Agent mode connection issues (Chat mode works fine)

Hi everyone,

I’m running into a problem with CursorAI and wanted to see if anyone else has experienced the same thing.

Whenever I try to use Agent mode, I get the following error:

Connection failed. If the problem persists, please check your internet connection or VPN
(Request ID: a0aa6584-253c-48a2-9fac-1b20439fab39)

Interestingly, Chat mode works perfectly fine.

Here’s what I’ve already tried:

  • Using public DNS (8.8.8.8 and 1.1.1.1)
  • Disabling HTTP/2
  • Restarting the CursorAI app

I’m not using any VPN or proxy, and my internet connection is otherwise stable.

If anyone has faced a similar issue or has any suggestions, I’d really appreciate your input. Thanks in advance!

CursorAI version data:

Versión: 0.49.4 (Universal)
Confirmar: 1.96.2
Fecha: ec408037b24566b11e6132c58bbe6ad27046eb90
Electron: 2025-04-22T00:33:06.906Z
ElectronBuildId: 34.3.4
Chromium: undefined
Node.js: 132.0.6834.210
V8: 20.18.3
SO: 13.2.152.41-electron.0

Yeah, I’m having the same problem, and there’s a spike forming on DownDetector, so I think this is another temporary issue on Cursor’s side. It’s happening with all the models I’ve tried.

Since yesterday, I started having issues with Cursor, nothing works, agent, ask, manual.

Ouch, in my case, ASK mode is working fine

in your case, only Agent fails ? or manual/ask as well ?

This has been happening to me for about 3 weeks. I tried disabling HTTP/2 as some have suggested but that didn’t do anything. Closing cursor entirely and reopening it fixes it for a little while (maybe half an hour) but it always messes up again.

Me too, it’s becoming unusable. I tried a clean install too. It keeps happening and I have hardly been able to get anything done for the last 2 hours.

Is it still down for you?

yep, now:
Connection failed. If the problem persists, please check your internet connection or VPN

(Request ID: 238d7774-dd8d-44ef-8c74-35a4c30bd6fc)

OK, that’s weird. Mine resolved itself within an hour or two, after whatever outage happened yesterday. (Here’s a screenshot of the spike on downdetector.)

I had assumed you were having the same problem as me, due to the timing, but it appears your problem is different.

I’m having the same problem repeatedly today. I’m using premium models and they are making a lot of tool calls but I get the error before the response is complete and code changes are suggested. Are we being charged for these errors? [Edit: Just checked an there is no charge for the errors]

I fixed my bug…
in MCP servers I had this (that I removed and Agent starts working)

,
  "security": {
    "trustedServers": ["xxyy.local.test"],
    "allowAllToolsFromTrustedServers": true
  }

Same issue here. Ask works, agent fails more often than not. Initially exiting and restarting worked, now not even that.

Ask mode works, but of course no apply model, so cursor is down for me.

Honestly the product is just getting worse and worse for me.