Skip to main content
Every read has a cache window, listed in the rate limits table. That window is the whole story of how to poll.

​
Poll on the window, not faster

A read is cached at the edge for its window: 5 seconds for a coin, 10 for the coin list, 30 for markets, 60 for pairs and the index. Inside the window every call returns the same answer, so polling faster than it buys you nothing. Polling on the window gives you every change there is. Repeats inside the window are served by the edge and don’t count against your limit. Only the call that reaches us counts.

​
Browser or server

From your users’ browsers. Cross-origin requests are allowed from any site, so call the API straight from the page. Each visitor has their own limit and the edge absorbs the repeats, so this scales with your traffic. You don’t need a proxy. From your server. Everything you send shares one address and one limit. Cache on your side and poll on the window. If a legitimate job needs more than the limit, ask and we raise it for that address.

​
Which read for which job

Don’t walk /coins/{mint} for every coin to build a list. /coins returns the same headline figures fifty at a time, and /index returns every mint in one call.

​
Read the 400

A 400 names the parameter that’s wrong and lists the values that work. Read it rather than guessing again: sorts are mcap, earnings, burned, fees, buyback, active, holders, reserve, paid and newest; the coin list caps limit at 50; /launches has no sort.