Is Codex down right now?

What to do with this result

An incident can explain failures affecting several users. An operational report does not prove your account, network or session is working: reports can lag, and local problems may never appear here.

• Reported disruption: read the incident’s affected features and updates before changing your configuration. Preserve your work and try again after the official resolution.
• No disruption reported: check the exact error. A disconnected stream, login failure and exhausted allowance need different next steps.
• Unknown or stale: open the official source directly. Missing data is not evidence that Codex is healthy.

Follow the outage-versus-local-error checklist →

Where the data comes from

We read OpenAI’s public component feed and incident feed hourly. We also check incident details for component impacts, including events whose titles do not name Codex. Failed checks retry no more often than every 15 minutes. Successful records are retained in our database; this page shows records first reported within the past 30 days and any unresolved records. The incident card calculates days since the official first-report time in your browser; the last-updated timestamp below refers to the most recent successful source check, not a new incident.

Missing or stale component data is shown as unknown. Reporting timestamps are not necessarily a measurement of the actual outage duration. We do not calculate an uptime percentage from them.

Need to distinguish an incident from an allowance refresh? Compare the public reset signals for today with the personal reset guide. The six reset scenarios are listed there; none can read your private account.

Is a usage limit a service outage?

Usually these require separate checks. A message saying you reached a usage limit concerns your allowance; use your account’s usage view. Public service status cannot read your balance or reset time.