Reference

Plans, usage and limits

Two plans, metered usage, and a monthly spend cap on every organization.

Checked 2026-09-28

There are two plans, Free and Pro. There is no team plan and no enterprise plan.

#Free

• No fee, and no payment method.
• No overage. Usage past the included allowances is refused, never billed.
• One project, one member, one database, and no custom domains.
• 10 concurrent runners.
• Control API 2 requests a second, with a burst of 20.
• AI inference 1 request a second, with a burst of 10.
• A small monthly AI allowance, and a one-time sandboxes allowance.

#Pro

• A monthly fee, with the same amount of usage included.
• A payment method is required. Usage past the included allowances is priced at the published list price; Pro billing is switched off at present, so no Pro charge is taken.
• Unlimited projects, members, databases and custom domains.
• 100 concurrent runners.
• Control API 50 requests a second, with a burst of 200.
• AI inference 100 requests a second, with a burst of 400.
• Rate limits rise with monthly spend: 150 requests a second past $500, and 500 requests a second past $5,000.

#Usage and prices

Usage is metered per service: compute, memory, storage, egress, build minutes, run minutes, emails, log ingestion and others.

Prices are published when billing opens.

Usage is reported per month per organization. You read it in the console, and it is what an invoice is built from.

#The monthly spend cap

Every organization has a monthly spend cap of US$200 by default, and you can change it. Removing the cap entirely requires a verified payment method.

#When a limit is reached

Past a plan limit, the call fails with PLAN_LIMIT_REACHED. Past the spend cap, it fails with SPEND_LIMIT_REACHED. Both answer HTTP 402, and both are on Errors.

A Free organization is hard-capped: past its allowances its calls fail rather than bill. That is the point of the plan, and it is why a Free organization never needs a payment method to stay safe from a surprise invoice.