Quota Reset Watch — public reset announcements with primary-source history

[Normr0i33aW1vis144ingo]

I maintain Quota Reset Watch, an independent public reset tracker for Codex and Claude.

It helps answer: “Was an extra reset announced, who was it for, and what did the original source actually confirm?” It does not read your account usage or request Codex credentials.

The site includes:

• Original announcement links, affected scope, and separate states for announced, rolling out and completed resets.
• A distinction between applied public resets and banked reset opportunities.
• Searchable history with UTC/local time zones, source health and last-success times.
• 24/48-hour historical estimates, comparable samples and previously saved forecasts for later evaluation.
• Free verified-email alerts: choose Codex or Claude, a 24/48-hour window and a 1–100% threshold, with an independent switch for confirmed public resets. No password is required.
• Public RSS, a read-only API and a remote MCP endpoint.

The forecasts use a small historical sample and are not an official schedule. Insufficient samples or stale monitoring data suppress probability alerts. A continuing high estimate does not produce an email every five minutes: re-alerting requires a new threshold crossing and at least 24 hours since the preceding forecast alert for that product. Claude may not yet have enough comparable history for an estimate.

Set up email alerts · API / MCP documentation

Useful guides, checked September 19, 2026:

• Compare free reset trackers by history, alerts and personal timers — includes this site and identifies us as the publisher; links to each tool’s own feature page.
• Why reset predictions differ and how to assess a forecast
• Email setup, threshold examples and missing-email checks
• Banked resets, expiry and changes to the weekly reset date

Public event metadata and corrections are available on GitHub. I am the project maintainer; the site is not affiliated with OpenAI or Anthropic. Feedback on confusing reset terminology or missing primary-source announcements would be useful.