OpenAI's Codex coding agent has been experiencing degraded performance since approximately November 1, according to user reports and third-party status monitors. Glitchwire confirmed the issues firsthand on November 4, encountering unresponsive servers and errors when attempting to run prompts.
StatusGator, which tracks OpenAI's official status page, recorded 58 user-submitted outage reports in the past 24 hours as of early November 4. Reports came from users in Pennsylvania, Wisconsin, South Carolina, Belgium, and New York. Complaints included "Server not responding," login failures for Pro accounts, and JSON error messages reading "server error encountered."
GitHub Issues Detail Scope of Failures
A GitHub issue filed November 4 describes the outage in blunt terms: a ChatGPT Pro subscriber in Europe could not start new Work conversations, continue existing ones, or submit work for more than 48 hours straight. The user sent HAR traces, screenshots, and request IDs to OpenAI Support on November 3.
According to that report, an AI Support Agent responded on November 4 with findings from server logs. OpenAI's automated review identified that a scheduled task had stopped during startup because a configuration file in the hosted Work execution environment could not be parsed. The response stated that no repair had been applied at the time of review.
A related issue documented at least eight failed scheduled-task runs during the incident window. Interactive access resumed on November 4, per user observation, though the root cause remained unresolved at the time of posting.
A Pattern of Codex Instability
The November outage is the latest in a string of Codex reliability problems. OpenAI's status history shows incidents including "Degraded performance in ChatGPT, Codex, and APIs" on October 21, "Degraded performance in GPT-6.1 Sol Codex and Work mode" on October 13, and a September 25 outage that prompted OpenAI to reset usage limits for all paid users.
Codex has grown rapidly since its desktop app launched in February 2026. By June, OpenAI reported more than 5 million weekly active users. An August post from OpenAI employee Tibo Sottiaux cited 20 million active users, though without a defined time window. The agent merged into the ChatGPT desktop app on July 9, and by late July, OpenAI said Codex and ChatGPT Work combined had crossed 10 million weekly users.
That scale makes reliability incidents more consequential. Codex runs each task in a sandboxed cloud environment, which isolates execution but also means outages can strand users mid-workflow with no local fallback. Developers relying on Codex for production-grade work have reported losing hours of work and burning through usage quotas on failed attempts.
OpenAI's official status page had not posted an incident notice for the November 1–4 period as of this writing. StatusGator noted that previous outages, including a server-not-responding incident on October 23 and a tools-unavailable incident on October 20, went unacknowledged on OpenAI's page. The company's AI agent infrastructure has drawn scrutiny this year for other reasons, but the more mundane question of uptime may matter more to the developers who depend on it daily.
OpenAI did not respond to a request for comment.
