Why requests to claude-xx uses custom azure api config?

Bug Report
1

I have my own azure openai api key configured, which is fine when using gpt models.
but when I switch to claude-xx model, the error indicates it’s trying make the request via azure server (check the snapshots below). why?

image
image3064×2070 461 KB

2

am I the only one who is facing the issue?

3

Hi @bigfatsea

You need to disable your API key every time you want to use models from your Cursor subscription.

1 Like
4

This seems a bit off. Shouldn’t it be:

  • If a custom key is available for the selected model, use it.
  • Otherwise, use the service provided by Cursor.

In my case, the Azure OpenAI key should only be used for GPT models. It doesn’t make sense to use this key when choosing Claude models.

1 Like
5

I agree, @bigfatsea , that it should be the way you described (more intuitive and flexible, in my opinion); however, the way it actually works is how @deanrie mentioned. There may be some logic on the back-end for why it’s designed that way but for now, this is how we work around it.

6

This might get fixed in the future, but for now, that’s how it works. Since everything runs from the same source, it’s probably hard to determine whether you want to use your API key or the models from your subscription. Both options include GPT models.