The client has been completely broken for a week now. All requests to the agent fail with the message taking longer than expected. Please note that agents work correctly through the browser and do not work at all in the client. Yesterday, there was a short period when the agent was working a little, but requests often failed with the reconnecting status.
Hey, this is a known issue affecting multiple users on v2.5.x. The fact that cloud agents work but the desktop client doesn’t points to a client network or connection issue.
Here are a few things to try:
Disable HTTP/2: Press Ctrl+,, search for “HTTP/2”, enable “Disable HTTP/2”, then restart Cursor. This has helped several users.
Run network diagnostics: Cursor Settings > Network > Run Diagnostics. Send a screenshot of the results here so I can see what’s failing.
Reindex the project: Go to Cursor Settings > Indexing & Docs > Codebase Indexing and click Resync Index. One user with the exact same symptoms fixed it by reindexing.
Try a different network: If you’re on a VPN, behind a corporate firewall, or can switch to a mobile hotspot, try that to rule out your ISP or router.
Test in an empty folder: File > Open Folder > pick an empty folder, start a new chat, type “hello”. Does it still hang?
Are you using a VPN or proxy? What model are you using (Auto, Claude, Gemini, etc.)?
I did some more tests based on the current situation.
My problem is fixed by disabling HTTP2 and reindexing (don’t know which of it solved the problem). However, after doing this, I can only use my OpenAI API Key, which generates response. The official Cursor auto mode still gets stuck on Taking Longer than Expected.
Hey @redfenix, your diagnostics are all green, so it doesn’t look like a network issue on your side. That means the problem is likely on the client or server side. To dig in, I need a Request ID from a failed or stuck request. You can find it in the top right of the chat panel, open the context menu, then click Copy Request ID. Without that, I can’t trace what’s happening on the backend.
Also, have you tried fully reinstalling Cursor? Since you’re on 2.5.25, try updating to 2.5.26 too. If you’re already on 2.5.26, please do a clean reinstall.
@Fernando-Fernandes, your case is different. In diagnostics, Authentication and Downloads are failing. A few things to try:
Remove the disableHttp1SSE setting and keep only disableHttp2: true. With both disabled, it forces HTTP/1.0, which could be why those two endpoints fail.
Check if anything on your Mac is blocking connections, like Little Snitch, firewall rules, DNS filtering, or security software.
I also need a Request ID from a stuck request so I can escalate.
Once I have Request IDs from both of you, I can pass this to the engineering team.
I have tried all the methods you suggested, but the issue of “Taking longer than expected” still persists. I would like to know how to solve it. Request ID: 9ccdfde1-6092-4365-8644-7ff412b7057a
Been getting this for the last week too. All the fixes for this error are either temporary or do not work, really doubt there is a consistent client side fix for this.
