Add the ability to fallback to platform.openai.com credits when usage runs out.

Resolved 💬 2 comments Opened May 4, 2026 by divyamagrawal06 Closed May 4, 2026

What variant of Codex are you using?

CLI, App

What feature would you like to see?

/add-api-key while being signed in with a ChatGPT account.

  • Fallback API key in cred store
  • User uses codex as normal
  • When limits run out, just switches to API key
  • This feature can be extended to the App and it benefits app users as well.

Why?

  • People using codex with an API key will likely have a ChatGPT account that they would prefer using before using API.
  • This is a big pain point as it means having to constantly log out and log in. Often, it means having to generate a new key as you would copy/paste the key only once.

Additional information

This was rejected previously for the following reason:

I think it's unlikely that we would implement a solution like the one you've outlined above. We don't recommend switching between accounts, so this isn't something we'd want to encourage. If you need to purchase additional usage when you run low on your ChatGPT subscription, you can purchase additional credits that are associated with your main account. For details, refer to this documentation. Using this approach, you don't need to switch accounts.

_Originally posted by @etraut-openai in #20899_

But my original request is about a different scenario: someone having prepaid API credits on platform.openai.com (tied to the same OpenAI/ChatGPT account) and wants to use that when the sub limit is hit, (without logging out.)

Currently the code explicitly blocks adding an API key while ChatGPT auth is active. So even though both billing systems share the same underlying account, there is no way to use API platform credits as a fallback. The user must log out, switch to API key mode, and lose their ChatGPT session. I have tried this and there is no way to fallback to OpenAI platform credits.

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗