yet have a fix, or they require broader architectural changes. Many bugs do get fixed, but we cannot realistically fix all of them, especially if they only appear in rare edge cases or have simple interim workarounds. That is why detailed reports, logs, steps to reproduce, and confirmations from multiple users are all very helpful for prioritization.
We are not able to respond to every single post, but we do review Forum reports and attempt to reproduce them. When you see follow‑up questions from our team, replies there are often what allow us to track down the root cause and get a fix into a release.
You are right that we can do better at closing the loop here in the Forum once something is fixed. We are actively working on improving communication around bug status, but with a small team we cannot yet provide full 1:1 support and status updates on every report. We are hiring to improve that, and feedback like yours is part of why we are pushing in that direction.
On the tone side: calling the Forum “fake” can easily overshadow the substance of your points and may discourage some people who are trying to help triage, reproduce, and route issues every day. I am not offended or defensive here, but I do want to keep discussion aligned with our Forum Guidelines so that it stays constructive, focused on problems and solutions, and welcoming for everyone participating.
I’d really appreciate your continued patience and your replies in your existing issue threads so we can tighten the signal between your reports and our internal tickets. Your support for the Cursor community, teaching at meetups, and pushing us to do better are all appreciated. Also, please encourage others to post detailed issues here as well, the clearer the overall picture, the better we can address the real pain points.
In my role as Community Developer I primarily bridge the communication between Forum & social media users with our engineers.
Please can you prioritize the MCP connection handling bugs - it’s driving me and loads of other people absolutely crazy, and I’ve posted at least a dozen replies now explaining exactly what you’ve done wrong, and how to fix it: you’ve got something calling async code in a loop and only using the last reply from that, leading to exponentially more idle but otherwise live, unused persistent SSE connections and eventually leading to total crash.
Do you have visibility into the ticket system? Can you find the sse mcp server connection bug-report, and tag it as priority #1 please!
Fixing your hooks so they work the way you documented is a runner up in the “please get this sorted” department!! (but not at the expense of fixing mcp connections).
The last should-be-very-high priority mistake you need to deal with is the mad idea of using a folder timestamp to generate a hash for indexing all our chats/agents. Timestamps change for many reasons, and this is the cause of every complaint you keep getting from irate users who’ve had all their chat histories vanish. While super-simple to fix, it will be a big challenge for you to somehow deal with the backwards compatibility issue this brings up. A good workaround for now, would be to store the folder hash inside a file in a local .cursor folder inside user projects - so anytime the hash changes, your code can find and learn the old one, and take appropriate steps to keep the history. If anyone is reading this, and wants to do this immediately, reply here: I’ve written a tool you can run, now, to record this info - but it’s based on a microsecond hash, so you MUST run it BEFORE you lose your chats, to backup that timestamp, otherwise your chats are effectively gone-for-good. Remember - a good chat is not just a chat (or agent) - it’s like a veteran employee who’s been trained up to know everything and expert in their niche. Losing a chat is like them getting run over by a bus!
BUT - you really need to sort out those MCP bugs above all else - same-named tools on different systems can be randomly invoked by agents [which one runs is a coin toss, depending on which async thread ended last], leading to potentially catastrophic consequences (like dev-tool commands running in prod-server environments)
Best of luck with the hiring! Do you take off-side applicants? Or do you take short-term contractors? Do you have a link for us to apply?
@condor@deanrie Guys - can we please have an update on your progress; none of the bugs I’ve reported should take more than a day to fix, and you are telling me “… it’s not accurate that we do not fix them.” but we are still not seeing any fixes.
Also, LetsEncrypt have rotated to their R12 intermediate cert today, so EVERY mcp server that renews certificates this month is going to stop working in Cursor again… until you also fix your broken TLS root cert bugs as well…