When generating code with Cursor, the editor always gets stuck showing Taking longer than expected. This issue has persisted for many days, and the software is completely unusable — not a single line of code can be generated successfully.
I have updated the application multiple times, but the problem remains unresolved. I access Cursor via VPN, and there were no such issues in the previous months.
I have checked solutions from similar existing issues.
my connection uses HTTP/1.1.
I ran Network Diagnostics — results show normal.
I do not know how to perform re-indexing codebase. My codebase indexing status constantly shows “Loading” and never completes.
Hey, thanks for the report. From your screenshot, diagnostics pass and HTTP/2 is already disabled, good.
Since you’re using a VPN, that’s probably the root cause. A few users with the same symptoms fixed it by switching networks or VPN servers. Even if diagnostics look green, the streaming connection to model providers can still get blocked or throttled by some VPN setups.
Here’s what I’d suggest:
Try a different VPN server or provider. If you can, try without the VPN to confirm it’s the cause.
Re-index your codebase. Open the Command Palette Ctrl+Shift+P, then type Cursor: Resync Codebase Index. This fixed the “Taking longer than expected” issue for another user with the same symptoms. It should also fix the “Loading” status you’re seeing in codebase indexing.
Test in an empty folder. Go to File > Open Folder, pick an empty directory, start a new chat, and type “hello”. This helps rule out project-specific issues.
Grab a Request ID next time it gets stuck. Open the context menu (top right of the chat) and select Copy Request ID. That’s the main thing we need to check on the backend.
Which model are you using? And can you confirm if the issue still happens in a brand new empty folder?
I create a new folder and start a new chat, I send a request, cursor create a new .py file but it seems the request is not finished. If I send another request it will be Queued. The request ID is 757d8ee2-6fa8-4224-99d6-c4b59d895ee6
Then I want to do some edit on the newly generated files so I send another request, it sill stucked during ‘planning next moves’ then shows ‘Taking longer than expected’.
To sum up, I paid for the Cursor Pro plan last week, but since then, none of my actual requests (excluding the test requests used to explain the issue in this bug report) have succeeded even once.
Could you also let me know if I am eligible for a refund and how to refund?
Not worked for me, I install 2.4.37. The agent can finish some basic task like writing a 50-line python program. But for some complex task, It stuck at ‘Planning next moves‘ again.
The Resync Codebase Index command looks like it was removed. Try opening Settings with Ctrl+Shift+J, then go to Indexing & Docs and click the button there.
The fact that in an empty folder the first simple request works, but more complex tasks hang points to a VPN or network issue. The streaming connection drops on longer responses. Try:
Switching to a different VPN server region, closer to the US or EU
If possible, testing without a VPN
Trying a different model, like GPT-5. Sometimes routing to a specific provider gets blocked
About a refund, email [email protected] with a description of what happened. The billing team can help.
Let me know if switching the VPN server or the model changes anything.
I don’t know if it’s a bug or if I’m just really stupid, but at this point I have tried everything and I am not getting any human response from the support mail.
The cursor agent is not working for me, it always starts by saying “planning next moves“, and then “taking longer than expected“, (the file in the prompt is 38 lines):
Then it just sits there forever (it has been at least an hour) and I don’t get any answer. Could someone please help me, I have done a lot of troubleshooting but I might have been checking down the wrong path. It has been like this for the past two days.
I used cursor and paid for pro plan for few month last year, it worked fine. But from last week, the software is completely unusable for me, I guess I have to refund.
It doesn’t sound like you’re doing anything wrong. From the behavior you describe, the Cursor agent is most likely getting stuck in the planning phase before executing any tool calls.
When the agent shows messages like “Planning next moves” followed by “Taking longer than expected” and then never proceeds, it usually means the planner is waiting on an internal tool call (like reading files, searching the workspace, or preparing edits) that never returns. In that state, the agent can sit there indefinitely without producing a response.
Given that your prompt file is only 38 lines, this should normally complete in a few seconds, so an hour strongly suggests the agent workflow is stuck rather than actually processing the task.
A few things that typically resolve this:
Fully restart Cursor (close the app completely and reopen it).
Reindex the workspace: open the command palette and run Cursor: Reindex Workspace.
Close and reopen the project folder (File → Close Folder, then Open Folder again).
If you’re using a .code-workspace, try opening the root folder directly instead.
Try switching models temporarily and then switching back.
If the issue persists, it may also be a backend agent service issue, so sharing the request ID from that run with the Cursor team can help them trace where the agent got stuck.
In short: the agent isn’t actually working for an hour it’s likely blocked in the planning stage, which is why you never receive a response.
Hey @Dirk, I can see the standard steps didn’t help. A couple things to try:
Delete the index completely: Settings Ctrl+Shift+J > Indexing & Docs > Delete Index, then wait for it to reindex from scratch. Resync isn’t the same thing. This fixed it for a few users with the same issue.
Check if you have a .cursor folder in your project root or in %USERPROFILE%\.cursor. If there are any skills files there, try temporarily renaming the folder to .cursor_backup and restart Cursor. In this same thread, @Johannes_Claesson found the cause was incorrectly formatted cursor skills.
Share the Request ID from the stuck request on version 2.4.37, we need it to check on the backend. In chat, click the three dots in the top right > Copy Request ID.
The team is aware of the issue, and every new Request ID helps us prioritize it.
