How do I preventing it from stopping??

Resolved 💬 11 comments Opened Sep 23, 2025 by jambudipa Closed Nov 11, 2025
💡 Likely answer: A maintainer (etraut-openai, contributor) responded on this thread — see the highlighted reply below.

I switched to Codex from Claude Code a while ago, but there are some infuriating problems that I am unable to solve.

The main one is that I cannot rely upon it to execute a task to completion. No matter how I prompt it, it seems, it will always sit down eventually and say, “just say the word and I will continue.”

Makes it fairly pointless as an agent, it must be continually supervised.

How do I fix this?

View original on GitHub ↗

11 Comments

FreddyE1982 · 10 months ago

Its not possible for now sadly. It has a maximum turn time limit (= the time it will work for before stopping to ask you stuff) plus it decides on its on how much of that time limit to actually use.

jambudipa · 10 months ago

Is the Rust source publicly available?

ghost · 10 months ago
Is the Rust source publicly available?

you must be kidding me

jambudipa · 10 months ago

No. So no then.

jambudipa · 10 months ago

But why would they do this? Seems utterly bizarre.

So I have to write a script that checks? And just keep reinvoking it?

ghost · 10 months ago
No. So no then.

If you're referring to the tool's open-source Rust code, you're already on the repository that contains it. I'm not sure if they have some guidelines in place to prevent people from adding more features on it such as continues run but you can download it, setup rust, and try to do that. if it works, let me know

ariawisp · 10 months ago

Yeah like 90% of my messages to Codex are "Continue" or "Proceed" lol

kidandcat · 9 months ago

Welcome to open source, were you can just adapt the code to your needs:
I have a fork with background process management like Claude Code, and a continue mode (Ctrl-G).

https://github.com/kidandcat/codex

etraut-openai contributor · 8 months ago

Thanks for the bug report. We've shared the feedback with the team members who are responsible for training our codex models. We're tracking this feedback and similar reports so we can make the model more robust in future iterations.

If you see other model behavior that you'd like to report, please use the recently-added /feedback command in the CLI. That will provide us with additional details including logs and conversation text.

tiagosiebler · 8 months ago

For what it's worth, I got frustrated to the point to modifying codex to address the behaviour, though I think it's more of a workaround. I'll try to keep the branch on my fork in sync with the latest changes for as long as I'm using it, but here it is for anyone that wants to see how I did it: https://github.com/openai/codex/pull/6197

It'll just auto continue if it has more to do. It's not perfect but it's a huge improvement, for me.

BruceBGordon · 5 months ago

totally not fixed. I'm running into these exact errors.