Runtime

How to pause a sandbox when it is idle

Set idlePauseSeconds; after that many seconds with no request the sandbox pauses, and the next request wakes it.

On Runtime an idle sandbox can stop billing compute by itself, and keep its memory, files and processes while it sleeps. Left running and idle for 30 days, a 2 vCPU, 4 GiB sandbox costs about $22.50. Paused, it costs $0.08 per decimal GB of saved state a month, and a pause took a median 234 ms, measured on 24 September 2026 (speed). On the free trial a paused sandbox counts toward neither the eight at once nor the 50 hours.

Set an idle timeout

TypeScriptimport { Sandbox } from "withruntime";const sbx = await Sandbox.create({ idlePauseSeconds: 600 }); // ten idle minutesawait sbx.exec("echo started > /workspace/log.txt"); // each request restarts the clockawait sbx.update({ idlePauseSeconds: 1800 }); // change it later; 0 turns it off
Pythonfrom withruntime import Sandboxsbx = Sandbox.create(idle_pause_seconds=600)  # ten idle minutessbx.update(idle_pause_seconds=1800)  # change it later; 0 turns it off
Terminalid=$(runtime sandbox create --name dev --get-or-create --idle-pause 900)runtime sandbox update "${id}" --idle-pause 0     # never pause for idleness

An agent over MCP sets it with runtime_sandbox_manage, "action": "update" and idlePauseSeconds (MCP).

What counts as activity

The idle clock restarts on every exec, file, process, terminal, desktop or preview request. When it runs out, the sandbox pauses as if you had called pause(). With automatic wake on, which is the default, the next of those requests wakes it and runs.

Setting Value
idlePauseSeconds 60 to 86,400; 0 is never
Default 300, only when timeoutSeconds is left out, and only until the first request
Not set by default With an explicit lease, an image or a snapshot
sbx.info.idlePauseUnusedOnly true for that default; false for an idle time you set
CLI --idle-pause <seconds> on create and update
Python idle_pause_seconds

What it saves

A worked example from the published rates, for a 2 vCPU, 4 GiB sandbox that waits on a person and is in use two hours a day:

Pattern over 30 days Compute Paused storage
Left running all month, mostly idle about $22.50 at $0.03125 an hour none
Idle pause, 60 hours running $1.88 at the same waiting rate $0.08 per GB of saved state

Busy time is billed on measured CPU, $0.025 per active vCPU-hour, so real use adds to both rows alike. Paused storage counts only the disk and memory blocks the sandbox alone owns, not the disk size you asked for (paused storage).

Mistakes to avoid

  • Setting it below what a slow step needs. A request resets the clock, but a long exec is one request. Pick an idle time longer than the gaps between your calls, not the length of a command.
  • Leaving an unattended job to run. Idle time counts requests, not CPU. A long job started with spawn and left alone is paused with the sandbox when the idle time runs out, and carries on only after a wake. Give unattended work a longer idle time, or 0.
  • Forgetting the lease. Idle pause and the lease are separate. A busy sandbox still pauses when its timeoutSeconds runs out, unless you extend the lease.
  • Turning automatic wake off by accident. With autoWake: false, an idle-paused sandbox answers sandbox_paused until you call wake().

Facts on this page were checked on 25 September 2026.

Claim 50 hours free