A plan is a monthly subscription, in USD through Stripe, that pays for the requests made with the keys it covers. Each plan's page lists its price and its limits. A workspace's API plan covers requests made with its API keys. It does not cover the keys Sail creates when you sign in to it, which have a Sail plan of their own. Requests a plan does not cover are charged to credits.
- A plan renews automatically each month on the day it started, and we charge the saved card until the plan is cancelled.
- The workspace owner can cancel from Billing at any time. The plan keeps working to the end of the period already paid for, then ends and is not charged again. Cancel a plan before you delete its workspace.
- A change from one Sail plan to another, made from Billing, applies at once, and the difference in price for the rest of the period is charged, or credited, on the next invoice. An API plan changes tier by cancelling it and subscribing to the new tier. The new tier starts when you subscribe, and the period already paid for on the old one is not refunded.
- If a payment fails, Stripe retries it and we email the workspace. The plan keeps working until 7 days after the first failed payment, then stops until the payment goes through.
- Plan payments are not refunded for part of a period, including when you cancel, except where the Australian Consumer Law requires it, where we end this agreement without cause, or as Fair use below says. Any other refund is at our discretion. If a workspace that already pays for a plan is charged for a second plan, we cancel the second plan and refund its charge.
A plan has no token limit. The tokens its requests use are listed on the usage page and not charged. While an API plan is in force, requests made with the workspace's API keys are not charged to credits, and the credits stay in the workspace for when the plan ends.
An API plan has a cap instead, in output tokens per second, shared by every API key in the workspace. Output tokens count in full and uncached input tokens count at a lower weight, set out on the plan's page. Cached input tokens are free. Over the cap, responses are slowed to it; a request is not refused for the plan. A workspace has a set number of API requests in flight at once, listed on the plan's page; a request over that number waits, and is refused with a retry time only after about 60 seconds of waiting. A Sail plan allows a set number of requests in flight at once, listed on its page, and a request over that number is refused until one finishes. Key rate limits still apply to both.
You may use a plan inside your own products and services, for your own end users. Do not resell, sublicense or share raw API access or a plan's capacity with anyone else, whether by giving them keys or through a proxy that passes their requests on. We may suspend a workspace that does, under Suspension and termination.
Fair use. A plan's tokens are unlimited: a plan has no token limit, and an API plan is paced to its cap in tokens per second. We may limit, slow or suspend a plan whose use is abusive or unlawful, is resold or shared outside the workspace, harms the service for other customers, or is far outside ordinary use of that plan. We email the workspace when we do, and say what we applied. If you do not accept a limit we apply, you can cancel the plan and ask us for a refund of the unused part of the period already paid for, pro rata.
Where we state a speed or priority for a plan, it is a typical figure measured on our fleet and not a guarantee. Speed depends on the model, the region, the length of the request and how busy the region is.
We may change a plan's price. The new price applies from the first renewal at least 30 days after we email you about it. You can cancel before then and pay nothing at the new price.