Could not reach provider

Support Bug Reports
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Using the Chat GPT model, I got several “Could not reach provider”. The problem is when this happens after some initial work. When I press “Try again” it starts planning and analysis again from scratch. This wastes a lot of my tokens.

Steps to Reproduce

Wait for “Could not reach provider”, press “Try again”

Expected Behavior

After “Try again” it should continue where it was when it lost connection. It appears it starts from the original chat prompt.

Operating System

Linux

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.2.44
VSCode Version: 1.105.1
Commit: 20adc1003928b0f1b99305dbaf845656ff81f5d0
Date: 2025-12-24T21:41:47.598Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Linux x64 6.14.0-1018-oem

For AI issues: which model did you use?

ChatGPT 5.2 Extra High

Does this stop you from using Cursor

No - Cursor works, but with this issue

2

Hey, thanks for the report.

Got it, the issue is that on retry the agent really starts over instead of continuing from where it failed. This is a known issue, and the team is working on improving the retry flow.

To debug the “Could not reach provider” error itself, we’ll need a bit more info:

  1. Copy the Request ID right after the error (Chat menu > Copy Request ID)
  2. Cursor Settings > Network > Run Diagnostics, then share a screenshot of the results
  3. What region are you in? Are you using a VPN or a proxy?

Temporary workaround if this happens often:

  • Try disabling HTTP/2: Settings CMD+, > search “HTTP/2” > enable “Disable HTTP/2”, then restart Cursor
  • If you’re using a VPN, try turning it off

Share the Request IDs and the diagnostics results, that will help the engineers figure out what’s causing the connection drops.

3

I do not have the request id available (closed this chat since). One additional observation: when this happens, it sometimes shows only “Try Again,” and sometimes it also shows the “Resume” button in the chat.

1 Like
4

It is happening all morning today. Could not reach OpenAI. Request ID: e188ec9f-6561-4358-ae44-2a167ff7e4c3

7

Thanks for the info. We checked and found an interesting error in the logs: “Append data exceeds size limit”. Something in your request is going over 50 MB.

A few questions to help us debug:

  1. Do you have a large .cursorrules file in your project? If yes, can you share its size?
  2. Are you using Claude Skills (imported from the .claude directory)?
  3. Can you run Cursor Settings > Network > Run Diagnostics and share a screenshot of the results?

This will help us figure out what’s blowing up the context to that size. Usually it’s one of these.

About the retry behavior (when the agent starts from scratch), that’s a known issue, and the team is working on improving the retry flow.

8

This topic was automatically closed 22 days after the last reply. New replies are no longer allowed.