Hey, thanks for the report. This is a known regression in Planning Mode that started in version 2.1.0 and is affecting multiple users. The team is already aware and investigating.
To help link your case, could you please share the Request ID from a failed planning attempt?
I’ll attach future request IDs here, plan mode doesn’t give the same request ID option in the same spot ( and I didnt know it was in top right at all times )
Thanks for sharing the Request ID and the screenshot with the “invalid string to replace” error! This is really helpful for the team to track your specific case.
The “invalid string to replace” error is related to the same Planning Mode regression that’s currently being investigated. Your request has been linked to the tracking thread.
As a temporary workaround while the fix is in progress:
Try starting a new plan for major changes instead of editing existing ones.
If you’ve saved plans to .cursor/plans/, you can edit the markdown directly there.
Request ID: dc52bf6d-75fa-4b11-a92a-8a14894a9b93
"{\"error\":\"ERROR_BAD_REQUEST\",\"details\":{\"title\":\"Bad request.\",\"detail\":\"Bad Request\",\"isRetryable\":false},\"isExpected\":true}"
ConnectError: [invalid_argument] Error
at oNc.$endAiConnectTransportReportError (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:6426:446069)
at iDo._doInvokeHandler (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7047:22831)
at iDo._invokeHandler (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7047:22573)
at iDo._receiveRequest (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7047:21335)
at iDo._receiveOneMessage (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7047:20152)
at ODt.value (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:7047:18244)
at ke._deliver (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:2962)
at ke.fire (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:49:3283)
at hmt.fire (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:6411:12156)
at MessagePort.<anonymous> (vscode-file://vscode-app/d:/Program%20Files/cursor/resources/app/out/vs/workbench/workbench.desktop.main.js:9179:18433)
This happens consistently after the initial plan. The initial plan creates consistently, but if I submit a follow up message in the chat with notes/comments/corrections to update the plan in plan mode, it happens fairly often
–// speculation only
Funny how it always happen to the great improved plan that included a bit of “scenarios”. To me they just ban when you had a plan for something good that you could just feed to LLM to execute. “You just found a secret formula - so we will ban you and tell it’s LLM’s going retarded mode for fun effects”.
Try clearing cache, then use specific mode. Worked once.
It’s honestly insane that this bug hasn’t been fixed for such a long time. This is literally one of the core features to Cursor, I don’t know how this wasn’t prioritised to be fixed ASAP weeks ago - IMO it absolutely should have been fixed within a few hours of the first report. Infuriating when this happens when you’re trying to plan a massive feature and have to start a new conversation just to get around this error. You even see it happening multiple times in Theo’s (T3) recent YouTube video that has 100k views.
My workaround that usually works is to say something like “delete the entire existing plan file completely and write the plan to a new plan file again from scratch”.
