Rate limits
The Kenso API limits calls two ways: per workspace, by the kind of call, and per operation, for each API key. A call has to fit both.
Kenso limits API key calls per workspace: per minute by the kind of call, and per day for all calls together. The limits depend on your workspace's plan, and Kenso can set other limits for a workspace.
GET), per minutePOST, PATCH, PUT, DELETE), per minutebulk-create, and imports), calls per minuteThe Free plan is on Free, the Premium, Done-for-you and Performance plans are on Elevated, and every other plan on Standard.
Every response reports the rate limit the request is closest to reaching, whether that is your plan's limit, your API key's own limit or a limit on the operation: RateLimit-Limit (the limit), RateLimit-Remaining (the requests left) and RateLimit-Reset (the seconds until it resets). A 429 from one of these limits carries the same three headers, with RateLimit-Remaining: 0, and Retry-After. Over your plan's limit, the response is 429 rate_limited, and its details names the class, limit and window you hit. A 429 quota_exceeded is a usage quota, not a rate limit: wait the Retry-After seconds. Spread bulk work over the day instead of sending it in bursts.
Responses to an API key also carry X-RateLimit-Remaining-Minute and X-RateLimit-Remaining-Day, your key's own counters. They are deprecated: read the RateLimit-* headers instead.
Each operation also has its own limit per API key, on top of the workspace limits, and some are much lower than the table. The operation's description states it, and its x-ken-rate-limit field gives it as limit calls per windowSeconds, with per saying what is counted: api_key, or ip on the few calls made without a key. Over an operation's own limit, the response is a 429 with the headers above and, for now, a plain-text body instead of the error body above.
Every operation's page in the reference shows its own limit under its endpoint: per API key, or per IP address on the few calls made without a key. Most are at least as high as the workspace limits, so the workspace limit is the one you meet first. Some operations that send, launch or start work are capped far lower.
These operations are capped under 100 calls a minute per API key, or per IP address where the table says so:
- Wait the
Retry-Afterseconds, then send the request again. - No
Retry-Afterheader: wait a moment, and double the wait after each further429. - A workspace limit answers with the code
rate_limited. An operation's own limit, and the per-key and per-IP limits in front of the API, answer with a plain-text body or{ "error": "Rate limit exceeded" }instead of the error object. Branch on the status, not oncode. - Watch
RateLimit-Remainingand slow down before it runs out. - Retrying a write is safe only with an
Idempotency-Key: see Idempotency.