OpenAI Models not working in IDE or CLI with "reconnecting" and "resource_exhausted"

Support Bug Reports
, ,
1

Where does the bug appear (feature/product)?

Cursor IDE (this is for the last ~3h)

Describe the Bug

Hi! I’m having a problem in the IDE and CLI when trying to use OpenAI models. It is working for Claude and Composer 1.5 models.

I don’t have any personal keys or endpoints enabled.

Steps to Reproduce

Select an OpenAI Model (tried gpt-5.3-codex, gpt-5.2-codex-high)

Operating System

MacOS

Version Information

Version: 2.4.31
VSCode Version: 1.105.1
Commit: 3578107fdf149b00059ddad37048220e41681000
Date: 2026-02-08T07:42:24.999Z (2 days ago)
Build Type: Stable
Release Track: Default
Electron: 39.2.7
Chromium: 142.0.7444.235
Node.js: 22.21.1
V8: 14.2.231.21-electron.0
OS: Darwin arm64 24.6.0

About Cursor CLI

CLI Version 2026.01.28-fd13201
Model GPT-5.3 Codex Extra High
OS darwin (arm64)
Terminal vscode
Shell zsh

Additional Information

Failure 1:

Request ID: cb8ad735-e778-410c-9ae5-b68468b9ab79
{"error":57,"details":{"title":"Provider Error","detail":"We're having trouble connecting to the model provider. This might be temporary - please try again in a moment.","isRetryable":true,"additionalInfo":{},"buttons":[],"planChoices":[]},"isExpected":true}
[resource_exhausted] Error
OTe: [resource_exhausted] Error
    at lpf (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9097:38348)
    at apf (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9097:37251)
    at Cpf (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9098:5088)
    at Tva.run (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9098:9098)
    at async Vyt.runAgentLoop (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34206:57104)
    at async mgc.streamFromAgentBackend (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34255:7835)
    at async mgc.getAgentStreamResponse (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34255:8576)
    at async qTe.submitChatMaybeAbortCurrent (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9180:14965)
    at async Xi (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:33004:3808)

Failure 2:

Request ID: 69a9a9c4-9e49-45b7-8f99-e4d2fc9d82bb
{"error":57,"details":{"title":"Provider Error","detail":"We're having trouble connecting to the model provider. This might be temporary - please try again in a moment.","isRetryable":true,"additionalInfo":{},"buttons":[],"planChoices":[]},"isExpected":true}
[resource_exhausted] Error
OTe: [resource_exhausted] Error
    at lpf (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9097:38348)
    at apf (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9097:37251)
    at Cpf (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9098:5088)
    at Tva.run (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9098:9098)
    at async Vyt.runAgentLoop (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34206:57104)
    at async mgc.streamFromAgentBackend (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34255:7835)
    at async mgc.getAgentStreamResponse (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:34255:8576)
    at async qTe.submitChatMaybeAbortCurrent (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:9180:14965)
    at async Xi (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:33004:3808)

Does this stop you from using Cursor

Sometimes - I can sometimes use Cursor

3

Failure in the IDE

image
image852×178 2.63 KB

image
image854×410 26 KB

Success in the IDE

image
image854×214 13.4 KB

4

Hey, thanks for the report. I can see the error logs and screenshots. It’s the classic resource_exhausted pattern.

The fact that Claude works but all OpenAI models fail could point to a temporary capacity issue on the provider side, but first let’s rule out the most common local cause.

Can you check one thing: in Cursor Settings > Tools & MCP, do you have any MCP servers connected? If yes, try temporarily disabling all of them. In a few identical cases, that was the root cause:

Also try:

  1. Log out and log back in (Cursor Settings > Log Out, then Log In)
  2. Start a new chat in a fresh project folder

I noticed you also posted in this thread: Composer 1.5 and Codex 5.3 show "Reconnecting" repeatedly, then a Provider Error and cannot connect to the model provider. The same steps apply there too.

If you don’t have any MCP servers and logging out and back in doesn’t help, let me know.

5

Disabling the MCP servers worked! I had previously tried it without success.

Since they work with other models, is there a reason they might fail specifically for OpenAI models? They’re particularly useful in my workflow so I would like to be able to re-enable them if possible.

Thanks very much for your help so far, much appreciated.

7

Glad disabling MCP helped.

About why it’s specifically OpenAI models, that’s a good question. Based on other cases, the issue usually isn’t model-specific. Here’s an example where both Anthropic and OpenAI were affected: Trouble connecting to the model provider for all Anthropic and Open AI models

Most likely, one of your MCP servers is conflicting specifically with OpenAI models. Different providers handle MCP tool calls differently, and that can cause patterns like this.

8

Thanks again, I’ll try to play around and see if I can make it work somehow.

1 Like