Rate limits

How many calls a key may make, and how to read where you stand.

For AIDownload .md All docs as one Markdown file llms.txt (index)llms-full.txt (all docs)OpenAPI (JSON)

Calls are counted per key and per kind of call, per minute, and there is an overall limit per key too. Every answer says where you stand.

Kind of call	Name	Per minute
`Read one record`	get	600
`List records`	list	240
`Change a record`	update	240
`Search records`	search	120
`Create a record`	create	120
`Key information`	me	60
`List properties`	properties	60

Headers

X-RateLimit-Limit: 240
X-RateLimit-Remaining: 238
X-RateLimit-Reset: 1794313700

Header	Meaning
`X-RateLimit-Limit`	Calls allowed in the window for this kind of call.
`X-RateLimit-Remaining`	Calls left in the window (the smaller of the kind's and the key's overall).
`X-RateLimit-Reset`	When the window resets, in seconds.
`Retry-After`	On 429: seconds to wait before trying again.

When the limit is passed

HTTP/1.1 429 Too Many Requests
Retry-After: 20

{ "statusCode": 429, "message": "Too many requests: slow down" }

Wait for Retry-After, then try again. Back off for longer each time a retry fails. A program that loops over records should read X-RateLimit-Remaining and slow itself down before it hits 429.