The quick answer
If Codex cannot start your next task, first check the account, sign-in method and exhausted window. A new conversation does not replenish account allowance. These are diagnostic scenarios, not verbatim error messages.
1. Confirm which account is being charged
Open the profile menu to check the account and workspace; in the CLI, run codex login status. Authentication documentation distinguishes ChatGPT subscription access from API-key billing.
- ChatGPT sign-in: open the Codex usage dashboard for the same account and workspace.
- API key: inspect the API organization's usage, billing and the returned error. A subscription reset timestamp does not diagnose an API limit.
- The two clients disagree: confirm that both use the same authentication route before comparing their balances.
2. Identify the constraint
Use the official usage dashboard to check remaining allowance and reset times. The information available through /status depends on the client; session context is not plan allowance. Usage guidance · CLI commands
- Five-hour allowance is exhausted: record its reset time and check the weekly balance.
- Weekly allowance is exhausted too: the earlier five-hour reset is not enough. Follow the reset-time guide.
- Allowance remains, but the task fails: retain the exact message and identify whether it concerns authentication, a tool or a service error; do not invent a quota recovery time.
ChatGPT Work and Codex share usage, so include concurrent Work activity in the investigation. Official pricing
3. Preserve progress and choose a continuation route
Save existing changes, test results and unfinished steps. The official pricing page says an active turn may continue after reaching a limit, subject to fair-use limits; a zero balance does not always mean an immediate interruption.
If waiting is possible, use the displayed recovery time. Plus and Pro users can purchase additional credits; Business, Edu and Enterprise workspaces with flexible pricing can purchase workspace credits. Check eligibility, balance and charges before continuing. This extends eligible use; it is not a promise that the included quota bar resets.
An API key offers separately billed local work where supported. Check feature availability and workspace policy first; it does not refill the subscription allowance. Switching models may change future consumption but does not restore spent quota.
4. Confirm the result
After the stated reset or account action, refresh the official dashboard and check the blocking window again. Then try the smallest needed next step. A timer expiring on this site is only a reminder to verify, not proof that OpenAI has restored access.
If allowance disappears after a few messages, review models, reasoning, tools, caching and subagents. Plan estimates and public reset history are on the Codex page.
Why is Codex still limited after a reset?
First distinguish a scheduled window reset from redeeming an earned reset. If only one window has recovered, another exhausted window can still block the next task. Compare the same account and each displayed window before trying again.
For an earned reset, check the account result before attempting another redemption. The official reset protocol distinguishes these cases:
- No earned reset available: there is no reward to redeem.
- No eligible window: the service has no window that this action can reset.
- Redemption completed: refresh account limits to see the actual result.
These are diagnostic situations, not verbatim client messages. A public announcement or a saved countdown cannot tell you which case applies. Follow the reset-time and redemption guide; if allowance is available but the task still fails, investigate the actual error rather than repeatedly redeeming rewards.
If the message includes 429 Too Many Requests or exceeded retry limit, use the Codex error troubleshooting guide to distinguish temporary throttling, API billing restrictions and connection failures.