I want to add a custom model. I can add a model and name it but I don’t see anywhere where I can adjust things such as api key, url, etc? I see some older screenshots that have an api keys section. I do not see that at the bottom. I do not see any way to edit models. Is there something I am missing? (If it makes any difference - I am on an enterprise plan)
Hey @something_something, thanks for the question, and you’re not missing anything!
Your team’s Cursor admin has turned on a setting called “Restrict Personal API Keys” for your organization. When that’s on, Cursor hides the whole API Keys section under Cursor Settings → Models, which is why you don’t see a place for an API key or base URL like in older screenshots.
A custom model needs its own key and endpoint, so custom models you add won’t work while that restriction is in place. You can remove the one you added with the trash icon next to it, and keep using the models your team has enabled in the list.
If you need a custom model or your own API key, reach out to your Cursor team admin. They can review “Restrict Personal API Keys” in the team dashboard under Settings → Models. Once they turn it off and you restart Cursor, the API Keys section will show up again at the bottom of the Models page.
Under API Keys section in Settings → Models I have added the following in OpenAI API Key section:
- API Key: set to custom api key
- Use OpenAI API Key: toggled on
- Override OpenAI Base URL: toggled on
- Base URL: set to our custom base url
I then have added a custom model that is available at our URL.
However, when I try to use the model, I get:
AI Model Not Found
Model name is not valid: “our-unique-model-name-here”
I have set this model up with other harnesses on openai protocol with success, so I know all the parameters are set correctly.
Does something else need to be configured?
Hey @something_something, thanks for following up!
Two things are getting in the way here:
-
Your team now allows personal API keys in general, but each provider also has its own switch, and the OpenAI one is still off for your team. Because you’re using a custom base URL, Cursor treats your model as an OpenAI API key request, so it’s refused. Your Cursor team admin can turn it on in the team dashboard under Settings → Models → Personal API Keys (BYOK) → OpenAI.
-
Cursor chooses which of your saved keys to send based on the start of the model name. If your custom model’s name begins with
claude-, Cursor uses the Anthropic API Key slot instead of the OpenAI one, so with only an OpenAI key saved the request goes out with no key. That’s what produces the “Model name is not valid” message. Using a name that doesn’t start withclaude-(for example an alias on your endpoint) avoids it.
Once your admin has enabled OpenAI:
- Open Cursor Settings → Models.
- Remove the current custom model with the trash icon.
- Add it again with a name that doesn’t start with
claude-. - Confirm “OpenAI API Key” is on and “Override OpenAI Base URL” still points to your endpoint.
- Restart Cursor, open a new chat, and select the model.
Admin has enabled OpenAI personal API keys. Our model name has been updated to not include claude-. I am no longer getting the AI Model Not Found,but now I am unable to make any requests because I get:
You've reached your monthly usage limit
Request higher limits to continue using Cursor.
Shouldn’t that be avoided with our custom model setup?
On your team’s plan, requests that use your own API key are billed by your provider for the model itself, but Cursor still adds a small per-token platform fee (the Cursor Token Rate). Since that fee is a Cursor charge, those requests still count toward the monthly spend limit your team sets for each member.
Your usage from earlier this month (on Cursor’s built-in models) already reached that per-member limit, so new requests are paused, including ones using your custom model.
To keep going, you can either:
- Ask your Cursor team admin to raise your monthly spend limit in the team dashboard (the per-member usage limit), or
- Wait for your team’s usage cycle to reset on October 1.
Once the limit is raised, restart Cursor and open a new chat, and your custom model should work.
Thanks @Colin. My admin raised my limit a bit and we were able to get our custom model working with all the information you outlined above.