I’ve been experiencing issues with Cursor in the past few days — all the requests I send cannot be processed properly. From the Cursor official website console, I can see that the server is receiving the requests normally and tokens are being consumed as expected. Please refer to the video in the attachment for specific symptoms. The version of Cursor I’m currently using is the latest version, 3.7.21. Thank you very much!
Steps to Reproduce
I’m not sure how to reproduce it. I first encountered this issue a few days ago, about three or four days ago. After trying multiple times with no success, I attempted many solutions, including restarting my system, downloading the latest version of Cursor from the official website, logging back into my account, creating a new agent to handle the problem, and asking other friends who use Cursor if they could solve it. In the end, none of these worked.
Expected Behavior
The agent is able to process my requests normally.
This is the token consumption status shown on the console. Looking at the token usage, the server did indeed correctly receive and process my requests, but the IDE side simply terminated.
Hey, thanks for the detailed report, the video, and the console screenshot. From the symptoms, this looks like a response streaming issue. Our backend is receiving and processing the requests (that’s why you see token usage in the console, the model is actually generating output), but the long-lived response stream isn’t reaching the IDE. That’s why it gets stuck on “Planning next moves” and then loops on “Reconnecting…” with no error. This isn’t wasted charging and it’s not a billing bug, the response just isn’t getting back to the client. Most often this is caused by an unstable network path or a route far from the US, including VPN or proxy.
Let’s narrow it down:
Open Cursor Settings > Network > Run Diagnostics and send the result. It often shows something like “Streaming responses are being buffered by a proxy in your network environment”.
In Settings > Network, enable HTTP Compatibility Mode (the “Disable HTTP/2” option or switching to HTTP/1.1), fully restart Cursor, and try again.
If you’re using a VPN or proxy, test both with it and without it. An unstable route to our servers is a common cause in this region.
To be honest, for regions far from the entry point, there isn’t always a complete client-side fix. The root cause is infrastructure on our side. We’re tracking the issue, but I can’t share an ETA yet. The steps above do fix the freezes for some users, so please try them and tell me what the diagnostics show.
