Skip to main content
Every machine starts from the same default size, every organization starts with the same quotas, and an idle machine hibernates on its own. Every limit on this page can be raised. If you need more machines or a different size, email contact@boxd.sh. The boxd team generally approves requests the same day.

What every machine gets

Disks are copy-on-write, so storage only grows as you actually write data. A fresh fork takes near-zero extra disk until it starts diverging from its source machine.

Machine sizes

A machine comes in one of four sizes. Naming either dimension picks the size, so --vcpu 1 means 1 vCPU with 4 GiB.
  • Creating. A size applies to fresh machines only. A fork and a machine created from a snapshot keep their source machine’s size, and passing a size there is an error.
  • Resizing. The new size is saved right away. A running machine reboots to apply it, so its disk persists and running processes are lost. A stopped machine picks it up at its next start. Resume or wake a machine in standby or hibernation first.
  • The org default. It only affects machines created afterwards. Every existing machine keeps its size.
  • The quota. This is the largest size a single machine may have, 8 vCPU / 32 GiB unless the boxd team raised it for your org.
The SDKs take the same sizes on create: see TypeScript and Python.

Organization limits

An org without a payment method is capped at 2 machines. Add a payment method and the cap goes to the full 50. Every machine counts toward the cap, hibernated and stopped ones included. Deleted machines don’t. Filling the cap with hibernated machines is normal use, since a hibernated machine bills for its disk only. With several regions, the machine limit applies in each region on its own. An org with a limit of 50 can have 50 machines in Frankfurt and another 50 in New York, and boxd manage machine-limit shows how many are in use in each.
Hitting a cap? Email contact@boxd.sh and describe what you’re running. The boxd team generally approves requests the same day.

Need more machines?

The org’s owner can ask for a higher machine limit without leaving boxd: from Request more machines on the console’s billing page, from the CLI, or from an SDK. Ask for the new total you want, above your current limit and up to 1000, and say why if you like. One request is open at a time, and the owner gets an email when the limit changes. Any member can see the limit, how many machines count against it, and whether a request is open.
In the SDKs: orgs.get_machine_limit() / orgs.request_machine_limit() in Python and orgs.getMachineLimit() / orgs.requestMachineLimit() in TypeScript. For more than 1000 machines, email contact@boxd.sh.

What each state costs

You don’t manage any of this by hand. A machine hibernates on its own when traffic stops and wakes when it returns. Usage is metered per second: You only pay for active vCPU, not for the vCPUs a machine reserves. A 2 vCPU machine that is active 10% of an hour pays for 0.2 vCPU-hours, and that is the typical case: on average, only 10% of a machine’s reserved vCPU is active. RAM bills on the memory the machine holds, not the GiB it was given, and disk bills on what has been written, not the 100 GB provisioned. Auto-suspend is off by default, and auto-hibernate kicks in after 4 hours without network traffic in either direction. Your app never notices the transitions. That is why a hibernated preview, an idle source machine, or a long-lived agent workspace can sit around for weeks and bill for disk only. See Suspend, resume, and hibernate.

Pricing

See boxd.sh/pricing for the current rate card, and email contact@boxd.sh for organization pricing.

Self-hosted

When you self-host boxd, you set the quotas yourself. boxd-ctl org set-quota overrides the platform defaults per org.