OpenAI's status page recorded a full outage of Codex late on September 25, lasting about 56 minutes across Codex Web, the API, the CLI and the VS Code extension. Service has recovered, and paid users were reportedly promised a usage-limit reset. For stores running work through agents, it is a prompt to plan for downtime.
Cloud sessions are officially available and out of research preview! They let you keep Claude Code working, even when your laptop is closed. Existing subscribers get a one-time credit to try them: $100 on Pro, $250 on Max. https://t.co/SvdB2ATkFw
— ClaudeDevs (@ClaudeDevs) · View on X
- What happened: OpenAI's status page lists "Issues with Codex" as a full outage from 10:58 PM to 11:54 PM on September 25, 2026, affecting Codex Web, the Codex API, the CLI and the VS Code extension (OpenAI status). The page does not state a time zone or a cause, and says all impacted services "have now fully recovered."
- What has been said about compensation: According to KuCoin, Tibo, head of Codex, said: "All paid users affected by this brief outage will have their usage limits reset." The official status page does not mention a reset.
- Not the only incident this month: OpenAI's status history lists other September incidents involving Codex, including "Elevated errors across ChatGPT and Codex" on September 3, "Codex GitHub Review and Pull Request Failures" on September 13 and "Elevated error rates for Codex and ChatGPT Work" on September 14. On September 3, 9to5Mac also reported disruptions at Claude and Grok. Unite.AI puts Codex uptime at 99.98% for June to September, so total downtime is small, but interruptions do happen.
- Why it matters more now: More work runs without anyone watching. Claude Code cloud sessions are now generally available, so agents keep working with the laptop closed (@ClaudeDevs). DigitalOcean's Managed Agents run Claude Code or Codex in the cloud (@digitalocean). @thekuchh described an agent that ran a Shopify store for about eight weeks while a person only approved spend, a self-reported case (@thekuchh). An outage on the API and CLI can stop background jobs like these, not just a chat window.
- What to do next: List the jobs your agents handle. Mark which can wait an hour, such as catalog edits and content drafts, and which cannot, such as customer replies or anything tied to live ads or orders. Keep a manual procedure for the second group, and after any outage, check jobs that were running for half-finished changes.
Did the outage affect Shopify stores directly?
Not directly. Codex is a coding and automation tool, so the outage affects workflows that call it rather than Shopify storefronts or checkout. There are no public reports so far of merchants harmed by the September 25 incident.
Will affected users be compensated?
According to KuCoin's report of Tibo's statement, paid users affected by the outage will have their usage limits reset. The status page does not mention this, and no other compensation has been reported.
Should I switch AI providers to avoid outages?
Switching alone may not help. On September 3, several providers had problems on the same day, according to 9to5Mac. A backup plan for the jobs that cannot wait matters more than which provider you use.