Is Cursor IDE secretly using OpenAI even when you select Claude?

I’ve been using Cursor IDE for a while, and I recently ran into something that seems odd. I had Claude 3.5 Sonnet selected as my assistant the entire time and never switched to OpenAI models. But then, when an error occurred, the message said:

“Unable to reach OpenAI. Try again.”

Why is an OpenAI-related error popping up when I’ve only used Claude? :thinking:

This raises some questions:

  1. Is Cursor making hidden API requests to OpenAI in the background, even if you select Claude?

  2. Are some features still relying on OpenAI despite user settings?

  3. Or is this just a UI bug that incorrectly labels errors?

If they’re routing requests to OpenAI without clear disclosure, that’s a transparency issue. Users should know exactly where their prompts are going,
especially when working in an AI-assisted coding environment.

Has anyone else noticed this? Would love to hear thoughts from the community and maybe even some clarification from Cursor’s team.

Hey, in the Composer, we use some of the smaller OpenAI models, namely gpt-4o-mini behind the scenes for certain purposes, and that is likely what happened here.

OpenAI had some downtime around when you were facing this I believe, which may have been the cause. Everything should now be working as expected!

1 Like