OpenAI API Key and Override OpenAI Base URL can't be modified

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

OpenAI API Key and Override OpenAI Base URL can’t be modified

Steps to Reproduce

Open Models settings, can input values.

Expected Behavior

Values can be inserted.

Screenshots / Screen Recordings

Operating System

Linux

Version Information

Version: 3.15.6
VS Code Extension API: 1.128.0
Commit: a1f686545fd0ce8917bbd2449f733551a9bce420
Date: 2026-08-06T01:41:03.876Z
Layout: IDE
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.291
OS: Linux x64 7.1.7-200.fc44.x86_64

Does this stop you from using Cursor

No - Cursor works, but with this issue

Hey, thanks for the report and the screenshot. I can see the Models settings section with the OpenAI API Key field.

This is a known issue in 3.15.x. The API Key and Override Base URL fields don’t get focus when you click them. It’s fixed in 3.15.20, so please update Cursor via Help > Check for Updates, or reinstall the latest build, and the fields should work again.

If you can’t update yet, here’s a workaround. Click next to the field and press Tab to move focus into the input, then type or paste the value as usual.

Let me know if the issue still happens after updating.

Thanks! This trick works. I have another question, do you know if Tab complete / Auto complete will be using my OpenAI? What I’m trying to do is to use our AWS Bedrock. I have managed to use it with agents, not sure about completion.

I have used help of this GitHub - bakwc/bedrock-openai-proxy: Proxy server to use AWS Bedrock models with OpenAI-compatible API. · GitHub

thanks!

Glad the trick worked.

On the second question: no, Tab or autocomplete doesn’t use a custom OpenAI key or Override Base URL. The BYOK setup your own API key and custom endpoint only applies to Chat and Agent, so that’s why you were able to connect Bedrock via a proxy for agents.

Tab runs on Cursor’s own dedicated model, and it can’t be redirected to a custom OpenAI-compatible endpoint. So you won’t be able to run completions through an AWS Bedrock proxy. That’s by design.

If anything else comes up, just message me.