Cursor and GPT 5.2 On Azure not working

[Gil_Maman]

Where does the bug appear (feature/product)?

Cursor IDE

---

Describe the Bug
When using a costum Azure OpenAI key to a GPT 5.2 on Azure, the LLM works fine until it needs to make code edits.

Then it writes the error: ”Connection failed. If the problem persists, please check your internet connection or VPN”

---

Steps to Reproduce
How can you reproduce this bug? We have a much better chance at fixing issues if we can reproduce them!

1. Deploy an gpt 5.2 on Azure
2. Use Azure OpenAI Costum key in Cursor
3. Write a prompt to edit a file

---

Expected Behavior
What is meant to happen here that isn’t working correctly?

The chat and editing should work

---

Screenshots / Screen Recordings
If applicable, attach images or videos (.jpg, .png, .gif, .mp4, .mov)

---

Operating System

• Macosx

---

Version Information

• For Cursor IDE: Menu → About Cursor → Copy

• For Cursor CLI: Run agent about in your terminal

Cursor info:

Version: 2.3.29

VSCode Version: 1.105.1

Commit: 4ca9b38c6c97d4243bf0c61e51426667cb964bd0

Date: 2026-01-08T00:34:49.798Z (2 wks ago)

Electron: 37.7.0

---

For AI issues: which model did you use?
Model name (e.g., Sonnet 4, Tab…)

gpt-5.2

---

For AI issues: add Request ID with privacy disabled
Request ID: a8aac1a6-9ccc-4241-92df-c4f1fcbb4f6b

---

Additional Information
Add any other context about the problem here.

---

Does this stop you from using Cursor?

• Yes - Cursor is unusable

[system]

Hi there!

We detected that this may be a bug report, so we’ve moved your post to the Bug Reports category.

To help us investigate and fix this faster, could you edit your original post to include the details from the template below?

Bug Report Template - Click to expand

Where does the bug appear (feature/product)?

• Cursor IDE
• Cursor CLI
• Background Agent (GitHub, Slack, Web, Linear)
• BugBot
• Somewhere else…

---

Describe the Bug
A clear and concise description of what the bug is.

---

Steps to Reproduce
How can you reproduce this bug? We have a much better chance at fixing issues if we can reproduce them!

1. …
2. …
3. …

---

Expected Behavior
What is meant to happen here that isn’t working correctly?

---

Screenshots / Screen Recordings
If applicable, attach images or videos (.jpg, .png, .gif, .mp4, .mov)

---

Operating System

• Windows 10/11
• MacOS
• Linux

---

Version Information

• For Cursor IDE: Menu → About Cursor → Copy
• For Cursor CLI: Run agent about in your terminal

IDE:
Version: 2.xx.x
VSCode Version: 1.105.1
Commit: ......

CLI:
CLI Version 2026.01.17-d239e66

---

For AI issues: which model did you use?
Model name (e.g., Sonnet 4, Tab…)

---

For AI issues: add Request ID with privacy disabled
Request ID: f9a7046a-279b-47e5-ab48-6e8dc12daba1
For Background Agent issues, also post the ID: bc-…

---

Additional Information
Add any other context about the problem here.

---

Does this stop you from using Cursor?

• Yes - Cursor is unusable
• Sometimes - I can sometimes use Cursor
• No - Cursor works, but with this issue

The more details you provide, the easier it is for us to reproduce and fix the issue. Thanks!

[deanrie]

Hey, thanks for the report.

This is a known issue with Azure OpenAI BYOK and code editing features. Cursor sends requests in a format that isn’t compatible with the Azure OpenAI API. It uses the wrong tools schema for file operations. Chat works, but applying code changes fails with this error.

Related threads with the same issue:

• Cursor Agent Mode cannot modify files when using Azure-deployed GPT-5 mini (communication works but no file editing)
• Azure OpenAI Chat Works, but Cursor “Apply” Fails (404) Due to Hardcoded /v1/responses Endpoint Mismatch

The team is aware and is working on improving BYOK support. A few related fixes have already shipped, but the issue still affects GPT-5.2.

I’ll pass this to the team.

[Gil_Maman]

Thanks! Is there any Azure models that are fully supported at the moment?

[deanrie]

Unfortunately, right now no Azure model works fully out of the box for code actions (Apply or Agent Mode) via BYOK.

Current situation:

• Older models (GPT-4o, GPT-4.1): they use the Completions API, but there are still issues with the tools schema
• Newer models (GPT-5, GPT-5.2, GPT-5 mini): they use the Responses API, which Cursor doesn’t support properly for Azure

The team is aware of the issue, and some fixes have already shipped, but the problem still remains for GPT-5.2 and other new models.

Workaround:
There’s a community solution from @gabrii, a proxy service that adapts Azure GPT-5 to work with Cursor: GitHub - gabrii/Cursor-Azure-GPT-5: A service that allows Cursor to use Azure GPT-5 deployments.

It works for GPT-5 and GPT-5 mini, and it requires a Pro or Business subscription for BYOK. Setup steps are in the README.

Let me know if you try the proxy and have any questions.

[Tomasz_Dysko (Tomasz Dyśko)]

@deanrie I don’t know if you understand but not fixing this is directly pushing people into copilot. A lot of teams have Azure resources and budgets there - not in custom SaaSes.

I’d be highly appriciated to assign it higher priority.

[Tomasz_Dysko (Tomasz Dyśko)]

any update?

[mar_on (Marek)]

I am experiencing exactly the same issue. I am also waiting for an update on this matter, as it is a significant blocker for our workflow.

[jka]

Hi, I would also appreciate an update on this. Thanks for the hard work!

[Qi_Shao]

I’m also waiting for this issue to be resolved. Please escalate the priority.

[FrankSaysHi]

Partial good news…

Today I noticed that an Azure deployment of gpt-5.2 works in agent, chat, debug, and plan modes UNTIL you add an image, and then you see an error like this:

```
Request ID: 7fb16581-8d2b-4c9a-b7dd-2eca8507ea89

{“error”:“ERROR_BAD_USER_API_KEY”,“details”:{“title”:“Invalid API key.”,“detail”:“Unauthorized User Openai API key”,“isRetryable”:false,“additionalInfo”:{},“buttons”:[],“planChoices”:[]},“isExpected”:true}

Invalid API key. Unauthorized User Openai API key
```

I am confident the image is the problem, because when I remove the image, the error goes away, and I get a successful chat response.

This is a problem, because without being able to add screenshots we can’t work efficiently.

A few days ago, I thought that I had the same problem as the original poster (not being able to use gpt-5.2 on Azure at all, or maybe I am mistaken), but I do know it’s not happening today. One thing that changed which might be relevant is that today I did temporarily turn on the OpenAI API Key under “API Keys”. I got that working, and then I turned it off in order to use OpenAI on Azure. I did this, because somewhere I had read that it might help.

I can confirm that the tokens are being used on OpenAI on Azure, because at https://cursor.com/dashboard I see this:

The problem is that when attempting to send an image, 0 tokens are sent.

It worked when sending chat messages without an image for both gpt-5-codex and gpt-5 models which I deployed to Azure recently. When it succeeded, I was using the following URL as the Azure OpenAI base URL in Cursor settings: https://XXXXXX.openai.azure.com/but when I used https://xxxxxxxx.cognitiveservices.azure.com/I was able to send a message and not get any response or error message.

I know that it’s using the model I deployed on Azure, because when I changed .com to .co in the base URL I started getting “Network Error: We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.”

It was unexpected to see that the deployment name in the settings did not seem to make a difference. For example, I could enter “gpt-5-codex” or “xyz” and it worked the same.

Does anyone else get similar results?

[FrankSaysHi]

Thanks for the link to the proxy solution,@deanrie.

In Cursor settings, I see many GPT 5.1, 5.2, and 5.3 variants available, but the only GPT5 variant seems to be GPT 5 Mini.

Are you expecting that a stronger GPT 5 variant should be available too? Thank you!