Runtime Cloud is a hosted fleet of Firecracker microVMs that an agent, or the program driving it, creates, runs commands in, pauses, forks and deletes over an API.

It is sold as compute for AI-agent workloads rather than websites or long-lived servers. The commercial pitch is that CPU is billed on measured use with a very low idle floor, so a sandbox waiting on a language model costs almost nothing.

Twenty products sit behind one API, CLI, MCP server and SDK set: sandboxes, secrets, images, volumes, snapshots, previews, desktop, metrics, webhooks, OpenTelemetry, TCP ports, private networks and more.

At a glance

Who it is forTeams building agent products that execute model-written code, and teams migrating off a rival sandbox SDK: drop-in shims exist for E2B, Daytona, Vercel and Blaxel.
IsolationFirecracker microVM on servers Runtime operates, with its own kernel; not a container and not gVisor. New sandboxes expose /dev/kvm, so nested virtualisation works. No GPU and no ARM: x86-64 only.
SizesNo fixed size menu: any vCPU / memory / disk combination inside the bounds, which are 16 vCPU, 64 GiB of memory and 400 GiB of disk on a paid account. CPU is shared by default; reserved guarantees every vCPU at full price.

How it bills

ModelPrepaid credit, no plan fee, no seats and no commitment.
MeterPer second: $0.025 per active vCPU-hour measured, with a 50-millicore floor per sandbox, plus $0.0075 per reserved GiB-hour of memory. Storage while paused or stopped is $0.08 per decimal GB per 30-day month.
While stoppedDisk and saved memory blocks keep billing at the storage rate. Only delete ends the charge.
Plan feeNone publicly. Plans exist in a private beta with no published prices.
MinimumNo minimum billing period. Top-ups are $10 to $10,000, and an automatic top-up earns 5% extra credit.

How Runtime lets you try before paying is on Trials and free tiers.

Where it stands out

  • Secrets the machine never holds: placeholders swapped into HTTPS headers for named hosts only.
  • OIDC workload identity that AWS and GCP can trust.
  • Pause with memory, and snapshots that keep memory by default.
  • SSO, SCIM, audit logs and roles free on every account.
  • Billed on measured CPU, so a box waiting on a model costs the floor, about $0.03 an hour.

Using it without writing code

DashboardYes, a detailed console: a Home report with a Needs you section, sandboxes with Overview, Terminal, Files, Activity and Settings tabs, a limits page, and usage and billing with daily spend by product.
Without codePartly, and only for the first step: one click starts a sandbox with a shell in the page on free hours. There are no app templates and no hosted agent, and the documented no-code path is pasting its llms.txt link into Claude Code, Codex or Cursor.
SignupA Google sign-in, no card and no waitlist. Free usage needs a verified sign-in and is withheld from data-centre or VPN addresses until credit is added.
Docs and supportExtensive, every page dated, with llms.txt bundles. Support is an AI agent in the same call, with humans for billing, security and data loss. The status page is public and specific.

Caveats and contradictions

Shown as the research found them. A contradiction means the provider's own pages disagree; an open question means the answer is not published.

  1. 01

    A stopped sandbox keeps billing Open question

    Only delete ends the storage charge. A stopped box is a bill, not a cleanup, and persistent: true bills memory around the clock.

  2. 02

    Uptime was below the published promise Open question

    The status page showed 96.37% over 7 days and 98.87% over 30 days against a 99% monthly promise at the time of research.

  3. 03

    Free hours: every month or once? Contradiction

    The start page says 100 free hours every month; the pricing and included-usage pages say an account that has never paid gets them once, across its first 30 days.

  4. 04

    The canonical state list is not published Open question

    The docs name transitions and mention starting, running, pausing, paused, resuming, stopping and stopped, but there is no quotable list of state strings.

  5. 05

    One region, in Virginia Open question

    Everything runs in us-east. There is no EU option, and no per-tenant residency control.

  6. 06

    Very young Open question

    A one-month-old repository, near-zero community discussion, and daily changelog entries including breaking changes. No preinstalled coding agents, and the vendor says no bridge harness has run on Runtime yet.

Sources

The three shared workloads

List-price illustrations, not quotes

A · 1,000 short tasks

$5.21 to $13.33 a month

2 vCPU / 4 GiB / 4 GiB disk, deleted after each task

  • Memory166.67 h × 4 GiB × $0.0075$5.00
  • CPU at the 50-millicore floor0.05 × 166.67 h × $0.025$0.21
  • CPU fully busy2 × 166.67 h × $0.025$8.33
  • Disk under 10 GiB is included while running, and deletion ends the charge.
  • The range exists because per-task CPU use is not published; the floor sets the bill for a task that mostly waits.

B · One always-on box

$48.15 to $120.24 a month

4 vCPU / 8 GiB / 50 GiB, 730 h, always on

  • Memory8 × 730 × $0.0075$43.80
  • CPU at the floor0.05 × 730 × $0.025$0.91
  • CPU fully busy4 × 730 × $0.025$73.00
  • Disk past the included 10 GiB(53.7 − 10.7) GB × $0.08$3.44
  • Always-on needs persistent: true or idlePauseSeconds: 0; otherwise the box pauses after 5 idle minutes.
  • Disk is an upper bound: only blocks the disk alone holds are billed, so a mostly empty disk costs less.

C · Three hours a day

$4.42 to $8.81 a month

2 vCPU / 4 GiB / 20 GiB, 90 running hours in a 720-hour month, stopped otherwise

  • Running, at the floor0.05 × 90 × $0.025 + 4 × 90 × $0.0075$2.81
  • Running, fully busy2 × 90 × $0.025 + 4 × 90 × $0.0075$7.20
  • Stopped disk21.5 GB × $0.08 × 630/720$1.50
  • A 30-day month is 720 hours, so 630 hours are stopped.
  • Pausing instead keeps memory and costs about $4.72 to $9.11.