Give your agents
room to work.
A short sprint. A monthly build. Your daily coding workload. Choose a useful amount of capacity, not an arbitrary number of API keys.
More work.
Not more gates.
API-key management, basic usage visibility and spending controls belong on every plan.
Prefer PAYG? Compare token ratesWhat changes when I upgrade?
In this lineup: included usage, parallel requests, sustained throughput, scheduling priority and workspace controls. Exact quotas and throughput limits must be costed and load-tested before launch.
Why not limit the number of API keys?
Keys separate projects and make credential rotation safer. They do not create compute capacity. All keys should share the account's underlying usage and throughput limits.
What happens when my allowance runs out?
Recommended policy: pause paid work by default, or continue through an explicitly enabled PAYG balance with a spending cap. No automatic surprise bill.
Is this dedicated hardware?
These self-serve plans do not promise reserved GPUs. Reserved capacity, custom hosting, fine-tuning and contractual SLAs should be separately scoped and priced. Existing offers are reproduced only in the reference view.
Why aren't final token quotas on the cards?
Raw token counts are not interchangeable across models or input/output mixes. Final usage allowances are still being costed against serving data and will be published before launch.