What is going on? Ask mode can’t implement anything, it repeatedly gets stuck in a loop which Cursor detects: “Unrecoverable agent model looping detected.” The AI already formed a very explicit implementation plan. It just decides midway through the code to re-evaluate. I’m not charged for these useless tokens, am I?
Let me reevaluate and take a different approach.
_(It may take a moment for the assistant to restart.)_
Hey, thanks for the report. Could you let me know which model you’re using? Also, could you share the request ID? For this, you need to disable privacy mode if it’s enabled, then reproduce the issue when it happens again and copy the ID.
I’m too optimistic. We’re back! Apparently even the LLM-favored task of writing new boilerplate code from scratch was too much.
Request ID: 017295e0-5d35-4055-b0bb-31b8b1aa8c60
{“error”:“ERROR_CUSTOM_MESSAGE”,“details”:{“title”:“Unrecoverable agent model looping detected”,“detail”:“Unrecoverable agent model looping detected.”,“additionalInfo”:{},“buttons”:,“planChoices”:},“isExpected”:true}
ConnectError: [aborted] Error
at nol.$endAiConnectTransportReportError (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:4814:319459)
at egr._doInvokeHandler (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:488:211942)
at egr._invokeHandler (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:488:211684)
at egr._receiveRequest (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:488:210449)
at egr._receiveOneMessage (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:488:209271)
at O_t.value (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:488:207365)
at ye._deliver (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:49:2962)
at ye.fire (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:49:3283)
at Prt.fire (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:4801:12154)
at MessagePort. (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:6983:18168)
How do I get a refund for the money and time I’ve wasted? Just email [email protected] and get a response from a useless support bot and never talk to a human. Sounds great
Sorry you are seeing this issue! While rare, models can sometimes get stuck into unrecoverable loops, which is the error you are seeing.
Are you able to share a request ID with privacy mode disabled, so we are able to see an example of a conversation that caused such a loop?
In the meantime, I’d recommend either not using Ask mode (the model may get confused as it tries to use a tool it’s used before in Agent mode, but is unable to use now), or attempting to switch to a specific, trusted model like Sonnet or Gemini!
