boat vs shellbox

Both give you a real Linux VM with root, Docker and nested virtualization, hosted in the EU. shellbox is a lean SSH box that pauses with its memory when you log off. boat is built to run many agents at once, with the agents and a desktop already inside and an optional proxy that keeps model keys out.

Where boat stands out

Agents and keys built in

Seven coding-agent harnesses, Claude Code and Codex among them, come preinstalled, and you prompt, steer and follow them from outside through one API. With the optional paid credentials proxy on, real model API keys stay outside the sandbox, and a credential can switch to another account when one runs out of credits.

Fleets, not a handful of boxes

boat plans run 100 sandboxes at once on $20 and up to 2,000 on $2,000, with 200 to 2,400 starts a day included. You can fork a running sandbox from its last disk save. shellbox allows 16 running slots per account, so 8 boxes of 4 vCPU, and copies a box with its memory once it is stopped.

A screen and more ports

Every boat sandbox has a 60 fps Linux desktop and Chrome for computer-use and browser agents. It also serves HTTPS on up to 50 ports, token-protected by default, plus raw ports on the sandbox IP.

The companies and products

boat

boat.dev runs sandboxes for AI agents: each one is a full Linux VM with its own kernel.

You start a machine with one command or one API call, then SSH in, run Docker, open a desktop, or let a built-in agent (Claude Code, Codex, OpenCode and others) work in it. The disk stays between sessions, and a stopped sandbox costs nothing.

shellbox

shellbox sells instant Linux boxes over SSH for builders and AI agents, with no signup: your SSH key is your account.

Each box is a microVM on Firecracker by default, or Cloud Hypervisor when you need /dev/kvm, running Ubuntu 24.04 with Docker on Hetzner servers in Helsinki. You can also create a box from a public OCI image. You manage everything with SSH commands that can return JSON. A box pauses with its memory when you disconnect and resumes when you connect again. It suits developers who want a cheap dev box for a few hours a day, and agents that already speak SSH.

Which use cases each one fits

For each use case, Battleships prices every sandbox provider on the same workload and sorts them from the cheapest monthly bill up. “5th cheapest of 365” means 4 of the other 364 providers run that workload, with every feature it needs, for less. “With limits” means the provider runs it but misses something, so it is listed after every full match. Open a use case for its whole ranking and what each provider misses.

Use case and its workloadboat: price rankshellbox: price rank
Software factory
100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month
A software factory runs many agents at once. shellbox's 16 running slots fit 8 boxes of 4 vCPU. boat's $20 plan runs 100 sandboxes at once, with 200 starts a day included.
5th cheapest of 365
$634 a month
66th of 365, with limits
$485 a month
Always-on AI employees
50 machines 24/7 · 2 vCPU / 4 GB
Both can keep a box running all month. boat adds preinstalled agents you drive from outside, and an optional paid proxy that keeps model keys out of the machine.
6th cheapest of 361
$657 a month
53rd of 361, with limits
$628 a month
Code interpreter bursts
100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month
shellbox fits short bursts well: a box pauses with its memory, comes back in seconds, and bills per minute with no plan fee. boat's $20 plan includes 12 starts a minute and 200 a day; extra starts cost $0.03 to $0.07 each.
189th of 336, with limits
$2,000 a month
73rd of 336, with limits
$67 a month
Sandboxes for your users
300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month
shellbox's account caps stop well short of hundreds of user sandboxes, and its terms forbid reselling or sublicensing access to the service.
3rd cheapest of 349
$209 a month
43rd of 349, with limits
$211 a month
RL rollouts
1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month
shellbox copies a stopped box with its memory snapshot, so rollouts can branch from the same running state. boat forks a running sandbox from its last disk save, without memory or running programs.
39th of 349, with limits
$2,000 a month
19th of 349, with limits
$1,000 a month
Agent evals
250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month
10th cheapest of 359
$200 a month
102nd of 359, with limits
$160 a month
CI runners
40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month
2nd cheapest of 374
$86 a month
144th of 374, with limits
$96 a month
24/7 dev boxes
1 machine 24/7 · 4 vCPU / 8 GB
This is shellbox's home ground: a box that pauses when you log off, keeps its processes, and works with VS Code Remote, Zed and mosh.
19th cheapest of 371
$26 a month
17th cheapest of 371
$25 a month
Computer-use agents
50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month
Computer-use agents need a screen. boat ships a 60 fps Linux desktop with Chrome ready to use. On shellbox you set it up: install a desktop, a VNC server and a browser, then reach them through SSH port forwarding.
7th cheapest of 367
$200 a month
99th of 367, with limits
$147 a month
Browser-use agents
50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month
boat has Chrome and a desktop in every sandbox. On shellbox you set it up: install Chromium with apt, headless or on a desktop you add.
7th cheapest of 356
$200 a month
168th of 356, with limits
$80 a month
Scraping & bots (startup scale)
50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month
6th cheapest of 343
$90 a month
122nd of 343, with limits
$98 a month
Mass scraping & bots (undetectable)
1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month
73rd of 327, with limits
$1,800 a month
78th of 327, with limits
$1,950 a month
Parallel red teaming
200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month
3rd cheapest of 351
$200 a month
34th of 351, with limits
$49 a month
GPU eval jobs
10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month
Not ranked: misses a requirementNot ranked: misses a requirement
macOS agents
5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month
Not ranked: misses a requirementNot ranked: misses a requirement

Key features

Featureboatshellbox
IsolationFull VM with its own kernel, on a physical host dedicated to boat; vCPUs are sharedFirecracker microVM by default, or Cloud Hypervisor; own kernel and disk per box
Biggest machine16 vCPU / 32 GB with 251 GB disk, on the $100 plan and above; disk is fixed per size (51 GB at 4 vCPU)16 vCPU / 32 GB with 400 GB disk (x8), no plan needed; 100 GB disk at 4 vCPU; size is fixed at creation
Machines at once100 at once on the $20 plan (200 starts a day included, extra starts $0.03 to $0.07), up to 2,000 at once on the $2,000 plan, more with extra seats16 running slots per account, so 8 boxes of 4 vCPU; the site publishes no way to raise it
Pause and resumeDisk is saved every minute and at each stop; memory and running programs are not saved, so a resume is like a rebootPauses with a memory snapshot; reconnect in about 3 seconds with processes and shell history intact
Stopping when idleAuto-stop timer, 1 hour by default, up to 30 days or off on a paid plan; it does not watch for idlePauses 120 seconds after the last SSH session ends; wakeup mode stops after idle
Wake on requestNo; you start a stopped sandbox by API or CLIYes; wakeup mode starts a box on an HTTPS request, and cron mode wakes it on a schedule
ForkFork a running sandbox from its last disk save (taken every minute); memory and processes are not copiedDuplicate a stopped box with its disk and memory snapshot; live duplicate is an open feature request
Nested virtualizationYes, /dev/kvm in every sandbox for Android emulators and VMsYes on Cloud Hypervisor boxes (--ch); not on the default Firecracker backend
Docker and rootRoot, systemd and Docker in every sandboxRoot and Docker on stock Ubuntu boxes; boxes made from OCI images do not include Docker
Custom imageNo; you start from boat's image or a saved snapshotYes; create a box from a public OCI image (linux/amd64, no private registries yet)
Inbound accessHTTPS URLs for up to 50 ports, token-protected by default; raw ports on the sandbox IP (IPv6 on every sandbox, IPv4 on some), which changes at each resumeOne public HTTPS URL to port 80, custom domains, and a stable public IPv6 for raw TCP and UDP; no IPv4
Desktop and browser1920x1080, 60 fps Linux desktop and Chrome, with computer tools for agentsNot built in; you install a desktop (for example Xvfb and VNC) and a browser yourself and reach them through SSH port forwarding
Coding agentsSeven harnesses preinstalled, Claude Code and Codex among them, with a prompt, steer and events APIYou set it up: install your agent, start it in tmux over SSH and tell it to stop the box when done
Keeping API keys out of the sandboxOptional credentials proxy ($1 a day while on, plus $1 per 1,000 requests) for model providers and custom HTTPS APIs; switch a credential to another account when credits run out; other secrets still enter the sandboxNo secret store; docs suggest SSH agent forwarding, pasting a token, or adding an SSH key inside the box
RegionsGermany, Finland and France; you cannot pick one; no US or Asia regionHelsinki, Finland only; a US region is an open request
Compliance and transparencySOC 2 is in progress. We're open about where we stand: live status and data retention policy.No SOC 2 report published
Every feature Battleships tracks (43)
Featureboatshellbox
Full VM (own kernel)YesYes
Docker insideYesYes
Nested virtualizationYesYes
RootYesYes
SystemdYesNot published
Desktop GUIYesNot published
BrowserYesNot published
Computer-use APIYesNo documented way
SSHYesYes
Persistent diskYesYes
Pause/resumeYesYes
Automatic snapshotsYesYes
Fork/cloneYesYes
Live fork (no pause)YesNo documented way
Memory forkNo documented wayYes
Custom image (Docker or snapshot)YesYes
≥24 h sessionsYesYes
Idle auto-stopNo documented wayYes
Wake on requestNo documented wayYes
HTTPS preview URLsYesYes
Raw TCP inboundYesYes
Public IPv4No documented wayNo documented way
Egress allowlistNo documented wayNo documented way
Secret proxyYesNo documented way
Secret proxy for any APIYesNo documented way
Preinstalled agentsYesNo documented way
Agent harness APIYesNo documented way
MCP serverNo documented wayNo documented way
GPUNo documented wayNo documented way
WindowsNo documented wayNo documented way
Arm64No documented wayNo documented way
Self-hostingNo documented wayNo documented way
BYOCNo documented wayNo documented way
Open sourceNo documented wayNo documented way
Scoped API keysYesNot published
Spend limitsYesNot published
Usage per sandbox (API)YesNo documented way
Audit logsNo documented wayNot published
SSONo documented wayNo documented way
SOC 2No documented wayNo documented way
HIPAANo documented wayNo documented way
EU data residencyNo documented wayYes
Zero data retentionYesNot published

Performance

boat28.06 runs/s
shellboxNot measured

shellbox is not in the public benchmarks we use, so we cannot say which is faster. In StarSling's run of 3 October 2026 (run 37073680030), boat did 28.06 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB, second of 15. That boat sandbox ran on an AMD Ryzen 9 9950X host; boat does not promise a CPU model.

V8 Web Tooling Benchmark at a 4 vCPU / 8 GB target, runs per second, higher is better. Source: StarSling's run of 3 October 2026 (run 37073680030), median of 5 sandboxes. Providers ran on different hardware in different regions. StarSling gives a shared rank when the run cannot separate two providers. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. shellbox is not in the public benchmarks we use, so we cannot say which is faster.

Pricing in five scenarios

Each bill is priced from the providers' public price lists by the open Battleships engine, with the cheapest plan that runs the scenario. Start limits, per-start fees and plan minimums are in the price, boat's included.

Scenarioboat, a monthshellbox, a month
One coding agent, workdays
4 vCPU / 8 GB, 50 GB disk, 8 hours a day, 22 days a month, files kept between days
$20 · $20/mo$6 · Prepaid
One agent that never stops
4 vCPU / 8 GB, 50 GB disk, running 24/7 all month
$26 · $20/mo$26 · Prepaid
50 agents in parallel
50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside
$317 · $20/mo × 3 seats$317 · Prepaid · with limits
Sandboxes for your users
2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once
$200 · $100/mo × 2 seats$90 · Prepaid · with limits
1,000 short jobs at once
2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month
$8,800 · $2000/mo · with limits$200 · Prepaid · with limits

boat: Plans cost $20, $100, $500 or $2,000 a month, and the whole plan price comes back as VM time for that month. Unused time expires at the end of the month. A 4 vCPU / 8 GB sandbox costs $0.036 an hour, billed per second while it runs. The $20 plan includes 200 starts a day; extra starts cost $0.03 to $0.07 each. A stopped sandbox has no compute charge and keeps its disk for at least 14 days; save a snapshot to keep it longer (10 saved snapshots included per $20, $0.056 a day for each extra one).

shellbox: shellbox is prepaid with no plan fee and a $10 minimum top-up. A running box costs $0.02 an hour per x1 slot, billed per minute, so $0.04 an hour at 4 vCPU / 8 GB. A paused box costs $0.50 a month per slot for its memory snapshot and disk. Boxes stop when the balance drops below $5 and are deleted at $0.

Limits in those scenarios

shellbox allows 16 running slots and 64 total slots per account, and a 4 vCPU box uses two, so only 8 such boxes run at once. That is why the 50-agent fleet and the 300 user sandboxes do not fit, and its site names no way to raise the cap. shellbox also says bandwidth is shared and may be throttled, and its terms bar resource-intensive background processes. boat's limits: the $20 plan includes 12 starts a minute, 60 an hour and 200 a day, and a start above 6 times that fails. boat never queues a start. Neither fits 1,000 jobs at once: boat's $2,000 plan includes 2,400 starts a day, charges for starts above 90 a minute and refuses them above 270. A boat resume brings back the disk only, with a new IP, and a machine failure loses what was written after the last 1-minute save.

Scenarioboatshellbox
One coding agent, workdaysNone for this scenario
  • egress price not published
One agent that never stopsNone for this scenario
  • egress price not published
50 agents in parallelNone for this scenario
  • Prepaid: max 16 concurrent
  • Prepaid: max 32 vCPU running at once (needs 200)
  • egress price not published
Sandboxes for your usersNone for this scenario
  • Prepaid: max 16 concurrent
  • Prepaid: max 32 vCPU running at once (needs 600)
  • egress price not published
1,000 short jobs at once
  • 2,400 starts/day included, paid above that (needs ~6,667)
  • 600 starts/hour included, paid above that (needs ~1,000)
  • starts above 90/min are paid, and starts above 270/min fail with 429
  • Prepaid: max 16 concurrent
  • Prepaid: max 32 vCPU running at once (needs 2,000)
  • egress price not published
boat's own limits
  • Plans start at $20 a month. The plan price comes back as VM time for that month. Unused time does not carry over. Source
  • Starts (create, fork, resume): the $20 plan includes 12 a minute, 60 an hour and 200 a day. The $2,000 plan includes 90 a minute, 600 an hour and 2,400 a day. Extra starts cost $0.03 to $0.07. A start fails with 429 above 3 times the included starts per minute, or 6 times per hour or per day. Source
  • A stopped sandbox has no compute charge. boat keeps its disk for at least 14 days. A saved snapshot keeps it longer; extra saved snapshots and the AI credentials proxy are paid. Source
  • Auto-stop defaults to 1 hour. You can turn it off on a paid plan. The trial allows 2 hours at most. Source
  • boat saves the disk every minute and at each stop. Memory, running programs and /tmp are not saved: a resume is like a reboot. A machine failure loses writes after the last good snapshot. Source
  • Machines run in Germany, Finland and France only. Snapshot copies can be stored outside the EU. Source
  • Public IPv6 and outbound IPv4. No dedicated IPv4. The IP changes at each resume. Source
  • Hosts are AMD Ryzen 7 9700X and Ryzen 9 9950X. vCPUs are shared. No GPU. You cannot boot your own Docker image. Source

Which one should you use?

Use shellbox when

you want one or a few personal SSH boxes that pause with their memory when you log off, small services that wake only on a request, or a box booted from your own OCI image.

  • Pauses with its memory. Useful for dev boxes you come back to: processes, shell history and TCP state are restored in about three seconds.
  • Wakes on a request or a schedule. Good for small web services and periodic jobs that should only run, and bill, while there is work.
  • Plain SSH, no account. Fits people who live in a terminal: your SSH key is the account, and VS Code Remote, Zed and mosh just work.

Use boat when

you run many agents at once and want them preinstalled and driven from outside, with a desktop and browser ready, an optional proxy that keeps model keys out of the machine, and hosting on many ports.

  • Agents and keys built in. On shellbox you install the agent and start it in tmux over SSH. Its docs suggest SSH agent forwarding, pasting tokens or adding an SSH key inside the box, where a prompt injection can reach them.
  • Fleets, not a handful of boxes. Fifty parallel agents or a few hundred user sandboxes do not fit in shellbox's 16 slots, and its site publishes no way to raise them.
  • A screen and more ports. shellbox routes its HTTPS URL to port 80 only, so an app with an API, a frontend and a preview needs a proxy you set up. Watching a browser agent there means installing a desktop and VNC and tunnelling to it yourself.

Questions people ask

Is boat cheaper than shellbox?

Not for one agent on workdays: a 4 vCPU shellbox box costs under $10 a month, and boat's smallest plan is $20. Per hour, boat costs $0.036 and shellbox $0.04 at 4 vCPU / 8 GB, so a box running all month comes to about $26 on boat and $29 on shellbox. A shellbox box that pauses while idle bills only $0.50 a month per slot for that time. For 50 agents, 50 boxes of 4 vCPU need 100 slots, past shellbox's 16 running and 64 total.

How many boxes can I run at once on shellbox?

Up to 16 running slots and 64 total slots per account, where a box uses as many slots as its size. That is 16 boxes of 2 vCPU, or 8 of 4 vCPU. boat runs 100 sandboxes at once on the $20 plan, with 200 starts a day included.

Can I run Claude Code on shellbox?

Yes. You install it in the box, start it in a tmux session over SSH, and can tell it to stop the box when it is done. On boat, Claude Code and six other agent harnesses are preinstalled, and you prompt and steer them through an API.

Does shellbox have an API?

Yes, over SSH: every command can return stable JSON, and your SSH key authenticates each call. There is no REST endpoint or SDK. boat has a REST API, SDKs and a CLI.

Do both support Docker and nested virtualization?

Yes. shellbox stock boxes ship with Docker, and boxes created with --ch on Cloud Hypervisor expose /dev/kvm. boat has Docker and /dev/kvm in every sandbox. shellbox can also boot a public OCI image; boat starts only from its own image or a saved snapshot.

What happens to my files when a box stops?

shellbox keeps the disk and a memory snapshot while the box is paused, and has no backups; a box is deleted when the balance reaches $0. boat saves the disk every minute and at each stop, but not memory or running programs. It keeps a stopped disk for at least 14 days, and a saved snapshot keeps it longer. If a boat machine fails, you lose what was written after the last good save.

Is boat SOC 2 compliant?

No. SOC 2 is in progress, and boat has no SOC 2 report today. shellbox documents no SOC 2 either. Live status is at status.ascii.dev and our data retention policy is at docs.boat.dev/data-retention.

Start a sandbox

Install the CLI, then run boat new.

7-day free trial

More comparisons