Auto Resume /Goal after limit reached once limit is reset (5h)

Open 💬 3 comments Opened Jun 18, 2026 by frank-bluemke

💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What variant of Codex are you using?

App

What feature would you like to see?

As a codex app user, when running the goal command on a lengthy body of work, after the 5 hour limit or weekly limit is reached and codex stops, it should auto resume (or provide an option to enable this) the goal so that you can issue a goal and have it complete the goal without having to manually resume it once your limit window is over.

Additional information

_No response_

View original on GitHub ↗

3 Comments

github-actions[bot] contributor · 2 months ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

#28897
#27590

Powered by Codex Action

edgebyte-ai · 2 days ago

This should also apply to regular Codex App conversations, not only /Goal runs.

When an active conversation stops because the shared five-hour usage window is exhausted, Codex already knows or displays when usage will become available again. Please provide an opt-in option such as "Automatically continue after usage resets."

Expected behavior:

Preserve the current conversation, pending objective, and task context.
Mark the conversation as waiting for quota and show the scheduled resume time.
At or after the server-provided reset time, re-check availability and enqueue exactly one continuation turn.
Keep normal approval and safety boundaries; do not automatically proceed when user input or approval is required.
Let the user cancel the scheduled continuation before it runs.

This would allow an unattended conversation to continue after the cooldown instead of requiring the user to return roughly five hours later and manually type "continue."

torba-lab · 20 hours ago

One additional concern: the 5-hour usage window is exhausted extremely quickly for GPT-5.6 Sol and Daybreak Blue at higher reasoning levels such as Extra High, Max, and Ultra - in my case sometimes within only 15–25 minutes of intensive agent work.

For long-running research tasks, this means a single task often has to be resumed across 3-4 separate 5-hour windows. Each long interruption also hurts exploratory momentum: after resuming, Codex tends to reconstruct the task from persisted context rather than continuing the same deep search trajectory, which can noticeably reduce the quality of the investigation.

I would strongly suggest removing the 5-hour window entirely for Extra High / Max / Ultra, and enforcing only the overall weekly allowance for those modes. That would let the same total quota be used continuously when needed, instead of artificially fragmenting one deep task across many hours.