Pricing questions

One rate card for everything, why idle bills zero compute, database tiers, per-request pricing (there is none), and a worked cost example.

What does PandaStack cost?

One rate card, two meters. Sandboxes, apps, and managed databases all bill the same way — there is no per-class pricing:

Meter	Rate
Active vCPU	$0.054 per vCPU-hour
Memory	$0.0162 per GiB-hour

CPU bills by active CPU-seconds actually burned, capped at the committed vCPU ceiling. An 8-vCPU sandbox that sits idle bills almost no CPU.
Memory bills by the RAM your workload actually keeps resident, capped at the committed (template-baked) size — a 4 GiB sandbox pays at most 4 × $0.0162 per hour, and usually less.

The live card is always available at GET /v1/pricing. Full details in Billing.

What does a hibernated sandbox cost?

$0 compute. Hibernated sandboxes, sleeping apps, and auto-suspended databases bill no CPU and no memory — only storage (volume storage beyond your plan quota is $0.15 per GiB-month). Billing is per second of actual runtime, so scale-to-zero workloads really do cost nothing while idle.

Do you charge per request or per invocation?

No. There are no per-request fees anywhere — not on apps, functions, or sandbox previews, and no separate build minutes (app builds run inside your app's VM at the same two rates). Network egress is metered but not billed on any plan, within fair use.

What do managed databases cost?

Databases pick a RAM tier at create time — 1g (default), 4g, or 16g — and then bill on the same two meters as everything else: the CPU they burn and the memory they hold resident. Backups, point-in-time recovery, and failover are included. A database that auto-suspends bills storage only.

pandastack database create --label staging --size 4g

To move tiers, clone into a new size (pandastack database clone <id> --size 16g) — that is the supported resize path. See Databases.

Can you walk through an example?

A code-interpreter sandbox (2 GiB template) that runs for 30 minutes, does about 6 minutes of busy single-core work (0.1 active vCPU-hours), and keeps roughly 1 GiB resident:

Meter	Usage	Cost
Active vCPU	0.1 vCPU-hr × $0.054	$0.0054
Memory	1 GiB × 0.5 hr × $0.0162	$0.0081
Total		≈ $0.014

About a cent and a half. For scale: a hobby app that sleeps when idle and is awake ~2 hours a day (say ~1 GiB resident while awake, a few active vCPU-hours a month) lands around a dollar a month — comfortably inside the free tier's $5.40 monthly credit. An app kept always-on is a different story: it holds memory resident around the clock and will exhaust the free credit in days, which is why always-on is a Pro-and-up pattern.

What happens when free credit runs out?

At 80% you get a warning email. At 100%, running resources are paused (billing stops with them), new creates return 402, and nothing is deleted. The pause lifts when you upgrade or when the monthly credit resets. See Billing.

Where do I see what I've spent?

The dashboard's Billing page shows the credit meter and month-to-date usage; GET /v1/pricing returns the current rates and tier catalog programmatically.