boat vs Replicate

These are different kinds of product. boat gives each agent its own full Linux machine to work inside, while Replicate runs AI models behind an API, on GPUs that scale up with requests and down to zero.

Where boat stands out

A whole machine to work inside

Each boat sandbox is a full VM with its own kernel, root, SSH, systemd and Docker, on dedicated AMD Ryzen 9 9950X hosts. It also exposes /dev/kvm, so Android emulators and VMs run inside it. Replicate runs one model container that you reach through its prediction API, with no shell to log into.

The whole disk stays, and you can fork it

boat saves the disk every minute and when the sandbox stops, and a stopped sandbox costs nothing and keeps every file. You can also fork a running sandbox into a copy with the same disk. A Replicate instance starts fresh from its image each time, and there is no snapshot or fork.

Agents, desktop and keys built in

Claude Code, Codex and five other agents come preinstalled, and you prompt, steer and follow them through one API. Each sandbox has a 60fps desktop with Chromium for computer use. Real API keys stay outside the sandbox in a credentials proxy that works with any HTTPS API.

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.

Isolation: Full VM per sandbox, own kernel
Hardware: Dedicated AMD Ryzen 9 9950X (Zen 5) hosts
Regions: EU: Germany, Finland, France
Plans: From $20 a month, all of it usage credit, 7-day trial

Replicate

Replicate runs AI models with an API: thousands of community and official models, plus your own models packaged with Cog.

You call a public model with one API request, fine-tune one on your data, or package your own model with Cog, Replicate's open-source tool that builds a Docker image with an HTTP prediction server. Replicate runs it on CPU or NVIDIA GPU hardware as deployments that scale up with traffic and down to zero. It is built for teams that serve models, not for agents that need a machine of their own, so this page compares the places where the two overlap.

Company: Founded 2019, Y Combinator Winter 2020, San Francisco; joined Cloudflare in November 2025 and continues as its own brand with the same API
Product: Model inference API: public models, fine-tunes, and custom Cog models run as autoscaling deployments
Isolation: Each model runs as a Docker container built by Cog; the docs do not name the isolation layer
Hardware: CPU (1 or 4 vCPU), T4, L40S, A100 80GB, H100 and H200, up to 8 GPUs per instance; most multi-GPU and all H200 need a committed spend contract
Run time: Each prediction times out after 30 minutes unless support raises it
Hosting: Infrastructure subprocessors (AWS, CoreWeave, GCP) listed in the United States; no region choice documented
Open source: Cog is open source and its images run anywhere Docker runs; the hosted platform is not

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 workload	boat: price rank	Replicate: price rank
Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month A software factory needs Docker, root, a shell and a disk that is still there tomorrow; a Replicate model is a stateless container you reach through its prediction API.	3rd cheapest of 367 $634 a month	268th of 367, with limits $4,340 a month
Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An assistant that lives on one machine all month keeps its whole disk on boat; a Replicate instance starts fresh from its image and each prediction stops at 30 minutes.	5th cheapest of 366 $657 a month	335th of 366, with limits $11,300 a month
Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Short runs fit Replicate's model: a request starts work on an instance that bills per second and scales back to zero when traffic stops.	186th of 345, with limits $2,000 a month	258th of 345, with limits $295 a month
Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month boat gives each of your users a full VM with a lasting disk and hosting; Replicate gives them a model endpoint that scales.	Cheapest of 361 $209 a month	317th of 361, with limits $3,705 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	174th of 360, with limits $17,900 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	67th cheapest of 364 $1,440 a month
CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month CI jobs that build containers or Android apps need Docker and /dev/kvm, which every boat sandbox has and Replicate does not document.	4th cheapest of 377 $86 a month	311th of 377, with limits $864 a month
24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A dev box needs SSH and a disk that keeps your installs; Replicate has no SSH and no persistent disk.	18th cheapest of 368 $26 a month	306th of 368, with limits $226 a month
Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month boat ships a Linux desktop streamed at 60fps with Chromium for computer use; Replicate has no desktop or browser product.	6th cheapest of 369 $200 a month	263rd of 369, with limits $1,296 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	322nd of 367, with limits $1,296 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	272nd of 346, with limits $439 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	221st of 338, with limits $8,775 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	306th of 360, with limits $855 a month
GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month Replicate has T4, L40S, A100, H100 and H200 GPUs billed per second; boat is CPU only.	Can't run it	22nd cheapest of 98 $1,338 a month
macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month	Can't run it	Can't run it

Key features

Feature	boat	Replicate
What you get	A full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts, which you or an agent work inside	A model packaged as a Cog container and called through a prediction API
Shell access	SSH, root and systemd, like a normal Ubuntu server	No SSH; you call the model through the HTTP API or client libraries
Docker and nested virtualization	Docker inside, and /dev/kvm for Android emulators and VMs	The model is the container; Docker inside and /dev/kvm are not documented
Files after a stop	The whole disk is saved every minute and on stop, and a stopped sandbox keeps it for free	No persistent disk; instances start fresh from the image, and API inputs and outputs are deleted after an hour by default
Fork and snapshots	Fork a running sandbox with its disk	No snapshots or fork; you add replicas of the same image
Run time	No session cap on paid plans; you set the time to live	Instances can stay on, but each prediction times out after 30 minutes unless support raises it
Scale to zero and wake	No; a stopped sandbox is started by API	Yes; deployments scale to zero and start again on the next prediction, with min and max instances you set
GPUs	No, CPU only	T4, L40S, A100 80GB, H100 and H200, up to 8 per instance
Ready-made models	None; you bring the model API you call	Thousands of public and official models for images, video, speech and text, plus fine-tuning
Inbound access	HTTPS URLs for up to 50 ports, token-protected by default; raw TCP; dedicated IPv4 or IPv6	One prediction API per model behind API token auth; no own IP or ports
Desktop and browser	Linux desktop streamed at 60fps with Chromium and computer use	No desktop or browser product
Coding agents	Seven agents preinstalled, with a prompt, steer and events API	None to run inside; Replicate hosts models your agent can call
Keeping API keys out of the machine	Credentials proxy for model providers and any HTTPS API; switch a credential to another account when credits run out	Secrets are sent to the model as inputs and redacted from logs afterwards
Self-hosting	No	Cog images run on your own infrastructure, anywhere Docker runs
Compliance and transparency	SOC 2 is in progress. We're open about where we stand: live status and data retention policy.	Not published

Every feature Battleships tracks (43)

Feature	boat	Replicate
Full VM (own kernel)	Yes	No
Docker inside	Yes	No
Nested virtualization	Yes	No
Root	Yes	Not published
Systemd	Yes	Not published
Desktop GUI	Yes	No
Browser	Yes	No
Computer-use API	Yes	No
SSH	Yes	No
Persistent disk	Yes	No
Pause/resume	Yes	No
Automatic snapshots	Yes	No
Fork/clone	Yes	No
Live fork (no pause)	Yes	No
Memory fork	No	No
Custom image (Docker or snapshot)	Yes	Yes
≥24 h sessions	Yes	Yes
Idle auto-stop	No	Yes
Wake on request	No	Yes
HTTPS preview URLs	Yes	Yes
Raw TCP inbound	Yes	Not published
Public IPv4	No	No
Egress allowlist	No	No
Secret proxy	Yes	No
Secret proxy for any API	Yes	No
Preinstalled agents	Yes	No
Agent harness API	Yes	No
MCP server	No	Yes
GPU	No	Yes
Windows	No	No
Arm64	No	No
Self-hosting	No	Yes
BYOC	No	No
Open source	No	Yes
Scoped API keys	Yes	Not published
Spend limits	Yes	Yes
Usage per sandbox (API)	Yes	No
Audit logs	No	Yes
SSO	No	No
SOC 2	No	Not published
HIPAA	No	Not published
EU data residency	Yes	No
Zero data retention	Yes	Not published

Performance

boat26.69 runs/s

ReplicateNot measured

There is no comparable measurement: Replicate has no result on the StarSling hpc-sandbox-benchmarks CPU test, so we cannot say which is faster at CPU work. boat did 26.69 runs per second on that test at 4 vCPU / 8 GB. Replicate's hardware is chosen for GPU inference, which this test does not cover.

Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Replicate 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.

Scenario	boat, a month	Replicate, 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	$57 · Usage · with limits
One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month	$26 · $20/mo	$237 · Usage · 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	$2,851 · Usage · 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	$1,620 · Usage
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	$3,580 · Usage

boat: Plans start at $20 a month, and the whole fee is usage credit. You pay per second while a sandbox runs, and a stopped sandbox costs nothing and keeps its disk.

Replicate: Replicate bills per second by hardware tier, from prepaid credit that expires after a year. Public models bill only while they run; private models and deployments bill setup, idle and active time, including a few idle minutes after traffic stops, and failed or canceled runs.

Limits in those scenarios

Replicate does not fit the workday agent, the always-on agent or the 50-agent fleet, because it has no persistent disk: an agent's files are gone when the instance scales down, and the fleet also needs Docker inside, which is not documented. Each prediction stops after 30 minutes unless support raises the limit, so long agent tasks have to be split. Replicate does not publish an egress or disk price, so the scenarios count none. boat's start limits are lower: at 1,000 short jobs at once, its $2,000 plan queues starts, while Replicate fits the burst.

Scenario	boat	Replicate
One coding agent, workdays	None that bite	missing persistent disk no persistent disk: files lost when it stops or sleeps disk beyond 0 GiB: price unknown egress price not published
One agent that never stops	None that bite	missing persistent disk no persistent disk: files lost when it stops or sleeps disk beyond 0 GiB: price unknown egress price not published
50 agents in parallel	None that bite	missing persistent disk, Docker inside no persistent disk: files lost when it stops or sleeps disk beyond 0 GiB: price unknown egress price not published
Sandboxes for your users	None that bite	disk beyond 0 GiB: price unknown egress price not published
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)	disk beyond 0 GiB: price unknown egress price not published

Which one should you use?

Use Replicate when

you need to call or serve an AI model on GPUs, from its catalog or packaged with Cog, as an API that scales with traffic.

GPUs for running models. For model inference or fine-tuning, Replicate offers T4 to H200 GPUs, up to 8 in one instance, billed per second.
Thousands of models, one API call away. Teams that want to call an image, video, speech or language model without hosting it can start with one request.
Scale to zero, and Cog runs anywhere. Models that wait for traffic can scale to zero and wake on the next prediction, and Cog images run on your own servers too.

Use boat when

your agents need a machine of their own: a shell, root, Docker, emulators and a browser, on a full VM whose whole disk is still there the next time it starts.

A whole machine to work inside. A coding agent installs tools, starts services and runs docker build as it goes. On Replicate every step would have to be a prediction call that stops at 30 minutes, and Docker or /dev/kvm work has no documented way to run.
The whole disk stays, and you can fork it. An agent that clones repos, installs packages and edits files loses all of it on Replicate when the instance scales down, so each run sets itself up again, and that setup time is billed on private models.
Agents, desktop and keys built in. Replicate has no agent harness or desktop to start from, so you build both, and its secrets are passed into the model as inputs, where a prompt-injected agent running there could read them.

Questions people ask

Is Replicate a sandbox for AI agents?

Not in the usual sense. Replicate runs AI models as containers behind a prediction API. It has no SSH, no desktop and no agent harness, while a boat sandbox is a full VM an agent works inside.

Is boat cheaper than Replicate?

Yes, boat costs less in all five scenarios on this page, such as $26 against $237 a month for one agent that never stops. For 1,000 short jobs at once, boat's $2,000 plan is still cheaper than Replicate's $3,580 but queues the starts, while Replicate fits the burst.

Can boat run GPUs like Replicate?

No. boat is CPU only. If your agent needs to run or fine-tune a model on a GPU, Replicate offers T4 to H200, or your agent in boat can call Replicate's API.

How long can a Replicate prediction run?

Predictions time out after 30 minutes, and you can ask Replicate for longer. A boat sandbox has no session cap on paid plans, so an agent can work for hours or days.

Can I run Docker inside Replicate?

Cog builds your model into a Docker image, but running Docker inside that container is not documented. A boat sandbox runs Docker, systemd and /dev/kvm like a normal Ubuntu server.

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