Codex rate limits and 429s: see headroom before the failure
A 429 is rarely the real problem. The real problem is finding out you were near the limit only after the session breaks. The Codex CLI tells you afterwards. CodexUse shows which account is getting hot, so you keep a second account ready and Open it before the interruption lands.
What a 429 really means
A 429 means the active account crossed a rate-limit window. Sometimes recovery is quick. Sometimes the wait is long enough to end the session.
Error: 429 Too Many Requests Rate limit exceeded. Please try again later.
The important point is not the exact wording. It is that the signal arrives after flow already broke.
Why the native flow is reactive
- You usually do not see a clean per-account headroom view before the failure.
- You may know you have another account, but not which one is actually safe right now.
- If you move to another account only after the error, the interruption already happened.
What the CodexUse trial adds
- Headroom: the percent left in each account's 5-hour, 7-day, and 30-day windows, so you see which accounts show Ready and which show Low.
- Refill timers: read the refill time reported by Codex. The refill controls attempt a short background request; they do not accelerate a refill or add quota.
- Auto-roll: when the app is idle, move to a healthier account automatically once the active one drops below your Roll below threshold.
Fastest preflight
codexuse profile list
codexuse run --profile Reserve -- "run the full test suite"Check each account's headroom first, then start the heavy run under the account with the most room. In the app, turn on Auto-roll to make that move for you.
When moving to another account helps and when it does not
- Helps: the next account is a different OpenAI account with more headroom.
- Does not help: you saved the same OpenAI account twice under two names.
- Still useful: even if you wait instead of moving, headroom tells you that early enough to plan around it.
When every account you hold is hot at once, there is nothing to move to. A custom-model account runs a non-OpenAI model in the Codex app on your own provider credential and consumes no OpenAI quota. It does not raise the limit on your OpenAI accounts — it moves work off them, and you pay that provider directly.
When there is no other Codex account
Moving work to a second account assumes you have one. If you do not, a custom-model account is the other route, paid by the token to a provider rather than drawn from any allowance. Your options when an account runs dry sets out the costs.
Troubleshooting table
| Symptom | Likely cause | Action |
|---|---|---|
| You Open another account and still get 429s | The new account is the same OpenAI account or is already hot too | Check whether the destination is a different account with real headroom |
| Headroom looked fine but the session still failed later | Usage changed during the run or the estimate moved | Keep reserve capacity and treat headroom as operational guidance, not an absolute guarantee |
| Auto-roll did not roll | The app was not idle or the threshold was not crossed | Open the other account yourself or tune the watch interval and Roll below threshold |
Operational habits that help
- Keep one lower-usage reserve account for long refactors or urgent fixes.
- Check headroom before the next heavy run, not midway through it.
- Use Auto-roll as a guardrail, not as a substitute for naming and maintaining accounts cleanly.
Related
Why am I getting Codex 429 errors?
A 429 means the current account crossed a rate-limit window. The main problem is that the native CLI usually tells you only after the failure.
Can moving to another account help with Codex rate limits?
Yes, if the other account is a different OpenAI account with its own headroom. Moving between two saved entries for the same account does not create new capacity.
What does the CodexUse trial add here?
CodexUse shows per-account headroom, refill timers, and Auto-roll during the trial. Pro keeps that proactive workflow unlocked after the trial.