Rate limits
Limits protect the service and grow automatically as your workspace tops up its balance.
Base URLhttps://api.kalle.build/v1AuthBearer kalle_sk_…VersionKalle-Version: 2026-11-01
Usage tiers
| Tier | Reached when | Requests per minute | Input tokens per minute | Output tokens per minute |
|---|---|---|---|---|
| 1 | Any prepaid balance | 50 | 40,000 | 8,000 |
| 2 | €100 topped up in total | 1,000 | 400,000 | 80,000 |
| 3 | €500 topped up in total | 2,000 | 1,000,000 | 200,000 |
| 4 | €1,000 topped up in total | 4,000 | 4,000,000 | 800,000 |
| Enterprise | By agreement | Custom, tier 4 until agreed | Custom | Custom |
Limits apply to each key, in one-minute windows. A message counts its estimated input when it starts and its real input and output when it ends. A single request larger than the input limit still runs when it is the first of its minute. Counting tokens is not limited.
Rate limit headers
| Header | Meaning |
|---|---|
kalle-ratelimit-requests-remaining | Requests left in the current minute |
kalle-ratelimit-input-tokens-remaining | Input tokens left in the current minute |
kalle-ratelimit-output-tokens-remaining | Output tokens left in the current minute |
kalle-ratelimit-reset | When the minute ends, as an RFC 3339 time |
retry-after | Seconds to wait after a 429 |