Using OpenAI api issues

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I have activated openAPI with key under /Models and now every internal model I try including composer 2.5 is trying to route via that openAI api.

Steps to Reproduce

  1. add openai api key, activate it
  2. try using composer model in chat

Screenshots / Screen Recordings

Operating System

Windows 10/11

Version Information

Version: 3.16.29 (user setup)
VS Code Extension API: 1.128.0
Commit: 6246455961129c32969845e00aa25d87ae926ec0
Date: 2026-08-18T01:26:26.285Z
Layout: Agent Window
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.291
OS: Windows_NT x64 10.0.26200

For AI issues: which model did you use?

spark, composer

For AI issues: add Request ID with privacy disabled

Request ID: 57ddb746-9d16-4e24-b2ad-f45d5e53ca40

Does this stop you from using Cursor

Yes - Cursor is unusable

Hey, thanks for the detailed report and the request ID. I’ll look into what’s happening. This isn’t a problem with your setup, it’s how BYOK works right now.

When your OpenAI API key is enabled in Settings > Models, Cursor attaches that key to all models except Claude and Gemini, including Cursor-hosted models like Composer 2.5 and Spark. But those models always run on Cursor infrastructure and can’t be billed to a custom key, so the request gets rejected with This model does not support custom API keys. Important: nothing is sent to OpenAI, the request is rejected before the provider is called.

What to do:

  • If you want to use Composer 2.5, Spark, Grok, or other Cursor models, turn off the OpenAI key toggle in Settings > Models. You can keep the key saved and turn it back on when you want to use OpenAI models.
  • As an alternative, use Auto mode. It won’t send BYOK creds when routing to internal models.

We hear you that switching the key on and off is annoying. This is a limitation of the current BYOK implementation, and I’ve shared this feedback with the team. I can’t share an ETA yet.

Let me know if turning off the toggle fixed it.

Yes it does work however super inconvenient. Thank you for the quick reply.

Hello! any status updates on this, it’s super inconvenient atm

Hey, there’s no timeline update yet. Like I mentioned above, this is a current limitation of how BYOK works: when the OpenAI API key toggle is enabled in Settings > Models, that key gets attached to every model except Claude and Gemini, including Cursor-hosted models like Composer 2.5, Spark, and Grok. Those models always run on Cursor infrastructure and can’t be billed to a custom key, so the request gets rejected. Nothing is sent to OpenAI, the request is rejected before the provider is called.

For now, the simplest workaround is:

  • If you need Composer 2.5, Spark, Grok, or other Cursor models, turn off the OpenAI key toggle in Settings > Models. You can keep the key saved and turn it back on when you’re using OpenAI models.
  • Alternatively, use Auto mode. It doesn’t send BYOK creds when routing to internal models.

I know switching the key on and off is annoying. We’re tracking this limitation, but I can’t share an ETA yet. If there’s an update, I’ll reply in this thread.