I have the same issue. I tried all solutions proposed but nothing works for me
Version: 1.6.26 (user setup)
VSCode Version: 1.99.3
Commit: 6af2d906e8ca91654dd7c4224a73ef17900ad730
Date: 2025-09-16T17:12:31.697Z
Electron: 34.5.8
Chromium: 132.0.6834.210
Node.js: 20.19.1
V8: 13.2.152.41-electron.0
OS: Windows_NT x64 10.0.22631
Same issue today. Tried removing all extensions, made sure I was up to date. restarted app and computer, used different models, nothing. Then as mentioned I went to my chat history and deleted the only one that I had on the project, which wasn’t that big, but that did the trick.
I had this a bunch today. A few times, I was able to “clear all chats”, but now it is just stuck and not working at all.
Same issue. Completely down. No chats work.
Seems to be an unresolved incident: Service Degradation. https://status.cursor.com/
I don’t think it’s working overall. Let’s all calm down and wait!
I mean, it’s not right that this and “we couldn’t contact the provider” are so frequent.
We pay to use top of the line models and end up using up our quota using less powerful models because every month there’s two weeks of mostly-down-time where the best models don’t work. This thread is from MARCH.
If a model actually fails we get refunded, but if we’re pushed to use a lesser alternative, we don’t. “Let’s just wait until they charge us a few more times and they’ll fix it at some point” isn’t a great solution.
Edit: Just hit my usage limit bashing my head against the same wall for a few hours with claude-4-sonnet because GPT-5 wasn’t available. I’m gonna cancel my subscription.
Pro Plan Summary
Sep 14, 2025 - Oct 14, 2025
| Item | Quantity | Cost |
|---|---|---|
| Included in Pro | ||
| claude-4-sonnet | 42.7M tokens | $20.11Included |
| claude-4-sonnet-thinking | 21M tokens | $13.70Included |
| gpt-5 | 47.1M tokens | $11.65Included |
| grok-code-fast-1 | 318.4K tokens | $0.03Included |
| Total | 111.2M | $45.49Included |
Everything I used other than gpt-5 was useless (I would not have used them at all if the very useful model I wanted to use was online) and the plan was $20. You charged me $20 to send $11.65 worth of mostly-useful inference to someone else and waste my time. I’m gone.
TLDR: This can be fixed by removing app data
Please follow this doc step by step to resolve the issue.
So are you deliberately forcing everyone to create a new chat?
This is ridiculous. I create a new conversation when this happens, and we start from scratch again. Then it happens again about 5 minutes later. It’s impossible to get any work done!
My resolution was removing the MS Powershell extension and a MCP that had a red icon. Restarted and so far after 20 mins no connection failures where as it was failing every prompt.
(edit: problem came back. The project was a zig lang test and the I trapped an issue with a STDOUT print and remembered the potential of such output messing with MCP tools calls. I changed print usage to use STDERR and shrank a buffer for when MCP tool calls can loop calling tools that return “large” values. I don’t know what triggers “large” return issues but stating what I did for completeness.)
Need a FAQ with what helped people with this as it is not intuitive without knowing what can cause the connection failures when not obvious per error help message.
good idea, solved my issue.
Cutting out ms powershell? Ah so it’s a windows thing. POSH (aka powershell) on windows is known to be an issue and requires an ambiguous amount to fix even before cursor days. We only noticed this recently at www.dev3lop.com - and it was not an issue in the beginning. I will admit I’m asking cursor to run really mundane tests and that would mean the posh tool is trying to log an enourmous amount. I notice it’s not filling up context, but rather this ai agent console. Pretty much my team is just saying ‘use your mac’ …
So my question; who above is using a windows machine.
It’s starting again…
