Using the API
Rate limits
Generous on purpose: far above what a person's CLI or a CI job sends, so only a flood is slowed.
At the edge
api.rykloud.com takes up to 20 requests a second from each address, with bursts of 200. Beyond that it answers 429 before the request reaches the API, so the body is the edge's, not the JSON error shape, and there is no Retry-After header: pause for a second or two and back off. A request body may be at most 1 MB at the edge and 64 KB at the API.
In the API
| What | Limit | Answer |
|---|---|---|
| Sending an invitation again | Once a minute, ten times at most | 429 resend_too_soon with detail.retryAfterSeconds; then 409 resend_limit |
| Organisations a person owns | Five | 409 organisation_limit |
| Live environments | Set per organisation (a new organisation starts with one) | 409 limit_live |
| Spending | The organisation's own budget and hard limit | 409 limit_spend, hard_limit |
| VPN keys | Five live keys per person per environment | 409 vpn_key_limit |
| The cloud shell (opened in the console) | By default three open per person, 20 per organisation, five new a minute, eight hours a day | 429, and a session past its hours ends |
The API sets no other limit on how often you call it. Reading logs is the one costly read (each runs inside the environment): poll it every ten seconds or so.