Source: https://spooly.eu/en/docs/limits

# Limits

The limits stop automations that run in circles. For comparison: the Home Assistant integration needs about 1,450 calls a day.

## Values

| Limit | Value | Code |
| --- | --- | --- |
| Calls | 10,000 a day per account, across all connections combined. The day ends at midnight (UTC). | `fair_use` |
| At once | 4 requests per account. There are also limits that all accounts share: 12 requests together, and during the key check 16 together and 4 per IP address. Above that, the API answers right away with Retry-After: 1. | `too_many_parallel` |
| Request size | At most 64 KB for `/api/v1` and `/api/spoolman`. Spooly rejects larger requests with 413 without reading them. | no code, see [error codes](https://spooly.eu/en/docs/errors) |
| Changes per minute | 60 per connection | `change_rate_limit` |
| Changes per day | 500 per connection | `change_daily_limit` |
| New spools per day | 100 per connection | `new_spool_daily_limit` |
| New keys | 30 per hour per account | `too_many_keys` |

## What counts

- Every valid request counts as a call, including the one an AI uses to fetch its tool list.
- Failed changes do not count against the change limits.
- `GET /me` shows today’s count in `calls_today` and `daily_limit`.

## How your program should react

All limits answer with 429. The wait in seconds is in the `Retry-After` header. Wait at least that long, then retry. With `fair_use`, it only continues after midnight (UTC).

`too_many_parallel` can also hit you when others are sending a lot right now, even if you send little yourself. Wait briefly and repeat the request.

Do not let an automation write on every reading. A scale that reports ten times a second needs a condition that waits for a steady value.