Usage limits and slowness

Kaivo works like the tools you may already know: no credits to count, but a usage limit per period. There are two: a limit per 5-hour session, and a limit per week. They reset on their own.

The unit is one request to the model. A task needs several: reading a file, searching the code, changing, checking… Tiny requests (like a conversation title) don't count. Your plan sets the size of the limits.

To see where you stand, type /usage in Kaivo: a window shows one bar per limit, with the percentage used and when it resets. The Account page of the website shows the same thing.

When a limit is reached, Kaivo tells you and gives the time it resets. You can wait, or keep going right away with your own API key: see “Add your own key (/custom-key)”.

The model may also be slow or very busy. Kaivo then spreads the load over several servers and retries on its own. If you see “Retrying due to rate limit”, that's this mechanism at work: wait a moment.