Usage, billing, and limits

VMYard bills machines per second while they run, at the hourly price of their plan. Stopped machines cost nothing. Every organization has a prepaid wallet; nothing is charged after the fact except a plan's renewal.

How usage is measured

Usage starts when VMYard observes a machine running and ends when it observes it stopped or gone. What a machine reports about itself never counts. A run keeps the hourly price it started with, and its charge after n seconds is floor(n × hourly price ÷ 3600), so partial seconds are never billed. Long runs are charged as they go, about once a minute.

See Sizes and prices for each size's resources and how to read them from the catalog.

All amounts in the API are integer micros of USD: 1,000,000 micros is $1. A field such as cost_micros: 30000 is $0.03.

The wallet

Usage is paid from three balances, in this order:

1. Trial: $1.50 of usage, which is 50 hours of default (100 hours of small), granted the first time the organization uses VMYard.
2. Plan: the current period's allowance, if the organization has a plan. Unused allowance expires at the end of the period.
3. Credits: bought in $20 packs. Credits never expire.

Owners and Billing members buy plans and credit packs, open the payment portal (cards, invoices, cancellation), and set two controls, all from the console's Billing page:

• Spend cap: the most credit spend allowed per UTC calendar month.
• Minimum balance: creates and starts need more spendable balance than this. The default is 0.

They can also cancel a plan at the end of its period or end the trial early, and turn on automatic refill to buy credit packs when the balance runs low.

Card details stay with the payment provider. Deployments without online payments answer those actions with 503 billing_unavailable; trial and operator-granted balances still work.

When the balance runs out

When the spendable balance (trial + plan + credits, with credits limited by the spend cap) reaches zero, VMYard waits a grace period (5 minutes by default) and then stops every machine that is still running. Those are ordinary stops: they emit machine.stopped. Buying credits or a plan before the stops begin cancels them. Usage metered during the grace period can make the credit balance slightly negative; the next purchase repays it first.

Limits

Creates, starts, and restarts are checked against the organization's limits. Replays of an already-accepted request are never refused.

Concurrent machines counts every machine that is meant to be running, including creates still in progress. Start windows are rolling and count creates, starts, and restarts. During the trial, only small and default sizes are allowed, and at most 2 machines run at once.

Card check

A deployment can require trials to verify a card before their machines start. GET /v1/billing reports it in card_check: required is true when the deployment asks for one, and verified_at is when a card was verified. Paid organizations never need one.

Until a card is verified, a trial's creates, starts, and restarts answer 402 card_verification_required. An Owner or Billing member starts the check from the console (the create page, the refusal, or the Billing page), or with the API:

curl -X POST "$VMYARD_API_URL/v1/billing/card-check" \
  -H "Authorization: Bearer $VMYARD_API_KEY"

The answer has a checkout_url: the payment provider's page, where the card is checked and screened for fraud but not charged. Once the provider confirms it, card_check.verified_at is set and machines can start. Card details never reach VMYard.

Suspended organizations

To stop abuse, the people who run the service can suspend an organization. While it is suspended, OrganizationDto.suspended_at is set, its machines are stopped (stop_reason: suspension), and anything that creates, starts, or reaches into a machine (commands, files, terminals, desktops) answers 403 organization_suspended. Open terminals and desktops close with reason revoked. Reads, billing, API keys, and settings still work, and nothing is deleted. Contact support to have a suspension reviewed. When it is lifted, the stopped machines stay stopped until you start them.

Trial machines that use every vCPU at full load for a long stretch, the pattern of crypto mining, are stopped for review (stop_reason: abuse) and members who operate machines get an abuse_stop email alert.

The SDKs raise BillingRequiredError, SpendCapReachedError, TrialMachineClassNotAllowedError, and LimitReachedError for these. Waiting doesn't clear concurrent_machines; stop a machine instead.

Read balances, limits, and usage

Every member, and every API key preset, can read billing.

vy billing                       # status, balances, plan, trial, spend cap
vy limits                        # each limit and its current use
vy usage --since 2026-10-01      # per machine, with a total
vy usage dev                     # one machine