I’d love to be able to add custom models (e.g. via Groq, Open Router) without having to turn off OpenAI models that are part of my Cursor Pro plan.
Is it possible to add another way to include these custom endpoints other than through the “OpenAI API Key” settings base url and key settings.
Current UX/functionality, of course, deactivates pro plan models by assuming you’re using OpenAI’s API instead.
+1
I’m a Pro user and I often use both the default cursor-fast/GPT4 models, and my API custom models (e.g. Claude 3) too.
Every time I want to switch between the two types, I’m required to lose time in enabling/disabling my OpenAI API Key in the Settings.
Why I can’t select both the types of models without going in Settings every time?!?
+1 for this feature
Also requested something quite like this:
+1 from me too. I’d love to be able to use my premium requests for chat and composer and use a free local model for command+k line edits
+1000. I thought this feature would be more requested as I find it essential: be able to 1. have a name of your choosing “Gemini 2.0 Free” as well as the model name “gemini-2.0-flash-exp”, and 2. specify which API each model uses.
If you want to use a specific model/API my understanding is you have to disable all other models and API, or Cursor gets confused, and I don’t know about you but I get confused too between the “anthropic/claude-3.5-sonnet” (API, I think) vs “claude-3.5-sonnet” (Cursor, I think) and ditto with the OpenAI models, etc. You could easily use the wrong model since there is no way to identify them. If there was a way to manage your own json, like the Continue extension provides you with, to override the Cursor Settings > Models, would be great too. Allowing the usage of your own keys sets Cursor apart of the rest but as it is the Cursor Settings > Models panel is not really usable without being able to edit your API models.
Thanks!
Bumping this. Since they removed all thinking models from inline chat, the only way to use gemini flash for inline chat is to use our own API key.
But then you can’t use gemini pro (from cursor) anymore because you get asked to disable your key.
This back and forth is annoying.
We need this to test new models or opensource one, Some time I want test local models. I do not want to disable my cursor pro.
This is the single biggest pain point that I have with working with Cursor. I want to be able to use Cursor Auto mode, Opus 4.5 (via anthropic.com), and GLM 4.7 (via z.ai), and switch between them without having to change my settings every time. I’ve started looking into other IDEs just because this is such a big problem. Different API’s have different rate limits and up times, making it very practical to need to switch between API providers easily.
I would hope this would be a relatively simple fix since it seems like it just needs more finely-customizable config mapping for API URLs. Any updates on if this feature is in-progress would be greatly appreciated.
This is indeed the biggest pain point of Cursor. It’s barely understandable to restrict non-Pro users from using custom APIs, but even as Pro users, our experience is still very poor. I hope it can be improved soon.