Starter
$5
≈ 115 hours on a 1 GB box
A sub-lunch way to try the whole loop.
Pricing
Prepay a credit pack and we meter your box by the second while it runs. Bring your own model key, so your only bill from us is the machine — never a markup on tokens.
Starter
$5
≈ 115 hours on a 1 GB box
A sub-lunch way to try the whole loop.
Standard
$20
≈ 463 hours on a 1 GB box
Enough runway for a real project.
Pro
$50
≈ 1,157 hours on a 1 GB box
Long-running boxes, top up rarely.
Packs are prepaid credit, not a subscription. Buy more whenever you want; a box needs at least $1.00 of credit before it will provision.
Rate card
Two box sizes at launch. You're billed for the seconds a box is actually running; a suspended box costs nothing.
| Box size | Per hour | Per second |
|---|---|---|
| 1 GBshared-1x · 1 GB RAM | $0.0432 (4.32¢) | 1,200 µ¢ |
| 512 MBshared-1x · 512 MB RAM | $0.0288 (2.88¢) | 800 µ¢ |
µ¢ = microcents (1¢ = 1,000,000 µ¢). Your rate is snapshotted onto the box when it's created, so a later price change never touches a box you're already running.
How billing works
A running box meters by the second at its size's rate. No rounding up to the hour, no minimum session length.
Leave a box idle and it suspends — a suspended box isn't running, so it stops costing you. Reconnect and it wakes in seconds.
Credit is prepaid. When a balance hits zero the box suspends; there's no overage bill and your balance can't go below zero.
You need at least a dollar of credit on hand before a box will start — a fail-closed floor, not a fee.
Packs don't renew and don't expire into a recurring charge. Top up when you want more machine-time; that's the whole billing model.
You bring your own OpenAI or Anthropic key, and model traffic goes straight from your box to the provider. They bill you for tokens; we never touch inference billing — zero token margin, structurally.
A cloud box, your agent, your key, metered by the second. Be first in line when early access opens.
No spam — one email when it's ready.