Reconnecting / stream stall on one machine only (same account works on other machine)

[ImranAhmed]

Where does the bug appear (feature/product)?

Cursor IDE (It worked normally on this machine all day, then started reconnecting with no change on my side (no reboot, no settings change, same network). That suggests a backend or session-state issue rather than a local/network fault.)

Describe the Bug

Subject: Reconnecting / stream stall on one machine only (same account works on other machine)
Description:
Cursor gets stuck on “planning next move” and then “Reconnecting” on one Windows PC. The same account works normally on another machine. Internet and other apps are fine on the affected PC.

Steps to Reproduce

What I’ve already tried:
Reboot
Renamed %APPDATA%\Cursor and restarted (fresh profile)
No VPN; wired connection; no firewall changes
Affected machine has no WiFi (could not test hotspot)
Console errors when it happens (Help → Toggle Developer Tools → Console):
Connect error in unary AI connect ConnectError: [invalid_argument] Error (during getTeamRepos / pollRepoBlocklist)
[transport] [NAL client stall detector] Bidirectional stream stall detected - no activity for threshold period
[transport] Agent client stream abort Error: stall_detector
Connect error in stream AI connect ConnectError: [internal] Cancellation token requested for stream
[AGENT_ERROR_DIAGNOSTICS] requestId=d111e7b3-37a0-4a2b-a936-432bfc383bd2 decision=RETRY (countAsTransportError=true)
[composer] Extension host became UNRESPONSIVE

Operating System

Windows 10/11

Version Information

Environment:
Cursor version: 2.4.28 (from console)
OS: Windows (x64), stable channel
Same Cursor account works on a different machine

Does this stop you from using Cursor

Yes - Cursor is unusable

[deanrie]

Hey, thanks for the detailed report. This looks similar to the situation in this thread where one PC didn’t work, and another one on the same network worked fine.

The errors (stall_detector, ECONNRESET, EAI_AGAIN) suggest that TCP connections to the servers are being actively dropped at the network level. This could be your ISP, router, or security software.

Let’s isolate the issue:

1. Test with a mobile hotspot (critical)

• Share internet from your phone (mobile data, not WiFi from your router)
• Connect your PC to the hotspot and try Cursor
• If it works, it’s 100% your network or ISP

2. Try disabling HTTP/2:

• CTRL + , → search for “HTTP/2” → enable “Disable HTTP/2”
• Restart Cursor

3. Network Diagnostics:

• Cursor Settings → Network → Run Diagnostics
• Send a screenshot of the results

Questions:

• Who is your ISP?
• Do you have any antivirus or firewall (Kaspersky, ESET, Windows Defender with extra rules)?
• In that thread, changing the router port helped, sounds weird but it worked

Let me know the hotspot result, that’s the main thing to figure out where to dig next.

[ImranAhmed]

The service came back after a few hours, no changes to local machine. I use Norton antivirus and ISP is Community Fibre. This is the first time I saw this after using extensively for a couple of years, annoying, as I lost several hours.