Agent sandboxes / Together Code Sandbox alternative

A Together code sandbox alternative for an evolving agent computer

Together’s current sandbox SDK uses snapshots as the reusable handle between VM runs. It can save the whole filesystem and optionally memory, but an ordinary termination without a saved-state policy is ephemeral. Compare that lifecycle with returning to one persistent sandbox identity.

Start buildingCompare capabilities ↓

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

When Together Code Sandbox fits

Together Sandbox fits teams using Together’s model platform who want snapshot-built VM environments, Python or TypeScript clients, and explicit saved-state policies for agent tools or RL rollouts.

When to consider Sandbox0

Consider Sandbox0 when the agent should retain one writable Linux environment across tasks, expand its RootFS while running, and let a control plane access persistent files without waking compute.

Capabilities

Together Code Sandbox and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeTogether Code SandboxA configured TTL counts from creation, not idle activity. The reviewed guide does not establish an unlimited continuous 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 filesTogether Code SandboxTermination snapshots preserve state when configured; an ordinary ephemeral teardown does not.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSTogether Code SandboxThe termination snapshot captures the whole filesystem.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 stateTogether Code SandboxSaved-state termination followed by creation from a snapshot; the resumed VM gets a new sandbox ID.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateTogether Code Sandboxsnapshot.memory opts into disk and memory capture; disk-only is the default.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.

The current migration guide fixes CPU and memory specs at creation and documents recreating from saved state for changed specs. Automatic online RootFS expansion is not established by the reviewed snapshot workflow.

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 →

Persist a snapshot handle across VM runs

Together’s migration guide maps hibernate to termination with a snapshot, and resume to creating a new sandbox from that snapshot. A sandbox ID ends at termination. Store the snapshot alias or ID when later work must return to its saved files and optionally retained processes.

Source 1 ↗

Memory capture and retention are explicit policy

The current SDK can save disk only or disk plus memory; memory is opt-in. Without a termination policy, the sandbox is ephemeral. This differs from older CodeSandbox SDK hibernation defaults, so migrate lifecycle code deliberately rather than just changing the package name.

Source 1 ↗Source 2 ↗

Code Interpreter sessions use another meter

Together publishes resource rates for Code Sandbox and a separate per-session rate for Code Interpreter. Use the VM resource rows for a general agent environment. The low per-session interpreter charge does not price a customizable VM with the same CPU, memory, and lifecycle.

Source 1 ↗

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.

Same size. Lower compute cost.

2 vCPU · 4 GiB RAM · 730 running hours / month

Save $64.82 / month

The same CPU and memory allocation, with runtime aggregated across sessions. Monthly allowances and credits below are applied to this workload.

Together Code Sandbox

Published Code Sandbox compute

$108.62

$0.1488/hour before monthly allowances

Sandbox0

Pay as you go · no monthly plan fee

$43.80

$0.06/hour · 59% lower in this example

Uses the VM Code Sandbox resource meters. Account or deployment fees require confirmation and are excluded here; the $0.03 Code Interpreter session is a separate product.

Compute and stated plan fees only. Storage, network, taxes and one-time credits are excluded; hardware and throughput can differ.

Calculation and billing assumptions

Together Code Sandbox: $108.624 compute after 0 included vCPU-hours and 0 included GiB-hours, plus $0.00 plan fee, minus $0.00 applied monthly credit.

Sandbox0: 4 GiB × 730 hours × its published compute rate. CPU uses the Cloud allocation shown on the pricing page.

Includes compute and the selected plan. Excludes storage, network, taxes, one-time grants and negotiated discounts. Assumes this workload can use the full listed monthly allowances. CPU and memory counts do not guarantee equal hardware or throughput.

Checked Together Code Sandbox rate source ↗Sandbox0 rates ↗

Together Code Sandbox

Pricing reviewed 2026-10-18

Code Sandbox CPU
$0.0446 per vCPU-hour
Rate source ↗
Code Sandbox RAM
$0.0149 per GiB-hour
Rate source ↗
Separate Code Interpreter product
$0.03 per 60-minute session
This is an interpreter session price, not the Code Sandbox VM rate.
Rate source ↗

Verify deployment terms and retained snapshot/storage charges for the selected account. Model inference and the separately listed shared-filesystem product are distinct charges; neither is silently bundled into the VM resource rows.

Check current Together Code Sandbox 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
Understand storage billing →

Before migrating from Together Code Sandbox

  1. Inventory Together snapshots, aliases, termination policies, memory capture flags, and TTLs before terminating a source sandbox.
  2. Replace saved-snapshot/new-ID orchestration with the intended Sandbox0 identity and pause model; port SDK exec, file, and ingress calls.
  3. Run an agent task that accumulates dependencies and data, and validate RootFS growth plus filesystem-only or experimental memory restoration.

Common questions

Is Sandbox0 a drop-in Together Code Sandbox 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 ↗github.com/togethercomputer/together-sandbox/blob/c78355868d306d3b8734b40c589a4c41869f2e52/docs/migrating-from-codesandbox.md ↗docs.together.ai/docs/together-code-sandbox ↗www.together.ai/pricing ↗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 →