boat vs Hetzner Dedicated

A Hetzner dedicated server is a whole physical machine that you order, install and look after yourself, and it bills until you cancel it. boat gives each agent its own full VM by API call, sized to the job, with snapshots every minute, stops that cost nothing, fork of a running machine and coding agents already built in.

Where boat stands out

A machine per agent in seconds, sized to the job

On boat one API call gives an agent its own VM, from 2 to 16 vCPU, and you can run 100 to 2,000 at once by plan. On Hetzner each order is a whole physical server, a first order is checked by hand, and the smallest one has 12 threads and 64 GB.

Free stops, snapshots every minute, fork while running

boat saves a sandbox's disk every minute and when it stops, and a stopped sandbox costs nothing. You can fork a running sandbox in one call. A Hetzner server bills while it is off until you cancel it, has no snapshots of its own, and a copy means ordering and installing another server.

Coding agents and a desktop built in

Every boat sandbox has Claude Code, Codex, OpenCode and other agents ready, and you drive agents inside from outside with a prompt, steer and events API. It also streams a 60fps desktop with Chromium for computer use. The credentials proxy keeps real AI keys out of the machine and can switch to another account when one runs out of credits.

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.

Hetzner Dedicated

Hetzner is a German hosting company founded in 1997; its dedicated root servers are whole physical machines rented to one customer.

You pick a server from the AX (AMD), EX (Intel), SX (storage), DX (Dell) or GPU lines, or a used one from the Server Auction, and get root on the bare hardware with a public IPv4 and unmetered traffic on a 1 Gbit port. You install the OS yourself through the Rescue System, installimage or your own ISO, and manage it through the Robot panel and API. It is a server rental, not an agent sandbox, so this page compares the two where they overlap: giving agents their own Linux machines.

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. “3rd cheapest of 367” means 2 of the other 366 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 rankHetzner Dedicated: price rank
Software factory
100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month
Both give root, Docker and /dev/kvm, but on Hetzner you rent a whole server per agent or split it into VMs yourself, while boat hands each agent its own VM by API call.
3rd cheapest of 367
$634 a month
175th of 367, with limits
$1,918 a month
Always-on AI employees
50 machines 24/7 · 2 vCPU / 4 GB
A Hetzner server can run assistants around the clock, but you build the agent setup, the HTTPS endpoints and the backups yourself; boat has them built in.
5th cheapest of 366
$657 a month
248th of 366, with limits
$3,400 a month
Code interpreter bursts
100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month
Neither fits one-minute runs well: a Hetzner server is ordered, set up and billed by the hour per machine, and boat's start limits make big bursts a compromise.
186th of 345, with limits
$2,000 a month
199th of 345, with limits
$6,800 a month
Sandboxes for your users
300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month
Handing each of your users a sandbox on Hetzner means renting servers and building the VM layer, snapshots and HTTPS yourself; boat does that with one API call per sandbox.
Cheapest of 361
$209 a month
313th of 361, with limits
$3,270 a month
RL rollouts
1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month
34th of 360, with limits
$2,000 a month
208th of 360, with limits
$68,000 a month
Agent evals
250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month
10th cheapest of 364
$200 a month
174th of 364, with limits
$1,635 a month
CI runners
40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month
Both run Docker builds on real Linux; a Hetzner server makes a strong always-on runner if you run the runner software and clean-up yourself.
4th cheapest of 377
$86 a month
76th cheapest of 377
$645 a month
24/7 dev boxes
1 machine 24/7 · 4 vCPU / 8 GB
Both fit a 24/7 dev box, and a Hetzner server gives far more cores and memory per box; boat's box is smaller but keeps its disk with snapshots every minute.
18th cheapest of 368
$26 a month
121st of 368, with limits
$68 a month
Computer-use agents
50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month
boat streams a 60fps desktop with Chromium and computer use; on Hetzner you install the desktop, the browser and the control loop yourself.
6th cheapest of 369
$200 a month
308th of 369, with limits
$2,180 a month
Browser-use agents
50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month
7th cheapest of 367
$200 a month
348th of 367, with limits
$3,270 a month
Scraping & bots (startup scale)
50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month
6th cheapest of 346
$90 a month
294th of 346, with limits
$654 a month
Mass scraping & bots (undetectable)
1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month
71st of 338, with limits
$1,800 a month
255th of 338, with limits
$13,080 a month
Parallel red teaming
200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month
2nd cheapest of 360
$200 a month
280th of 360, with limits
$545 a month
GPU eval jobs
10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month
Hetzner rents GPU servers with Nvidia RTX PRO Blackwell cards by the month, and boat has no GPUs.
Can't run itCan't run it
macOS agents
5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month
Can't run itCan't run it

Key features

FeatureboatHetzner Dedicated
IsolationEvery sandbox is a full VM with its own kernelA whole physical server for one customer, with your own kernel on the hardware
CPUDedicated AMD Ryzen 9 9950X (Zen 5) hosts; 26.69 runs/s at 4 vCPU / 8 GBRyzen 7 PRO 8700GE (AX42), Ryzen 9 7950X3D (AX102), EPYC 9454P (AX162) and Intel lines; no comparable measurement
Machine sizes2 to 16 vCPU and up to 32 GB per sandbox, picked per sandboxA whole server: from 12 threads / 64 GB (AX41) up to 96 threads (AX162) or 172 threads (DX294)
Getting a new machineOne API call creates a sandbox in secondsAn order per physical server; first orders are checked and may need ID or an advance payment
Docker and nested virtualizationDocker, root, systemd and /dev/kvm on every machineRoot on the hardware, so Docker and KVM work once you install them
StoppingA stopped sandbox costs nothing and keeps its diskA turned-off server bills until you cancel it
SnapshotsAutomatic snapshots every minute and on stopNone for the server; you order a Storage Box and run your own backups
ForkFork a running sandbox in one call (disk, no memory)None; you order another server and install it again
Reaching the machineHTTPS URLs for up to 50 ports, token-protected; raw TCPPublic IPv4 on any port and unmetered traffic on 1 Gbit; HTTPS is your own web server and certificates
Desktop and computer use60fps desktop and Chromium, computer use and browser control, embeddable streamNone built in; a KVM console for admin, and you install a desktop yourself
Coding agentsClaude Code, Codex, OpenCode and more preinstalled, with a prompt, steer and events APINone; you install and drive agents yourself, for example over SSH
Keeping keys out of the machineAI credentials proxy for model providers and any HTTPS API you register; switch a credential to another account when credits run outNo credentials proxy; secrets live on the server you run
Outbound rulesOpen internet on every plan; no provider-side allowlistFree Robot firewall on the switch port: stateless, up to 10 rules each for incoming and outgoing traffic
GPUs and WindowsNo GPUs; no Windows guests (Wine for Windows apps)GPU servers with Nvidia RTX PRO 4000 and 6000 Blackwell cards; Windows Server 2025 as an add-on
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)
FeatureboatHetzner Dedicated
Full VM (own kernel)YesYes
Docker insideYesYes
Nested virtualizationYesYes
RootYesYes
SystemdYesYes
Desktop GUIYesNo
BrowserYesNo
Computer-use APIYesNo
SSHYesYes
Persistent diskYesYes
Pause/resumeYesNo
Automatic snapshotsYesNo
Fork/cloneYesNo
Live fork (no pause)YesNo
Memory forkNoNo
Custom image (Docker or snapshot)YesYes
≥24 h sessionsYesYes
Idle auto-stopNoNo
Wake on requestNoNo
HTTPS preview URLsYesNo
Raw TCP inboundYesYes
Public IPv4NoYes
Egress allowlistNoYes
Secret proxyYesNo
Secret proxy for any APIYesNo
Preinstalled agentsYesNo
Agent harness APIYesNo
MCP serverNoNo
GPUNoYes
WindowsNoYes
Arm64NoNo
Self-hostingNoNo
BYOCNoNo
Open sourceNoNo
Scoped API keysYesNot published
Spend limitsYesNot published
Usage per sandbox (API)YesNo
Audit logsNoNot published
SSONoNo
SOC 2NoNo
HIPAANoNo
EU data residencyYesYes
Zero data retentionYesNot published

Performance

boat26.69 runs/s
Hetzner DedicatedNot measured

There is no comparable measurement for Hetzner dedicated servers: they were not part of the public StarSling benchmark run where boat did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB. Results would depend on the model you rent, from the older Ryzen 5 3600 in the AX41 to the Zen 4 Ryzen 9 7950X3D in the AX102.

Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Hetzner Dedicated has no published run in this suite yet, so there is no like-for-like number to compare.

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.

Scenarioboat, a monthHetzner Dedicated, 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$19 · Usage account · with limits
One agent that never stops
4 vCPU / 8 GB, 50 GB disk, running 24/7 all month
$26 · $20/mo$67 · Usage account · with limits
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$946 · Usage account · 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$3,225 · Usage account · with limits
1,000 short jobs at once
2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month
$2,000 · $2000/mo · with limits$21,500 · Usage account · with limits

boat: You pick a monthly plan from $20, and all of it becomes usage credit spent per second on running sandboxes. A stopped sandbox costs nothing and keeps its disk.

Hetzner Dedicated: Hetzner bills each server by the hour, rounding up, until it reaches the monthly price, and keeps billing while it is off until you cancel it. Most models add a one-time setup fee per order, from $59 on the AX42 to $359 on the AX162, and the public IPv4 is $1.90 a month on top. The prices here rent one whole AX41 (12 threads, 64 GB, no setup fee) per agent.

Limits in those scenarios

The smallest Hetzner server priced here has 12 threads and 64 GB, so every scenario rents a whole server per agent and none fit the size asked for; you could split one server into several VMs, but then you build and run that layer yourself. Hetzner publishes no cap on how many servers you can rent, but it checks first orders for plausibility and may hold large ones, so 50 or 300 servers at once is a conversation, not a click. On boat the limit is a sandbox count (100 to 2,000 by plan) plus starts per minute, so the 1,000-job burst gets queued and needs more starts per day than the $2,000 plan allows.

ScenarioboatHetzner Dedicated
One coding agent, workdaysNone that bite
  • smallest size offered is 12 vCPU / 64 GiB
One agent that never stopsNone that bite
  • smallest size offered is 12 vCPU / 64 GiB
50 agents in parallelNone that bite
  • smallest size offered is 12 vCPU / 64 GiB
Sandboxes for your usersNone that bite
  • smallest size offered is 12 vCPU / 64 GiB
1,000 short jobs at once
  • max 2,400 starts/day (needs ~6,667)
  • max 600 starts/hour (needs ~1,000)
  • bursts above 90 starts/min are queued (a burst of 333 takes ~4 min to start)
  • smallest size offered is 12 vCPU / 64 GiB

Which one should you use?

Use Hetzner Dedicated when

you want a few big, steady machines you run yourself, with GPUs, Windows Server, unmetered traffic or a firewall in front of the hardware.

  • Big whole machines with unmetered traffic. Useful for heavy, steady work: one server goes up to 96 threads, and traffic on the 1 Gbit port is unmetered.
  • GPU servers. Useful if your agents train or serve models: Hetzner rents servers with Nvidia RTX PRO 4000 and 6000 Blackwell cards.
  • Windows and a firewall in front of the server. Useful if the machine must run Windows Server, or be fenced off by rules outside the OS itself.

Use boat when

you want each agent on its own full VM that is ready to work in seconds: /dev/kvm, coding agents, a desktop, HTTPS ports and minute-by-minute snapshots built in, fork while running, and stops that cost nothing.

  • A machine per agent in seconds, sized to the job. Agent work comes in bursts of many small machines. On Hetzner you either rent whole servers per agent or build and run your own VM layer on them, with its scheduling, images and clean-up.
  • Free stops, snapshots every minute, fork while running. Agents break their machines and sit idle between tasks. On Hetzner a broken install means a reinstall from your own backups, and an idle server keeps billing until you cancel it and lose its disk.
  • Coding agents and a desktop built in. A Hetzner server starts as an empty OS install, so you set up, update and supervise each agent, desktop and key yourself, and every agent on the server can read any key you put there.

Questions people ask

Is boat cheaper than a Hetzner dedicated server?

Per agent, mostly yes, because each Hetzner scenario rents a whole AX41: one agent on workdays comes to $20 on boat and $19 on Hetzner, one agent that never stops comes to $26 on boat against $67, and 50 agents in parallel come to $317 against $946. If you pack several agents onto one big server and run the VM layer yourself, Hetzner can cost less per agent.

Can I run nested virtualization on a Hetzner dedicated server?

Yes. You have root on the whole physical machine, so KVM, Docker and your own VMs work once you install them. On boat, Docker, systemd and /dev/kvm are ready in every sandbox.

Could I build my own agent sandboxes on Hetzner servers?

Yes, with KVM or Firecracker on top of the servers you rent. You would then build the images, snapshots, forks, HTTPS routing, key handling and scheduling yourself, and run them day to day. boat is that layer, already built, with coding agents and a desktop in every sandbox.

Do I pay for a Hetzner server that is turned off?

Yes. Hetzner bills a product for as long as it is on your account, even when it is turned off, until you cancel it. On boat, a stopped sandbox costs nothing and keeps its disk.

Where does my data run?

Hetzner's dedicated servers run in Falkenstein and Nuremberg (Germany) and Helsinki (Finland). boat runs only in the EU (Germany, Finland and France), with EU data residency and a zero data retention option.

Is boat SOC 2 compliant?

SOC 2 is in progress. We're open about where we stand: live status at status.ascii.dev and our data retention policy at docs.boat.dev/data-retention.

Start a sandbox

Install the CLI, then run boat new.

7-day free trial

More comparisons