Billing

Locked prices

These come from one constant table (never computed at runtime from anything configurable): hourly = monthly ÷ 667, i.e. paying hourly costs about 9.5% more than committing to monthly for the same full-time usage -- committing saves.

billing_mode: chosen per desktop

billing_mode ("monthly" or "hourly") is chosen once, at POST /v1/desktops and stored on the desktop row -- it isn't a global account setting. It defaults to hourly when omitted:

curl -sS -X POST "$BASE/desktops" \
  -H "$AUTH" -H 'Content-Type: application/json' \
  -d '{"tier": "default", "billing_mode": "monthly"}'

It's returned on every DesktopJson and factored into that desktop's projected cost in GET /v1/usage.

Hibernated desktops cost nothing

Only awake time is billed. A hibernated desktop keeps its full disk state but stops accumulating awake-seconds entirely -- hibernate whenever you're not actively using a desktop and its compute cost drops to $0 until the next wake.

Awake-seconds metering

Awake time is tracked as a sequence of open/close events -- a desktop becomes "guest responsive" (opens the meter) and later hibernates, is force-recovered after a host failure, or is destroyed (closes it). Each desktop's raw awake-seconds are then weighted by its tier (small = 0.5x, default = 1.0x, large = 2.0x) into weighted_awake_seconds, and its existence over the queried range is separately expressed as desktop_months (a 30-day-month convention: overlap of the desktop's lifetime with the query range, divided by 30 days in seconds).

Projected cost: GET /v1/usage

Query a time range (Unix milliseconds) to get per-desktop and org-total usage and projected cost at the locked prices above:

END=$(($(date +%s) * 1000))
START=$((END - 86400000)) # last 24h
curl -sS "$BASE/usage?start=$START&end=$END" -H "$AUTH"

{
  "per_desktop": [
    {
      "desktop_id": "d_...",
      "awake_seconds": 36000.0,
      "weighted_awake_seconds": 36000.0,
      "desktop_months": 0.0139,
      "billing_mode": "hourly",
      "projected_cost_cents": 14
    }
  ],
  "org_awake_seconds_weighted": 36000.0,
  "org_desktop_months": 0.0139,
  "org_projected_cost_cents": 14
}

projected_cost_cents is computed per desktop from its billing_mode: hourly desktops are awake_seconds / 3600 * hourly rate; monthly desktops are desktop_months * monthly rate -- both rounded half-up to the nearest whole cent. This is display/projection math, meant for showing an estimated running cost -- not the exact-cent precision a future Stripe metered-billing integration will need for real invoices.

See API reference: usage for the full route contract.