Does reaching one's monthly usage limit prevent from using an Azure OpenAI key with Cursor IDE?

Support Help
, ,
1

I use Cursor IDE. I have reached my monthly usage limit. Therefore, I have added my Azure OpenAI key:

image
image1933×1286 123 KB

However, I’m still getting the error message when trying to use the model that is accessed via my Azure OpenAI key:

You’ve reached your monthly usage limit. Request higher limits to continue using Cursor.

image
image3840×1760 215 KB

Did I misconfigure something or does reaching one’s monthly usage limit prevent from using an Azure OpenAI key with Cursor IDE?

3

Hey, this isn’t a misconfiguration. Using your own API keys (BYOK), including Azure OpenAI, now requires a Pro subscription. On the Free plan, your own keys won’t work even if they’re set up correctly.

It used to work as an undocumented option, but the behavior changed. More details here: Ask mode with my own API Key required Pro Subscription

Options:

  • Upgrade to Pro ($20/month), then your Azure OpenAI key will work
  • Stay on Free and use Cursor’s built-in free models without your own keys

Let me know if you have any questions.

4

Thanks, I use Cursor Enterprise.

5

Hmm, from what we can see on our side, your account shows as Free.

If you’re actually on Enterprise, then BYOK with Azure OpenAI should work. A couple things to check:

  1. Make sure you’re logged in with your Enterprise account, the email that’s tied to the Enterprise team.
  2. Check with your Enterprise admin whether BYOK is disabled at the team level. In Enterprise there’s a setting called Restrict personal API keys, and it can block BYOK.

Let me know what you see in settings and we’ll figure it out.

6

Oh interesting, thanks very much, this is good to know, I’ll check with my admin. I use Cursor with a different account (enterprise account).

1 Like
7

Since today it April 1, I received my new monthly Cursor pocket money and wanted to retest BYOK with Azure OpenAI: no complain about the key, but also, no output. Any idea? I feel Cursor should give some kind of message (invalid key, BYOK disabled, invalid, model deployment name, etc.) so maybe worth raising an issue internally on your side. I’m leaving my enterprise email in the screenshot if you want to see the log + model deployment name to show it matches the selected model in Cursor chat:

image
image3316×1680 112 KB

Looking at https://cursor.com/dashboard/usage, it says there is some error but doesn’t give any details:

image
image2582×477 12.5 KB

The key is fine because if I mess it up on purpose, then I get a clear API key error (“User API key etc.”):

image
image2594×817 35.4 KB

Another bug is that it tries BYOK even for models that are not configured for BYOK e.g.:

image
image2629×1616 111 KB

Entire error message:

Request failed with status code 400: {"error":{"type":"client","reason":"invalid_input","message":"model is required","retryable":false}}

Thanks!

8

Thanks, I see the screenshots and the dashboard. I can confirm both bugs.

  1. “Errored, No Charge” on gpt-54-nano-podcast: This is a known issue with Azure OpenAI BYOK. Cursor isn’t passing the model parameter correctly to the Azure API. A few users in similar threads are hitting the same thing: Not able to use azure api key and Azure Open AI any model. The team is aware, but there’s no ETA yet.

  2. Routing non-Azure models through an Azure endpoint (error on claude-4.6-opus-max): Good catch. Cursor shouldn’t try to use Azure BYOK for models that aren’t configured there.

  3. Unclear error messages: I agree. Cursor should give clearer feedback like invalid key, BYOK disabled, deployment mismatch, etc.

As a workaround, you can try this proxy service: GitHub - gabrii/Cursor-Azure-GPT-5: A service that allows Cursor to use Azure GPT-5 deployments. · GitHub. It adapts the Azure OpenAI format for Cursor. Not ideal, but some Enterprise users are using it.

Also, send your Cursor version Help > About Cursor > Copy. I want to make sure we didn’t miss any recent fixes.

9

Thanks, currently I have Cursor version 3.0.13 but when doing these bug reports I was on earlier versions. I’ll let you know when I try again.

10

this repo has not been update in months, the last model it supports is 5.2

11

if anyone figured out this problem I’d be glad to know.

I add the url from foundry, with the V1/responses or should I not? deployment name is a problem, in azure you actually need to deploy any model you gonna use, so if you have 4 different models, how do you add them to your azure openai, oh wait cursor never fixed that, or did they, it’s been there forever but never working, really strange.