boat vs Celesto Cloud
Both give an agent its own small Linux computer in a microVM. boat gives it a fuller machine with Docker, nested virtualization and agents built in, and lets you run a whole fleet, while Celesto's paid offer stops at 10 running sandboxes and its core is open source.
Where boat stands out
A whole machine, not just a shell
Every boat sandbox is a full VM with root, systemd, Docker and /dev/kvm, up to 16 vCPU / 32 GB, on dedicated AMD Ryzen 9 9950X hosts. Celesto's hosted computers run as a normal user, and its docs do not cover Docker or nested virtualization there.
Room for a real fleet
boat's $20 plan runs 100 sandboxes at once and larger plans go up to 2,000, multiplied by seats. Celesto's paid offer allows up to 10 running sandboxes, subject to host capacity.
Agents, keys and hosting built in
Claude Code, Codex and five other agents come preinstalled, and you prompt, steer and follow them through one API. Real API keys stay outside the sandbox in a credentials proxy, and each sandbox can serve 50 token-protected HTTPS ports and raw TCP.
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
Celesto Cloud
Celesto AI says it gives agents real computers: a hosted API that creates an isolated Linux computer for each agent, on top of an open-source microVM core.
Celesto Cloud's Computers API creates a Linux computer from a template (scratch or coding-agent), runs commands in it, publishes up to four ports as HTTPS URLs, and stops, starts or deletes it. The same open-source core, SmolVM, runs on your own machine with Firecracker on Linux or QEMU on macOS. Celesto also sells managed agents that run your own agent for your end users with per-user budgets.
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 | Celesto Cloud: 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 with Docker, databases and test suites at once; Celesto stops at 10 running sandboxes and does not document Docker in hosted computers. | 3rd cheapest of 367 $634 a month | 23rd of 367, with limits $1,408 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB Both let a computer run as long as you like, but boat adds preinstalled coding agents you steer by API and keeps real API keys outside the sandbox. | 5th cheapest of 366 $657 a month | 33rd of 366, with limits $1,095 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Short code runs on a small machine are where Celesto does well: it has a tiny 1 vCPU size and no monthly fee. | 186th of 345, with limits $2,000 a month | 58th of 345, with limits $50 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Giving each of your users a sandbox needs more than 10 running at once, and boat's HTTPS URLs are token-protected while Celesto's published URLs are public. | Cheapest of 361 $209 a month | 50th of 361, with limits $349 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 | 29th of 360, with limits $1,500 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 | 112th of 364, with limits $400 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month CI that runs Docker builds or Android emulators needs Docker and /dev/kvm, which boat documents and Celesto's hosted docs do not. | 4th cheapest of 377 $86 a month | 147th of 377, with limits $192 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB Both keep the disk across stop and start, so either works as a dev box; boat adds raw TCP and up to 50 HTTPS ports. | 18th cheapest of 368 $26 a month | 47th cheapest of 368 $58 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month boat streams a 60fps Linux desktop with a computer-use agent built in, while Celesto's hosted SDK offers a VNC display connection you drive yourself. | 6th cheapest of 369 $200 a month | 46th of 369, with limits $400 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 | 57th of 367, with limits $150 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 | 86th of 346, with limits $75 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 | 57th of 338, with limits $1,500 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month Celesto can create a computer with the internet turned off, which suits isolated attack targets, though only 10 run at once. | 2nd cheapest of 360 $200 a month | 15th of 360, with limits $75 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 | Can't run it | Can't run it |
Key features
| Feature | boat | Celesto Cloud |
|---|---|---|
| Isolation | Full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts | MicroVM with its own kernel (Firecracker in the open-source core) |
| Biggest machine | Up to 16 vCPU / 32 GB on the $100 plan and above | Up to 8 vCPU / 32 GB (XL) |
| Root and Docker | Root, systemd and Docker inside every sandbox | Runs as user ohm; Docker inside hosted computers is not documented |
| Nested virtualization | Yes, /dev/kvm for Android emulators and VMs | Not documented for hosted computers |
| Keeping files | Disk saved every minute and on stop; a stopped sandbox costs nothing | Root disk and workspace kept across stop and start; stopped root disks kept free for 30 days; home folder kept only if set at create |
| Fork | Fork a running sandbox with its disk | Homepage lists branching a running sandbox; the hosted Computers API and SDK show no fork call |
| Sandboxes running at once | 100 on the $20 plan, up to 2,000 on larger plans, multiplied by seats | Up to 10 on the paid offer, subject to host capacity |
| Inbound access | HTTPS URLs for up to 50 ports, token-protected by default; raw TCP | Up to four ports at a time as public HTTPS URLs with no built-in auth |
| Outbound rules | Open internet, no provider-side allowlist | Internet on or off, set at create; no domain or IP allowlists |
| Desktop and browser | 60fps Linux desktop with Chromium and a built-in computer-use agent | SDK opens a VNC display connection and a Chrome DevTools browser connection |
| Coding agents | Seven agents preinstalled, with a prompt, steer and events API | coding-agent template with Node, npm and git; a Pi extension runs Pi against a remote computer |
| Keeping API keys out of the sandbox | Credentials proxy for model providers and any HTTPS API; switch a credential to another account when credits run out | No credentials proxy for hosted computers; the Pi extension keeps model keys on your own machine |
| Regions | EU only (Germany, Finland, France), with EU data residency | No region choice documented; infrastructure subprocessors in the United States |
| Self-hosting | No | Open-source Apache-2.0 core runs on your own Linux or macOS machine |
| Compliance and transparency | SOC 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)
| Feature | boat | Celesto Cloud |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | Not published |
| 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 | No |
| Wake on request | No | No |
| 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 | No |
| GPU | No | No |
| 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 | Not published |
| Usage per sandbox (API) | Yes | No |
| Audit logs | No | Not published |
| SSO | No | No |
| SOC 2 | No | No |
| HIPAA | No | No |
| EU data residency | Yes | No |
| Zero data retention | Yes | Not published |
Performance
There is no comparable measurement: Celesto Cloud has no score in the StarSling hpc-sandbox-benchmarks run we use. boat did 26.69 runs per second on its Node.js build-and-test loop at 4 vCPU / 8 GB. We will add Celesto when it has a score from the same test.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Celesto Cloud 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 | Celesto Cloud, 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 | $14 · Usage-based paid |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $58 · Usage-based paid |
| 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 | $704 · Usage-based paid · 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 | $150 · Usage-based paid · 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 | $300 · Usage-based paid · with limits |
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.
Celesto Cloud: Celesto charges an hourly rate for each named size while a computer runs, with no monthly platform fee. Stopped root disks are kept free for 30 days, and the egress price is not published. Celesto recently halved its rates; sandboxes created before the cut keep the old rates.
Limits in those scenarios
The limit that bites on Celesto is the 10 running sandboxes on its paid offer, which is why it does not fit 50 agents in parallel, sandboxes for your users, or 1,000 jobs at once. Docker inside a hosted computer is not documented either, so a fleet that needs it is unconfirmed. boat runs those fleets, but its start limits are lower than a burst needs: at 1,000 short jobs at once, even the $2,000 plan queues the starts.
| Scenario | boat | Celesto Cloud |
|---|---|---|
| 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 Celesto Cloud when
you want an open-source sandbox you can also run on your own machines, or a handful of small sandboxes with no monthly fee.
- Open source you can run yourself. Teams that want to run sandboxes on their own hardware can use the Apache-2.0 core with Firecracker or QEMU.
- Agents for your end users, with budgets. If you sell an agent to your own users, Celesto can run it per user with spend caps and an audit trail.
- Offline sandboxes and a big workspace disk. A sandbox can be created with the internet off, and its workspace can grow far past the root disk.
Use boat when
your agents need a whole machine with Docker, emulators, a desktop and their own coding agent, and you want to run dozens or thousands of them at once.
- A whole machine, not just a shell. Agents that run docker compose, build images or boot an Android emulator need these on day one. Without them the task fails, or you rebuild the work to fit a plain user shell.
- Room for a real fleet. Fifty agents in parallel, or one sandbox per user of your product, does not fit in 10 slots. On Celesto the rest wait in your own queue, so the work takes five times as long.
- Agents, keys and hosting built in. On Celesto you install and drive the agent yourself, and its keys sit inside the machine it controls unless you run it from your own laptop as the Pi extension does. Only four ports can be public, with no login.
Questions people ask
Is boat cheaper than Celesto?
For an agent that never stops, yes: about $26 a month on boat against $58 on Celesto. For one coding agent on workdays, Celesto costs less, about $14 against boat's $20 plan. Fleets of 50 agents or sandboxes for your users do not fit on Celesto, which runs at most 10 sandboxes on its paid offer.
Can I run Docker in a Celesto computer?
Celesto's hosted docs do not say. Its computers run as the user ohm, and the coding-agent template lists Node, npm and git. boat gives every sandbox root and Docker, plus /dev/kvm for emulators and VMs.
How many sandboxes can I run at once?
Celesto's paid offer allows up to 10 running sandboxes, subject to host capacity, and the trial allows one. boat allows 100 on the $20 plan and up to 2,000 on larger plans, multiplied by seats.
Can I self-host Celesto?
Yes. Its core is open source under Apache-2.0 and runs on your own Linux machine with Firecracker or your Mac with QEMU. boat is only offered as a hosted service.
Where do Celesto sandboxes run?
Celesto does not let you pick a region, and its subprocessor list names AWS and Google infrastructure in the United States. boat runs only in the EU, in Germany, Finland and France.
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