boat vs Replicas
Replicas is a cloud coding agent product: you hand it a task from Slack, Linear or GitHub and it works in a Linux VM. boat is the machine layer under that kind of product, with its own kernel, /dev/kvm, forks of running sandboxes and coding agents you drive from outside.
Where boat stands out
A whole machine you control
A boat sandbox is a VM with its own kernel, root, systemd, Docker and /dev/kvm, and you can reach it on raw TCP as well as HTTPS. A Replicas workspace is also a VM, but it is shaped around one coding agent task, with Docker started on demand and no /dev/kvm or systemd in its docs.
Files that stay, and forks of running work
boat snapshots the disk every minute and on stop, and a stopped sandbox keeps its files at no charge. You can also fork a running sandbox into a new one to try two ideas from the same point. Replicas archives a sleeping workspace after 7 days idle and deletes it 30 days later, and its fork moves a chat, not a machine.
Real keys never enter the sandbox
boat's credentials proxy holds your real model and API keys outside the sandbox and works for any HTTPS API you register. It can also switch a credential to another account of the same type when one runs out of credits. Replicas keeps plugin keys in a gateway, but agent credentials and environment variables are placed inside the workspace.
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
- Regions
- EU: Germany, Finland, France
Replicas
Replicas is a Y Combinator startup (Spring 2026) that runs coding agents like Claude Code, Codex, Cursor and Opencode in cloud machines.
Replicas is a finished coding agent product more than a sandbox API. Each task gets a workspace, an isolated Linux VM with your repositories, a desktop with Chrome and the agent you chose, and you start work from Slack, Linear, GitHub, a web app, a mobile app, a CLI or a REST API. Its own data processing terms name E2B as the provider of the isolated VMs, so the machines underneath are E2B sandboxes.
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 | Replicas: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Both fit; Replicas gives each coding task a ready workspace with Slack, Linear and GitHub triggers, while boat adds /dev/kvm, systemd and forks of a running sandbox. | 3rd cheapest of 367 $634 a month | 19th cheapest of 367 $5,790 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB Both fit; Replicas workspaces sleep after 30 minutes idle and wake on a message, while a boat sandbox keeps running until you stop it. | 5th cheapest of 366 $657 a month | 29th cheapest of 366 $7,534 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Replicas fits short runs better here, since boat's limits on starts per minute turn big bursts into a compromise. | 186th of 345, with limits $2,000 a month | 47th cheapest of 345 $800 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Replicas is built for your own team's coding tasks and caps an organization at 100 API replicas at once, so hosting sandboxes for your users is a compromise there. | Cheapest of 361 $209 a month | 164th of 361, with limits $2,520 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 | 158th of 360, with limits $12,000 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 | 187th of 364, with limits $1,920 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month Replicas has no CI runner product and caps API replicas at 100 at once, while boat sandboxes run Docker and /dev/kvm for builds and emulators. | 4th cheapest of 377 $86 a month | 108th cheapest of 377 $1,152 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB | 18th cheapest of 368 $26 a month | 108th cheapest of 368 $301 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Both fit: every Replicas workspace has a Linux desktop with Chrome, and boat streams a 60fps desktop you can embed in your app. | 6th cheapest of 369 $200 a month | 36th cheapest of 369 $1,760 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month Replicas has Chrome on a desktop for its coding agent; boat adds browser control you drive from your own code. | 7th cheapest of 367 $200 a month | 109th of 367, with limits $960 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 | 74th cheapest of 346 $1,170 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 | 293rd of 338, with limits $23,400 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 | 44th of 360, with limits $580 a month |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month | Can't run it | Can't run it |
| macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month Neither runs macOS machines, though Replicas can hand a workspace a cloud iOS simulator through its partner Limrun. | Can't run it | Can't run it |
Key features
| Feature | boat | Replicas |
|---|---|---|
| What you get | A sandbox VM you control by API, CLI or SSH, with coding agents built in | A coding agent product: tasks start from Slack, Linear, GitHub, web, mobile, CLI or API |
| Isolation | Every sandbox is a full VM with its own kernel | Every workspace is an isolated Linux VM, run by E2B |
| Docker and nested virtualization | Docker, root, systemd and /dev/kvm on every machine | Docker installed, started on demand with sudo; no /dev/kvm or systemd documented |
| Desktop and browser | 60fps desktop and Chromium, computer use and browser control, embeddable stream | Linux desktop with Chrome in every workspace, viewed in the Replicas app |
| Idle and wake | Runs until you stop it; a stopped sandbox costs nothing and keeps its disk | Sleeps after 30 minutes idle with no charge, and wakes on the next message |
| Keeping files | Snapshots every minute and on stop; stopped sandboxes are kept | Sleeping workspaces are archived after 7 days idle and deleted after 30 more days |
| Fork | Fork a running sandbox into a new, independent one | Fork a chat to another agent harness; no machine fork |
| Coding agents | Claude Code, Codex, OpenCode, pi and more, with a prompt, steer and events API | Claude Code, Codex, Cursor, Opencode, Pi and more, with messages, queue steering and an events stream |
| Work triggers | Your own code calls the API | Slack, Linear, GitHub and GitLab triggers, automations, a mobile app and an MCP server |
| Keeping keys out of the sandbox | AI credentials proxy for model providers and any HTTPS API you register; switch a credential to another account when credits run out | A gateway keeps keys for Slack, Linear, Sentry and catalog plugins outside; agent credentials and variables are put in the workspace |
| Reaching the sandbox | HTTPS URLs for up to 50 ports, token-protected; raw TCP; dedicated IPv6 or IPv4 | HTTPS preview URLs per port, optionally behind a Replicas login; SSH and a local tunnel |
| Mobile apps | Android emulators run inside the sandbox on /dev/kvm | Cloud iOS simulators, Android emulators and Xcode sandboxes through partner Limrun |
| Where it runs | EU only: Germany, Finland, France, with zero data retention | Primary hosting in Ohio, US; dedicated or self-hosted for Enterprise |
| Compliance and transparency | SOC 2 is in progress. We're open about where we stand: live status and data retention policy. | SOC 2 |
Every feature Battleships tracks (43)
| Feature | boat | Replicas |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | No |
| Root | Yes | Not published |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | Yes |
| Browser | Yes | Yes |
| Computer-use API | Yes | Yes |
| SSH | Yes | Yes |
| Persistent disk | Yes | Yes |
| Pause/resume | Yes | Yes |
| Automatic snapshots | Yes | Yes |
| 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 | Yes |
| Agent harness API | Yes | Yes |
| MCP server | No | Yes |
| GPU | No | No |
| Windows | No | No |
| Arm64 | No | No |
| Self-hosting | No | Yes |
| BYOC | No | No |
| Open source | No | No |
| Scoped API keys | Yes | Not published |
| Spend limits | Yes | Not published |
| Usage per sandbox (API) | Yes | No |
| Audit logs | No | Yes |
| SSO | No | Not published |
| SOC 2 | No | Yes |
| HIPAA | No | No |
| EU data residency | Yes | No |
| Zero data retention | Yes | Not published |
Performance
There is no comparable measurement for Replicas: it was not in the public benchmark we use. boat did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB in StarSling's run of 2026-09-29, on dedicated AMD Ryzen 9 9950X (Zen 5) hosts.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Replicas 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 | Replicas, 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 | $76 · Developer |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $315 · Developer |
| 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 | $3,802 · Developer |
| 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 | $1,080 · Developer · 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 | $2,400 · Developer · 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.
Replicas: Replicas bills per minute while a workspace is awake, with a one-minute minimum each time it wakes, and nothing while it sleeps. Developer is $50 a month and Team starts at $200, each including that much usage, and usage beyond it costs 1.5 times the included rate.
Limits in those scenarios
Replicas caps an organization at 100 API and automation replicas active or preparing at once, which is why 300 sandboxes for your users and 1,000 short jobs at once do not fit. Its biggest workspace is 8 vCPU / 32 GB, sleeping and archived workspaces cost nothing to keep, and it does not publish an egress price. On boat the limit is a sandbox count (100 to 2,000 by plan) plus starts per minute, so the 1,000-job burst is queued and needs more starts per day than the $2,000 plan allows.
| Scenario | boat | Replicas |
|---|---|---|
| One coding agent, workdays | None that bite |
|
| One agent that never stops | None that bite |
|
| 50 agents in parallel | None that bite |
|
| Sandboxes for your users | None that bite |
|
| 1,000 short jobs at once |
|
|
Which one should you use?
Use Replicas when
you want a ready coding agent your team starts from Slack, Linear or GitHub, that sleeps when idle, and you would rather not build that layer yourself.
- A finished coding agent product. Useful if you want a ready coding agent your team calls from Slack, Linear or GitHub, with no code to write.
- Sleeps when idle, wakes on a message. Useful if your agents work in short bursts: Replicas workspaces sleep after 30 minutes idle and wake on the next message.
- iOS simulators, US hosting, self-hosting. Useful if you test iOS apps, need US hosting, or must run the whole product inside your own infrastructure.
Use boat when
you are building the agent product yourself, or your agents need a full machine: /dev/kvm, systemd, raw TCP, forks of running work, keys kept outside and data in the EU.
- A whole machine you control. Android emulators, Firecracker or QEMU tests and services that expect systemd need a full machine. On Replicas the workaround for mobile is a separate partner, Limrun, rather than an emulator in the workspace.
- Files that stay, and forks of running work. An agent that comes back to a project after a month should find its machine where it left it. Without a machine fork, trying three fixes from the same state means rebuilding that state three times.
- Real keys never enter the sandbox. Code an agent runs can read anything inside its machine. When the agent's own keys sit in the workspace, a bad dependency or prompt injection can copy them out.
Questions people ask
Is Replicas a sandbox like boat?
Not quite. Replicas is a coding agent product: it takes a task, starts a Linux VM workspace with your repositories and runs Claude Code, Codex or another agent in it. boat is the sandbox layer itself, a full VM you control by API, with coding agents built in.
What runs under Replicas workspaces?
Replicas' data processing terms name E2B (FoundryLabs) as the subprocessor that runs its isolated virtual machines. Its docs describe each workspace as an isolated Linux VM with a desktop, Chrome and Docker.
Can I run Android emulators or Docker on Replicas?
Docker is installed and starts on demand with sudo. Replicas does not document /dev/kvm; for mobile it connects workspaces to cloud iOS simulators and Android emulators from Limrun. On boat, Android emulators run inside the sandbox on /dev/kvm.
How does Replicas bill?
Per minute while a workspace is awake, at a rate set by its size: small, large or ultra. The Developer plan costs $50 and Team starts at $200, billed monthly, each including that much usage, with overage at 1.5 times the included rate. Seats do not change the bill.
Where does my data run?
boat runs only in the EU (Germany, Finland and France), with EU data residency and a zero data retention option. Replicas' primary hosting is in Ohio, United States, and Enterprise customers can choose a dedicated or self-hosted deployment.
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. Replicas says it has SOC 2 Type I, with Type II in progress.
Start a sandbox
Install the CLI, then run boat new.
7-day free trial