I'm confused about using your own API vs a Cursor plan, how do I know which it's using?

General
1

I’m confused about how all this works together. I have added my own API keys into cursor, and I realize some features like the ‘apply’ button can’t use the API.

I am really impressed with Cursor and would be more than willing to pay for it just for the specific Cursor exclusive features, but I don’t want to be locked into the model limits they have, I just want to charge those to my own API. Like, I’m definitely going to use more than 10 o1-mini requests per day, and I will pay whatever API costs necessary to keep that functional. I really wish they would just put a price on the software, and charge separately for credits cause I don’t want to have to think about limits.

But I can’t really even tell whether it’s using my API or its own, or both for different things depending on the circumstance. The whole thing is frustratingly opaque.

Can anyone provide some clarity on this?

1 Like
2

Hi @plymouthvan ,

I am away from computer at the moment so can’t provide a more detailed response to all your queries, however have you viewed your usage requests when logged in at https://www.cursor.com/settings?

It might give you some more insight on which models are being used for which requests etc.

You can view them by going to the link above and then clicking on a link which toggles open usage information, eg:

{"chat":{"modelIntent":"gpt-4o"}}

{"contextChat":{"modelIntent":"gpt-4o-mini"}}

{"fastApply":{"isOptimistic":true}}

{"fastApply":{"isOptimistic":false}}

{"composer":{"modelIntent":"gpt-4o"}}

{"slashEdit":{"modelIntent":"gpt-3.5-turbo"}}

{"cmdK":{"modelIntent":"gpt-4o"}}

Cursor Tab uses a custom model.

A few other resources that may be relevant:

https://docs.cursor.com/advanced/models

https://docs.cursor.com/advanced/api-keys

If you add an API key, enable it, and then select the associated LLM when making a prompt, it will use your API key, at your own cost.

Otherwise, it will use Cursor’s built in access, which is probably cheaper as they have access to bulk pricing etc.

Some other users have previously requested on the forum the ability to just pay for Cursor Tab, and nothing else, but that is not a service offering at the moment.

3

In Cursor Settings | Models, if your API Keys are turned off (see screenshot), you’re using Cursor’s API’s. If they are turned on, it’s using yours. Note that in my experience, if you turn on even one of your API’s, say for OpenAI, Cursor won’t let you use their models like Anthropic. All or nothing.

image
image2360×906 168 KB

1 Like
4

I definitely have my keys set up, but it still lets me use the apply feature, so it’s just unclear what model its using.

5

I’m still on the trial pro plan right now, and I have my keys set up. My understanding is that the “apply” feature requires the cursor’s model, and I’ve used it a bunch, but the usage report in my settings are listing no usage at all. For what it’s worth, my API is racking up usage on o1-mini, and the Apply feature still works, so I guess it’s using both. I wish the app would just be transparent about this.

6

Hi @plymouthvan ,

Just to clarify:

  1. Have you gone to https://www.cursor.com/settings?

  2. Can you see the usage requests, like those shown here:

settings_usage_requests
settings_usage_requests1621×981 108 KB

  1. Which API Keys have you set up in Cursor and enabled?

  2. Can you provide specific steps of what you are doing, e.g:

I am sending a prompt in Ctrl + L Chat with the gpt-4o LLM selected

  1. Can you specify the expected behavior? e.g:

I expect this request will use my own API Key and I will be able to see this request listed at cursor.com/settings

  1. Can you specify the actual behavior? e.g:

What is actually happening is …

Additionally…

If you have specific questions you want answers to, please list them.

For example:

  • What model does Cursor Tab use?

  • What model does Apply use?

  • If I have my own API Key enabled, does it change which model is used by Cursor Tab and Apply?