Limits keep one busy integration from slowing down everyone else. Each site has its own limits for each caller, so traffic to one site never uses up the limits of another.
Every request to a site, at site/{site_id}/, counts toward all three of these limits:
Going over a limit blocks further requests to that site for 1 minute, or for 1 hour when it is the hourly limit. A blocked request gets HTTP 429 and this response:
{
"success": false,
"errors": [
"Please wait before trying again."
]
}
All requests from one IP address, to any site, also share a limit of 5,000 a minute. Going over blocks that address for 3 minutes, with HTTP 429 and the message The request limit has been exceeded. Please wait before sending more.
The Get and List endpoints have one more limit: 10 requests in 5 minutes to the same URL, for each site and caller. The query string is not part of the URL here, so every page of a list counts toward the same 10. Going over blocks that URL for 5 minutes, with the message Access to this resource has been throttled. Please wait before trying again.
Asking for a record that does not exist counts as two requests toward every limit above.
Get Payment works differently: it has neither of these rules, and instead each payment can be read once every 55 seconds. Each id you read it by counts on its own. Reading the same payment again sooner returns HTTP 400 with success false and the message You have hit the throttle limit for this payment, please try again in a minute
A request body can be up to 1 MB of JSON. A larger body gets HTTP 413 and Request body too large. Sending it again fails the same way, so split it into smaller requests.