Hey, I’m a little bit confused as when I go to settings I can add custom model name and that’s it. I cannot configure this, so what’s the point of this setting? I want to use my company proxy + api key + model name but I cannot do so.
I could use the OpenAPI key section but there is a problem, it tries to find model:
“model”: “gpt-4o”
where my company named it: openai/gpt-4o so I also cannot use it. Can someone tell me how to properly configure this?
Hey, try disabling all models and leave only the one you added. This should resolve the issue during API key verification.
not really, I cannot set openai/gpt-4o model, I believe the Add Model function doesn’t support /. After adding this model it just doesn’t appear into the list. @deanrie
You might be clicking the “Add model” button. Try using Enter for this after you enter the model name.
It works! Thank you!!! You saved my life 
@deanrie I was happy too early. Now I got this error:
This model does not support tools. Please select a different model and try again.
is this sth on our side or Cursor setting?
Is it in a pop-up window? Or is it an error in the chat window? If you’re trying this in agent mode, switch to ask mode.
Hey, do you have any MCP servers enabled?
@danperks Nope, I didn’t do that, should I?
ok trying with local MPC, when running it locally it starts on port 4333, however cursor looks for either 80 or 3000, I’m using stdio and also looked into docs:
Hey, it looks like you have an overridden base URL for the OpenAI API key. Try resetting to default, and it should work.
Hi,
Did you sovle the problem? I got the same problem.
Hey, is it working in ask mode for you? I also tried creating a custom mode, and everything works fine there too.
Hi, I think the issue happens when you first use cursor premium model with agent mode (need call tools like read or search) then you change it to use Optimus-alpha with ask mode.
Yes, it seems you are right, you need to create a new chat for this to work, thanks.
cc @Kazyn
Thanks for the response. So it is a bug, please fix it as switch models between cursor premium and custom API should be a frequent operation.
