Agent sandboxes / Agent 37 alternative

An Agent37 sandbox alternative for a growing persistent computer

Agent37 already retains the whole writable filesystem through normal stop, start, restart, and resize. Its default instance and full VM have different compute rates and kernel capabilities. For an evolving agent, examine image-update behavior, disk growth, and process continuity before comparing unit prices.

Reviewed 2026-10-18 · Published by the Sandbox0 team

When Agent 37 fits

Agent37 fits teams that want persistent instances with managed agent integrations, opt-in auto-sleep, URLs, SSH, and an optional full VM for Docker or hardware-accelerated Android. Its low standard rates describe shared-host compute, not identical hardware to every other sandbox.

When to consider Sandbox0

Consider Sandbox0 when the entire durable RootFS should expand online without restarting the running agent and supported file operations should work while compute stays paused. Test any full-VM dependencies against gVisor before migration.

Capabilities

Agent 37 and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeAgent 37Opt-in auto-sleep can retain execution state, with documented cold-boot fallback. Confirm any active-lifetime limits for the selected instance; sleep alone does not establish unlimited runtime.View evidence ↗Sandbox0ttl and hard_ttl default to 0 (disabled). Configure idle pause and deletion policies for the same persistent sandbox identity; an idle running sandbox still accrues compute charges.Read documentation →
Persistent filesAgent 37Files persist through stop/start, restart, resize, and host moves; image update resets the OS layer outside retained home data.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSAgent 37Normal lifecycle retains files across the whole filesystem, with distinct image-update semantics.View evidence ↗Sandbox0Persistence extends beyond /workspace to paths such as /usr/local, /etc, and /home. Runtime tmpfs and virtual filesystems are excluded.Read documentation →
Release compute, keep stateAgent 37Auto-sleep supports memory restore with documented cold-boot fallback.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateAgent 37Sleeping can retain processes; memory restoration is not guaranteed for every wake.View evidence ↗Sandbox0Experimental memory checkpoints can preserve compatible process state. Use memory: true on both pause and resume on a compatible deployment. A filesystem checkpoint alone does not preserve running processes.Read documentation →

Online RootFS growth

On growth-enabled images, Sandbox0 automatically expands the writable RootFS as free space runs low, within a configured sparse logical ceiling up to 64 TiB. Agent processes keep running. Growth is enabled in the qualified Cloud region; confirm activation for other regions. Existing images and snapshots retain their captured storage geometry. The ceiling is logical capacity, not preallocated host disk.

Agent37 supports explicit disk growth, but its VM documentation grows the filesystem on the next boot and resize discards memory. This differs from automatic online growth of Sandbox0’s mounted RootFS without restarting the agent.

Files while compute is paused

Sandbox0's Files API reads persistent files without resuming compute. Filesystem-only pauses also permit edits; retained-memory pauses are read-only and reject mutations with 409. SSH, SFTP, file watching, and runtime-only mounts require running compute.

See paused file operations →

Ordinary lifecycle retention differs from an image update

Agent37’s lifecycle guide preserves files across the filesystem during stop/start and resize, including host moves. Update resets the operating-system layer while keeping the home data directory; boot hooks can reinstall some tooling. Track system changes separately if template updates are part of the application workflow.

Sleep can continue processes or cold-boot

Opt-in auto-sleep saves execution state and bills disk only. The guide explicitly permits a fresh boot if a checkpoint is unusable, the instance moves hosts, or it enters cold storage. Persist essential work to files and make startup idempotent rather than relying on every wake retaining RAM.

VM resources and disk growth change the comparison

The full VM rate is 2.5 times default compute, with the same disk rate. VM resize cold-boots; growing the disk extends capacity and the filesystem grows on the next boot. That workflow retains data but differs from automatic online RootFS expansion while the agent continues running.

Pricing

Compare billing models, then price your workload

CPU allocation, memory, billing minimums, disk, plan fees, and supported workload differ. These meters are not equal-performance machine quotes. Confirm region, resource units, and included allowances before estimating a total.

Agent 37

Pricing reviewed 2026-10-18

Default instance CPU
$0.80 per vCPU per 730-hour month
Standard compute shares host resources; running usage is metered by the minute.
Rate source ↗
Default instance RAM
$0.70 per GB per 730-hour month
Rate source ↗
Retained disk
$0.09 per GB per 730-hour month
Continues while the disk is retained, including sleep and stop.
Rate source ↗
Full VM compute
2.5× the default compute rate
Disk retains the same rate; full VM kernel capabilities differ from the default gVisor instance.
Rate source ↗

These are Agent37’s own published rates, excluding its comparison table of other providers. Preserve GB units and the 730-hour reference month. Instance class, shared versus reserved hardware, model/tool usage, and retained disk affect the total; no equal-throughput or universal savings claim follows from the unit rates.

Check current Agent 37 pricing ↗

Sandbox0

$0.015/GiB-hour compute (billed by the second at approximately $0.000004167/GiB-second), plus $0.02/GiB-month of retained incremental RootFS and memory checkpoint data. Shared base images and platform recovery retention are excluded; paused sandboxes accrue storage charges only.

Compute / GiB-hour
$0.015
Retained data / GiB-month
$0.02
Paused compute
$0.00

A 4 GiB agent, 100 running hours

$6.00 compute

4 GiB × 100 hours, billed by the second. Keeping 20 GiB of billable incremental data for a full billing month adds $0.40. CPU is derived from memory by the platform; choose a tier that fits your workload.

Understand storage billing →

Before migrating from Agent 37

  1. Record instance type, billed disk, OS-layer changes, home data, image-update hooks, and auto-sleep or cold-storage behavior.
  2. Export required files and port templates, agent integrations, credentials, URLs, and service startup; assess full-kernel and Android requirements separately.
  3. Validate system-package retention, growing outputs, paused file operations, required process restoration, and the actual compute class and storage budget.

Common questions

Is Sandbox0 a drop-in Agent 37 replacement?

Evaluate API compatibility and product scope. A shared sandbox category does not make SDKs, lifecycle semantics, or limits interchangeable.

Does keeping files also keep a running agent?

Filesystem state and volatile process state are separate. Sandbox0 requires explicit compatible memory checkpoints to preserve running processes; its default filesystem resume starts fresh processes.

Sources and review scope

The Boat team's Battleships directory is the discovery source. References below describe the product and plan discussed on this page. Check current availability, regional terms and billing units before migrating. Where the cited material does not establish a capability or price, the comparison identifies that uncertainty.

Related comparisons