Agent sandboxes / Shardflux alternative

A Shardflux alternative for a growing durable RootFS

Shardflux has two materially different modes: a default processful microVM retaining files and execution state, and an opt-in file-first tree with a fresh VM for each command. It also supports paused file reads. Compare the selected mode, remote durability receipt and file-client wake behavior with Sandbox0’s durable growing RootFS.

Start buildingCompare capabilities ↓

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

When Shardflux fits

Shardflux fits agents that benefit from keyed workspaces, process snapshots, revision-checked file edits and searches, or a deliberately file-first workflow. Its shared organization allowances and opt-in capped overage can be useful when aggregate agent usage is predictable.

When to consider Sandbox0

Consider Sandbox0 when system packages, configuration and data across the writable RootFS should persist and expand online, while supported paused-file operations leave guest compute stopped. Memory continuation is an explicit compatible experimental mode rather than the filesystem-only default.

Capabilities

Shardflux and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeShardfluxPersistent workspaces live until explicit deletion; session workspaces have a separate idle/close deletion policy.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 filesShardfluxPersistent processful workspaces retain VM files; session lifetime and file-first mode have different deletion/path boundaries.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSShardfluxProcessful suspend captures VM disk and memory; file-first mode only retains /home/user, and template immutable paths have separate update behavior.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 stateShardfluxProcessful suspend releases compute and restores execution; remote durability requires its own completion receipt.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateShardfluxProcessful checkpoints capture memory and processes; this does not apply to file-first execution.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.

Shardflux documents explicit disk_gib resize for running or suspended workspaces, subject to plan ceilings and its accounting of workspace changes. Template immutable paths can follow newer published versions. Neither explicit resize nor a file-first tree establishes automatic online growth of the entire writable RootFS as free space runs low.

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.

Shardflux serves suspended read/stat/list/search requests from saved disk without compute. Writes and patches wake processful workspaces; when an offline read is unavailable or over budget, its SDKs/CLI/MCP wake the machine and retry. File-first mode supports direct tree mutations without a VM, but only for /home/user. Sandbox0’s supported paused Files API keeps compute stopped, permits mutations during filesystem-only pauses and returns conflicts instead of silently waking or dropping memory state.

See paused file operations →

Processful and file-first are different persistence contracts

Default processful workspaces keep a VM’s files, memory and processes and support suspend/resume. File-first mode instead versions /home/user, runs each command in a fresh VM and discards processes, memory and files outside that tree. Mode is fixed when the key is created. Persistent and session lifetimes also differ; session close or its idle expiry deletes the workspace.

Source 1 ↗Source 2 ↗

A finished local suspend is not yet a remote durability receipt

Suspend initially seals memory/disk on the host and releases compute; asynchronous copying may still be pending. The operation result exposes durable and checkpoint identity, and durable:true waits for remote persistence. If the host disappears before that copy completes, recovery can use an earlier durable checkpoint with lost_suspend details. Do not turn vendor latency samples into a durability guarantee.

Source 1 ↗

Memory metering needs the current policy, not just the headline

The homepage describes full working allocation and a brief parked-memory tail. The current limits/workspace guides describe default elastic held memory with a 1 GiB floor, fixed allocation as another option, and compressed parked residency. The overage example still uses the shorter parked-tail description. Confirm effective elastic/fixed and parking policy rather than using one illustrative session as a universal bill.

Source 1 ↗Source 2 ↗Source 3 ↗

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.

Shardflux

Pricing reviewed 2026-10-18

Developer
$9/month
Organization allowance: 100 CPU-hours, 800 RAM GiB-hours and 100 GiB retained state.
Rate source ↗
Pro
$29/month
Organization allowance: 500 CPU-hours, 4,000 RAM GiB-hours and 250 GiB retained state.
Rate source ↗
Startup
$79/month
1,500 CPU-hours, 12,000 RAM GiB-hours and 1,000 GiB retained state, shared across the organization.
Rate source ↗
Opt-in CPU overage
$0.04 per CPU-hour beyond allowance
CPU time actually used; overage is disabled by default.
Rate source ↗
Opt-in RAM overage
$0.015 per GiB-hour beyond allowance
Confirm elastic/fixed and parked-memory metering against the active policy.
Rate source ↗
Free
$0/month with 15 CPU-hours
RAM is capacity-as-available; retained state allowance is 10 GiB.
Rate source ↗

Allowances belong to the organization, not each workspace. Paid overage has a user-set period cap no greater than the plan fee; Free has no overage. Retained state/shared-volume storage is quota-limited without storage overage, and outbound transfer blocks at its allowance. The temporary 30-day Pro signup trial is not permanent free Pro.

Check current Shardflux 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 Shardflux

  1. Identify each key’s processful/file-first mode and persistent/session lifetime; wait for and record a durable checkpoint before retiring its source host.
  2. Export system and user state with template immutable paths and shared volumes accounted for; recreate credentials, Linux tooling and revision-sensitive file workflows.
  3. Test process continuity deliberately, files outside /home/user, online RootFS growth, no-wake file access and organization-wide allowance/overage costs.

Common questions

Is Sandbox0 a drop-in Shardflux 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 ↗docs.shardflux.dev/concepts/workspaces ↗docs.shardflux.dev/concepts/file-first.md ↗docs.shardflux.dev/concepts/lifecycle.md ↗shardflux.dev/ ↗docs.shardflux.dev/limits.md ↗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 →