Updated · 2026-07-13

Pricing

Compute is billed per second at flat unit rates (per vCPU-month + per GB-RAM-month) against the size you chose; egress per GiB; built-in models per token; the inbox is free within caps. All rates live on the pricing page and via the pricing / llm_list_models tools. A stopped VM bills nothing.

Rates for every service live on the pricing page — compute, egress, built-in models, and email, in one place. Your agent can also read live rates on demand: pricing (or GET /v1/pricing) for the unit rates + worked-example prices, and llm_list_models (or GET /v1/llm/models) for per-model LLM rates. This page describes how billing works.

VM compute

Compute is billed per second at flat unit rates — a rate per vCPU-month and a rate per GB-of-RAM-month — against the size you chose. You pick vcpuMillicores + memMib at create_vm and change them whenever you like with resize_vm; the meter prices exactly that allocation, so the bill never includes headroom you didn’t ask for. See VM sizing. The rootfs disk is included (not separately billed); grow it per-VM with create_vm({ diskGib }) or attach a volume for more.

For the exact unit rates, worked-example prices (“static site”, “small API”, “Next.js app”, …), and your VM’s live run-rate, ask your agent:

pricing

Or run:

curl https://api.superjolt.com/v1/pricing

Egress

Outbound traffic is metered separately per GiB (inbound is free). The rate is on the pricing page and in the pricing response as egressMicroCentsPerGib.

Built-in models & email

Built-in models bill per token against your balance, and the inbox is free within caps (sandbox sending is rate-limited; your own domain bills through your own provider). The per-model rates and the email caps are on the pricing page; live model rates also come back from llm_list_models.

What each state costs

  • running — billing per second at the run rate of its current size.
  • stopped — billing nothing. stop_vm releases the CPU, memory, and IP and archives the disk (and any attached volume) to object storage; there is nothing left running to charge for. start_vm restores it — see below. Archives are held ~30 days.
  • Suspended — billing nothing. Same mechanism as stopped, but triggered for you: your balance hit zero, or you hit the monthly spend cap with the hard cap on. The VM stays visible so you know to act. Held ~30 days; a top-up (or raising the cap) resumes it automatically. After that the archive is purged.
  • Trash — billing nothing. A VM you deleted, held for 24 hours before permanent purge; resume_vm brings it back within the window.

Stopping is the way to stop paying for a VM you’re not using — without losing it. Starting it again restores the disk onto a host with capacity and boots it, which takes tens of seconds (a download plus a boot) rather than being instant.

Capping your spend

set_budget_settings sets a monthly spend cap for the account. You get email and in-app alerts at 50% / 80% / 100%, and you can optionally turn on a hard cap that Suspends every running VM at 100% — visible and recoverable, exactly like a zero balance. get_budget_settings shows the current setting. Spend resets at 00:00 UTC on the 1st.

The account_status tool returns your current burn rate per minute, hour, and day, plus runway in time-to-zero.