# Limits

Credits, rate limits and Kelly Wire limits per plan, and what happens when you reach one.

Every call is metered in [credits](/glossary#credit), the same on every surface, and each plan sets an allowance, a rate limit and Kelly Wire limits. Nothing is charged during the beta.

## Plans

| Plan | Credits | Per | Rate limit | Wire connections | Wire topics |
| --- | --- | --- | --- | --- | --- |
| Anonymous | 200 | UTC day | 60 calls per 600 seconds, per tool | 2 | 20 |
| Free | 6,000 | billing month | 120 calls per 60 seconds | 3 | 50 |
| Pro | 25,000 | billing month | 300 calls per 60 seconds | 5 | 200 |
| Developer | 100,000 | billing month | 600 calls per 60 seconds | 10 | 500 |
| Scale | 1,000,000 | billing month | 3000 calls per 60 seconds | 100 | 5,000 |

Anonymous allowances are per address per UTC day; account allowances are per billing period. The numbers are proposals until traffic sets them.

## Prices per operation

| Operation | Price |
| --- | --- |
| `GET /health` | [object Object] credits |
| `GET /v1/status` | [object Object] credits |
| `GET /v1/reference` | [object Object] credits |
| `GET /v1/reference/{topic_id}` | [object Object] credits |
| `POST /v1/wire/tickets` | [object Object] credits |
| `GET /v1/stream/{topic}` | [object Object] credits |
| `POST /v1/answers` | [object Object] credits |

A refused request costs nothing.

## Reaching a limit

A call over the rate limit or past the allowance answers `429` with `rate_limited` or `quota_exhausted`, and the rate-limit headers say when to try again. See [retries](/policies/retries).

## Kelly Wire

Per address, at most 5 open connections and 30 connection attempts per 60 seconds. A frame is at most 65,536 bytes, and a client may send 20 messages per second.

---

Last updated October 16, 2026

Support: support@askkelly.ai · Security and disclosure: https://docs.askkelly.ai/security

Information is the edge. Odds are information, not advice · 21+ · Get help