i have add required domain to my company’s whitelist.
but it aways show connection failed,please try again when using agent .
i am a pro user,but cursor ide show the free plan.
is there any other domains i should add to network whitelist ? thanks.
this occured on my company’s win10 pc,but my home’s win11 pc and mac work fine.
agent log:
Request ID: b80592f3-aac1-49b4-aafe-ba2660569580
MainThreadCursor disposed while stream f8bbf6ee-075e-4127-9493-e159540ecdcb was active
ControlledExecDisposedError: MainThreadCursor disposed while stream f8bbf6ee-075e-4127-9493-e159540ecdcb was active
at lub.dispose (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:38300:49595)
at ibl.disposeCustomers (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:41101:9954)
at ibl.dispose (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:41101:9804)
at u0C.dispose (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:45757:14316)
at c0C.stopAllInReverse (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:45757:13450)
at async UHd._doStopExtensionHosts (vscode-file://vscode-app/c:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:45756:80535)
Steps to Reproduce
1.open cursor
2.login cursor pro account
3.ask any thing using agent
Hey, thanks for the detailed report and the logs. On our side, requests from this corporate PC aren’t reaching the backend at all. Request ID b80592f3-aac1-49b4-aafe-ba2660569580 shows zero hits. This is a typical case where a corporate firewall or SSL inspection is blocking the traffic. The Free plan in the IDE is for the same reason. The client can’t fetch your subscription status, so it falls back to Free. ControlledExecDisposedError is a secondary cleanup after a broken stream, not the root cause.
What to ask your IT or network team to do:
Add these to the whitelist, including subdomains:
*.cursor.sh
*.cursor-cdn.com
*.cursorapi.com
Exclude these domains from SSL inspection or DLP if you use Zscaler, Netskope, or similar tools.
Allow HTTP/2 bidirectional streaming for these domains. As a workaround in Cursor, go to Settings, search for HTTP/2, and enable Disable HTTP/2. This switches it to HTTP/1.1.
After the network is fixed, run Cursor Settings > Network > Run Diagnostics. Everything should be green. If Pro still shows as Free, close Cursor, delete the Cache and CachedData folders in %APPDATA%\Cursor, then reopen it and log in again.
Let me know what the diagnostics show after the changes.
Reinstalling Cursor does not clear %APPDATA%\Cursor, so you might still have broken cache or tokens from when the connection was not working. Let’s try this:
Fully close Cursor. Check Task Manager and make sure there are no Cursor.exe processes.
Open %APPDATA%\Cursor in File Explorer and delete, or rename to be safe, these folders: Cache, CachedData, Code Cache, GPUCache, Network.
Start Cursor and sign in again.
Next, since your colleague on the same corporate setup has it working, this is likely something user-level on your machine. Please check:
Windows proxy: In PowerShell run netsh winhttp show proxy, then open Settings > Network & Internet > Proxy. Compare with your colleague.
Env vars: Run echo $env:HTTP_PROXY, echo $env:HTTPS_PROXY, echo $env:NO_PROXY in PowerShell. If you have anything set and your colleague does not, that’s the cause.
EDR, DLP, antivirus: If your account has something active that your colleague does not, like CrowdStrike, Netskope Client, Zscaler Client Connector, or a Defender for Endpoint policy, ask IT to check whether it is intercepting traffic for your user.
Share the output. It will show whether it fails on SSL handshake, times out, or returns 403.
Also please send a fresh output from Cursor Settings > Network > Run Diagnostics after clearing the cache. Paste it as text in the thread. We’ll see which endpoints are failing.
Hey, thanks for the logs. Looks like it’s not the network. curl to api2.cursor.sh returns 200 OK, and Marketplace, Authentication, Downloads, and CDN are all green in diagnostics. The issue is something else:
Extension host terminated unexpectedly 3 times within the last 5 minutes
The extension host is crashing, and all AI features crash with it since they run there. That’s why you see Canceled and Timeout waiting for EverythingProvider in diagnostics. Your colleague works on the same corporate setup, so this looks like a user-level issue on your machine.
Let’s check this step by step:
Start without extensions. Open PowerShell and run:
cursor --disable-extensions
If the agent works, a third-party extension is crashing the extension host. Re-enable extensions one by one to find the one causing it.
Extension Monitor. In Cursor Settings using Cmd/Ctrl+,, enable Application > Experimental > Extension Monitor Enabled, restart Cursor, then run Cmd/Ctrl+Shift+P and select Developer: Open Extension Monitor. This shows which extension is using too much CPU or memory, or crashing.
Windows Event Viewer. Open Event Viewer > Windows Logs > Application, filter by Source Application Error or .NET Runtime. Look for events with Cursor.exe at the time the extension host crashes. It often shows what killed the process like an ASR rule, antivirus, or something else.
Defender or EDR user-level policy. Since it works for your colleague but not for you, your account might have an extra policy. Ask IT if your account has Attack Surface Reduction rules or a CrowdStrike or Defender for Endpoint policy that your colleague doesn’t have. Especially rules like Block execution of potentially obfuscated scripts or Block child processes from Office or browser-like apps.
New Windows profile. If possible, sign in with a different Windows account on the same machine, install Cursor, and test. If it works there, the issue is in your user profile like policy, AV exclusions, or something user-specific.
Send the result from step 1 and a screenshot from Event Viewer showing the crashes, and we’ll dig deeper.