Cooldown for ChatGPT

Shows a live countdown from the Retry-After header when ChatGPT returns HTTP 429. Not affiliated with OpenAI.

When ChatGPT says "Could not load this ChatGPT conversation" or a request quietly fails, the reason is often rate limiting: the server answered with HTTP 429 and a Retry-After header telling the browser how long to wait. The ChatGPT web app doesn't show that number. Cooldown for ChatGPT does. What it does • Shows a small panel with a live countdown for each rate-limited request. It uses the same time ChatGPT itself uses to block its Retry button, so you know when Retry can work at all • Turns green when retrying is allowed, and flags 429 responses that came without a wait time • When ChatGPT shows its conversation load error, reports what was actually observed for that conversation: a wait still running or over, a page that never requested the chat, a network-level failure, a request that succeeded, or nothing recorded. A 429 from before a reload is still mentioned • Shows a 429 badge on the toolbar icon while a limit is active • Keeps a short history of recent 429s and network errors in the toolbar popup • Optional auto reload (off by default): once the wait is over, reloads the tab that couldn't load the chat. It never reloads early, never touches other tabs, waits while you have unsent text, waits longer before each further reload, and stops after 3 reloads in 10 minutes. • English and Korean interface (English by default, switch in the popup) What it doesn't do • It only reads response status and headers. It never blocks or modifies requests, never sends requests of its own, and doesn't try to get around limits. • It doesn't read your messages or ChatGPT's replies. On the page it only looks for ChatGPT's error text and, before an auto reload, whether the message box is empty; nothing from the page is stored or sent. • It collects nothing and sends nothing anywhere. Observations live in memory and are cleared when the browser restarts. Good to know • Waiting out the countdown doesn't always fix the error. The server can answer with a new 429 after the wait, ChatGPT sometimes doesn't request the chat at all after a reload, and the error can also come from network failures or other server-side problems. The panel shows what was actually observed, so you can tell these cases apart instead of waiting blindly. Open source, and contributions are welcome. Bug reports, ideas and pull requests: https://github.com/snowmuffin/cooldown-for-chatgpt Independent project. Not affiliated with or endorsed by OpenAI. ChatGPT is a trademark of OpenAI.