Connection error for specific applications

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

“Connection failed. Please try again, or contact support team if the issue persists.”

After submitting new message the model gets stuck at “Reconnecting” and after a couple of minutes get the same issue notification. Would believe there’s a connection issue on either parties UNLESS given the Cursor IDE works well for other projects flawlessly. Restarting the computer usually makes ONE more message possible to send until’ the error message arrives again.

Steps to Reproduce

Start new chat, send message, receive error.

Expected Behavior

Cursor IDE supposed to be working per expectations.

Screenshots / Screen Recordings

Operating System

MacOS

Version Information

Version: 3.10.11 (Universal)
VS Code Extension API: 1.125.0
Commit: 4ef9fe3d055f8c4523179a090f14eb835bc3c940
Date: 2026-07-03T04:44:14.398Z
Layout: IDE
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.256
OS: Darwin x64 22.6.0

For AI issues: which model did you use?

Composer 2.5, Auto

For AI issues: add Request ID with privacy disabled

Request ID: e7b7908c-ac5e-42e2-8500-ea7c3f5ac87e
Canceled
RetriableError: Canceled
at Rqi (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9056:8640)
at Vap (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9057:11962)
at Object.classify (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9057:16467)
at Duf (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9057:5856)
at async eWo.runWithSharedTurnRunner (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9057:15819)
at async EV.runAgentLoop (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:20449:3041)
at async Uva.streamFromAgentBackend (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:20477:14168)
at async Uva.getAgentStreamResponse (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:20477:23591)
at async UIe.submitChatMaybeAbortCurrent (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9219:2399)
at async RT (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:19508:83887)

Does this stop you from using Cursor

Yes - Cursor is unusable

Hey, thanks for the detailed report with logs and the Request ID. From what we can see, that failed request e7b7908c-... never reached our servers, and most of your requests on the same days went through fine. So this doesn’t look like a server issue or an account problem. It looks more like an intermittent drop on the client or network side. The fact that only one project fails, and a reboot gives you exactly one message, also points to local connection state issues like HTTP/2 stream reuse, VPN or proxy behavior, or a workspace-scoped stall.

Here’s what to try, in order:

  1. Update Cursor from 3.10.11 to the latest version. We’ve shipped a few fixes since then related to connection issues and empty workspace behavior.
  2. Turn on Disable HTTP/2: App Settings Cmd+, then search for HTTP/2 and enable Disable HTTP/2. This forces HTTP/1.1 and can help if your HTTP/2 transport is unstable.
  3. If you’re using a VPN or proxy, turn it off and test again. Also check for any corporate firewall or antivirus doing HTTPS or SSL inspection.
  4. Run Cursor Settings > Network > Run Diagnostics.
  5. If the problematic project opens as NO FOLDER OPENED, open it as a folder instead.

Let me know if this helps, and if it does, which step fixed it.

Thanks for running through all the steps and sharing the new Request IDs. Both (9c780662-... and 5e29dca3-...) have the same RetriableError: Canceled signature as the first one. This looks like a client or transport drop. The request gets cut off somewhere between your machine and our servers, so it never actually reaches the backend. Since this is now happening across multiple projects on the same machine and network, it points less to a specific project and more to something at the network or system level.

Let’s narrow down where it’s breaking. The most useful test is:

  1. Connect to a totally different network. Use a mobile hotspot from your phone and try in Cursor. If the issue goes away on hotspot, it’s your router, ISP, or DNS, not Cursor. This is the key step.
  2. Check system-level software that can drop connections even without a VPN. Firewalls or network filters like Little Snitch or LuLu, employer MDM profiles, DNS filtering like NextDNS, Pi-hole, or a custom DNS. Temporarily disable and test.
  3. Is there a timing pattern? For example, does it fail after being idle, after N messages, or with larger contexts. A pattern like reboot then exactly one message then the error again often means a stuck connection state like keep-alive or a socket that some network device between you and us is resetting.

Let me know what happens on the hotspot test. That will rule out half the options right away.