We have to open Codex App to load chats in IDE extension
Resolved 💬 14 comments Opened Apr 20, 2026 by topemalheiro Closed Apr 30, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of the IDE extension are you using?
latest
What subscription do you have?
Business
Which IDE are you using?
VS Code
What platform is your computer?
Microsoft Windows NT 10.0.22631.0 x64
What issue are you seeing?
We have to open the Codex App to load chats into the extension, which didn't happen before.
What steps can reproduce the bug?
Open VS Code with Codex without opening Codex app first.
What is the expected behavior?
Open Codex extension and all chats load.
Additional information
_No response_
14 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Can you describe the behavior you're seeing? Do the chats just not show up in the IDE Extension? Or do they show up but fail to load when you click on them?
I click on a chat, and I get in, but the chat is blank. Only when I open the Codex app and select the same chat/or have that chat open when the Codex app opens (if not, I have to open said chat there) does the chat load in VS Code.
Pay attention to the computer where you can reproduce my case.
This never happened to me before, even more so because I didn't even have the Codex app installed at the start; perhaps it's the Codex app that causes the bug itself at most.
Like this:
<img width="1899" height="929" alt="Image" src="https://github.com/user-attachments/assets/5b7b1806-748a-47f7-a7a3-1e202d05efc0" />
Happened again, correction, I have to press in an existing chat in the Codex app to load the one I have on VS Code. Doesn't seem to have a set behaviour, perhaps.
<img width="312" height="92" alt="Image" src="https://github.com/user-attachments/assets/cbdc7efb-cd8c-497c-9314-e16c6e1684f0" />
<img width="1882" height="874" alt="Image" src="https://github.com/user-attachments/assets/a39605f8-25aa-4b2b-90f0-9b69d123ab91" />
For some reason I'm having a somewhat similiar issue, although for me, sometimes even the Codex App won't load the chat. And if it will, and somehow I manage to send the prompt, it often keeps thinking... and thinking... and thinking...
It seems that resetting usage for everyone at the same time so last day they spend it all, wasn't such a great idea if it's an issue related to "too much usage"
I've also been noticing that sometimes sending prompts often fails, and the send button stays with the throbber.
<img width="1880" height="123" alt="Image" src="https://github.com/user-attachments/assets/4d76f7a3-55a5-4d23-bce5-4d6013991307" />
This was the same chat I had to open the Codex app so it would appear. I haven't done anything since then.
Hi, after the last update Codex is completely broken on VSCode. Opening even average-length (10 messages) conversations results in a blank page. I am stopping usage of Codex until this is solved, the extension is in a completely unusable state.
Yeah, I get you. Sometimes, it feels like companies go out of their way to make the experience worse. I'm still wondering why they hid rate limit metering in a dropdown inside a dropdown, sounds like a shortsighted, useless design decision (like we even care for design).
In this case I am not even lodging an enshittification complaint (even though there's plenty of that going on), it's that someone at OpenAI must have pushed something that can only be described as completely broken.
Yeah, this may have to do with them trying things out. But for sure, if this is related to new stuff they're implementing, they shouldn't have pushed this version since it's broken.
It looks like they are (only) debugging in prod lol.
that's IMO the result of uncontrolled agent coding in production, and things generally are just going to be worse (in general, not on this repo specifically). Agent programming gives you urgency to not review everything it has written to not feel behind, because humans in the loop can be seen as bottleneck, but agent coding quality is not there, yet.
They should've tested it themselves, I don't know why they didn't, regardless of being coded by AI agents or not IMO.
This is still happening.
It's solved.