Rate limits
The requests a caller may send, with a key and without one, and how a 429 reads.
Every request pays a token bucket: a rate it refills at, and a burst it can spend at once.
| Caller | Per second | Burst | Counted by |
|---|---|---|---|
| No key | 20 | 60 | Your client address |
| API key | 30 | 120 | Your key |
| API key, high class | 100 | 400 | Your key |
Without a key, every caller behind one address shares that address's bucket. A key's limits are its own, whatever addresses its requests come from. A key's class is set when it is issued.
Without a key, token images (GET /v1/tokens/{address}/image) pay a bucket of their own: 50 a second, burst 400.
Routes with an allowance of their own
These routes also give each caller an allowance: each key, or each client address for a request without one.
| Route | Allowance for each caller |
|---|---|
A wallet's swaps (GET /v1/wallets/{address}/swaps) | 1 page a second, burst 20 |
GET /v1/stream | 4 open streams |
A high-class key has three times each of these.
Their 429 answers rate-limited with Retry-After and no bucket.
Reading your limits
GET /v1/limits lists every bucket your requests pay and what each has left: Data API.
It charges only its own request.
It does not list a route's own allowance.
Every answer carries two headers for the bucket nearest its end:
ratelimit-policy: "ip";q=60;w=3
ratelimit: "ip";r=59;t=1q is the burst, w the seconds to refill from empty, r the requests left at once, and t the seconds until the bucket is full.
A 429
{"error":"rate limited","code":"rate-limited","retryable":true,"retryAfterSeconds":1,"bucket":"ip"}Wait the Retry-After seconds, then send the request again.
Pace a client by RateLimit before it meets a 429.