Agent sandboxes / tama alternative

A Tama sandbox alternative when an agent changes the whole system

Tama offers Linux machines, GPU workflows, CLI and SDK access, and stop/start with saved state. Its durable storage boundary is a 50 GiB workspace. Decide whether the agent only needs that working directory or also needs installed tools and system configuration outside it to return.

Start building Compare capabilities ↓

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

When tama fits

Tama fits GPU-dependent experiments, model work, and machine workflows that benefit from its shell, snapshot, and conditional GPU-memory restoration. Sandbox0 is not a direct replacement for a GPU machine or CUDA execution requirement.

When to consider Sandbox0

Consider Sandbox0 for CPU agent environments whose durable state includes system-installed packages and configuration beyond a workspace, with automatic online RootFS growth and supported paused file operations. Benchmark the actual agent and hardware requirements before migrating.

Capabilities

tama and Sandbox0: what persists?

Feature comparison with documentation and verification status
Capability	tama	Sandbox0
Environment and runtime lifetime	tamaNot established in this review	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 files	tamaThe 50 GiB /workspace disk survives a normal snapshotting stop and later start; deletion destroys it.View evidence ↗	Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFS	tamaThe guide explicitly limits durable edits across stop to /workspace.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 state	tamaNormal stop snapshots state and permits start; --no-snapshot is explicitly not resumable.View evidence ↗	Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process state	tamaWarm start is conditional on a captured memory checkpoint. Desktop snapshots and disk forks boot cold.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.

Tama documents a fixed durable 50 GiB workspace, not automatic online expansion of the complete writable RootFS. Sandbox0’s growth-enabled RootFS serves an agent whose package installs and outputs accumulate beyond a working directory.

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 →

The durable workspace defines what survives

Tama’s guide identifies /workspace as its durable 50 GiB disk and the only location whose edits survive a stop. Keep build outputs and checkpoints there, and separate image-provided software from changes made elsewhere during execution. Sandbox0’s persistent writable RootFS targets the wider system state, excluding runtime-only mounts.

Source 1 ↗

A warm start, a fork, and a public template have different state

Tama can resume captured process memory, including supported GPU state, when a memory checkpoint exists. Desktop machines do not capture RAM. Snapshot forks cold-boot from disk; public-template launches across workspaces also boot cold, so memory does not cross that boundary. Do not assume that every reuse path continues the original process.

Source 1 ↗

Use current offers and an idle policy suitable for the job

Tama publishes machine offers through its API and meters running seconds. The guide describes stopping as ending charges while retaining the saved disk. Available hardware and machine billing shape matter: a CPU unit rate alone is not a full GPU-machine quote. Its optional idle timer and explicit heartbeat need attention when a long background task otherwise looks inactive.

Source 1 ↗Source 2 ↗

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.

tama

Pricing reviewed 2026-10-18

Published CPU component: $0.011660 per vCPU-hour; Converted from the offers API’s integer microdollar rate; not a complete GPU-machine quote.; Rate source ↗
Published RAM component: $0.005523 per GB-hour; Preserves the API’s GB label. Check the selected machine offer and allocation.; Rate source ↗
Stopped machine: No machine charges while stopped, per current guide; Normal stop saves the workspace. Permanent deletion and stop without a snapshot have different consequences.; Rate source ↗

Offers and availability can change. GPU machine shapes and whole-node charging must be checked with the selected offer rather than inferred from the CPU component rates. This comparison makes no GPU-equivalence or universal savings claim.

Check current tama 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 tama

Save and export required /workspace data, identify image or system changes outside it, and classify CPU, GPU, desktop, and memory-continuity requirements.
Move CPU-compatible agent work into a Sandbox0 template and port command/file calls, credentials, ports, and startup behavior; retain a suitable GPU backend where needed.
Check repeated system changes, growing artifacts, paused file access, process restoration requirements, and cost using the actual machine offers.

Common questions

Is Sandbox0 a drop-in tama 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.

Battleships research record ↗tama.computer/llms.txt ↗api.tama.computer/v1/offers ↗Sandbox0 RootFS capacity →

Related comparisons

Agent 37 vs Sandbox0 →AgentComputer vs Sandbox0 →ainclave vs Sandbox0 →Alibaba Cloud Agent Sandbox / FC / AgentRun vs Sandbox0 →Alibaba Cloud AgentBay vs Sandbox0 →Amika vs Sandbox0 →