boat vs Cua Fleet
Both give a computer-use agent a cloud desktop to work in. boat gives each agent a full VM that keeps its files, can be forked and comes with coding agents built in, while Cua is strongest as open-source tooling for driving desktops on Linux, Windows and your own machines.
Where boat stands out
A disk that stays, and forks of a running machine
A boat sandbox saves its disk every minute and when it stops, and a stopped sandbox costs nothing. You can start it again days later with every file in place, or fork a running sandbox into copies that share its state.
A whole VM, not a container
Every boat sandbox is a full VM with its own kernel, root, systemd and Docker, plus /dev/kvm for Android emulators and VMs. You can host what the agent builds on HTTPS for up to 50 ports, or reach it on raw TCP and its own IP.
Coding agents and keys built in
Claude Code, Codex and five other agents come preinstalled, and you drive agents inside from outside with one prompt, steer and events API. Real API keys stay outside the sandbox in a credentials proxy, and a credential 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.
- Isolation
- Full VM per sandbox, own kernel
- Regions
- EU: Germany, Finland, France
Cua Fleet
Cua (Cua AI, Inc., San Francisco) makes open-source tools that let AI agents use computers, plus Cua Fleets, a cloud that runs sandboxes for them.
Cua Fleets gives you pools of cloud sandboxes that you claim from the Cua SDK or CLI and release when done: gVisor containers or KubeVirt VMs, Linux or Windows, amd64 only. Around it sit the MIT-licensed Cua Driver, which clicks, types and reads accessibility trees on macOS, Windows and Linux, Lume for local macOS VMs, and Cua Spaces, a desktop app that runs agent desktops on machines you own. It is built for teams training, evaluating and running computer-use agents.
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 | Cua Fleet: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month A software factory needs files that survive between runs, Docker and preinstalled coding agents, which boat has, while a released Cua sandbox may be recycled. | 3rd cheapest of 367 $634 a month | Can't run it |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An assistant that lives on its machine for weeks needs a disk that is still there tomorrow; Cua has no suspend in the cloud, and boat keeps the disk through every stop. | 5th cheapest of 366 $657 a month | Can't run it |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Short runs that finish in minutes suit Cua's claim-and-release pools, where a warm sandbox is ready and releasing it ends the charge. | 186th of 345, with limits $2,000 a month | Can't run it |
| 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 they come back to works better when the disk persists and you can host their app on HTTPS ports, as on boat. | Cheapest of 361 $209 a month | Can't run it |
| 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 | Can't run it |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month | 10th cheapest of 364 $200 a month | Can't run it |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month CI jobs that build Android apps or VM images need /dev/kvm, which boat exposes and Cua's Fleets docs do not mention. | 4th cheapest of 377 $86 a month | Can't run it |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A dev box that runs all month needs a disk that survives restarts and a way to reach it on any port, which is what boat is built for. | 18th cheapest of 368 $26 a month | Can't run it |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Both fit: Cua brings open-source drivers and Windows desktops, and boat brings a 60fps Linux desktop with Chromium on a full VM that keeps its state. | 6th cheapest of 369 $200 a month | Can't run it |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month Both run a real browser for the agent; boat adds a disk that keeps logins and profiles between sessions. | 7th cheapest of 367 $200 a month | Can't run it |
| 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 | Can't run it |
| 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 | Can't run it |
| 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 | Can't run it |
| 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 | Cua Fleet |
|---|---|---|
| Isolation | Full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts | gVisor container or KubeVirt VM, amd64 only |
| Operating systems | Linux (Ubuntu) | Linux and Windows in the cloud; macOS only locally with Lume |
| Disk after a stop | Kept; saved every minute and on stop, and a stopped sandbox costs nothing | No suspend in the cloud; a released sandbox may be recycled, so you copy results out first |
| Pause and resume | Stop and start again with the same disk | Suspend, resume and restart are local only |
| Fork | Fork a running sandbox with its disk | Snapshots not implemented; snapshot-derived images are rejected |
| Session length | No cap on paid plans; you set the time to live | No fixed cap; a 15-minute claim TTL is renewed while your process holds it, and idle managed pools are deleted after 30 minutes |
| Nested virtualization | Yes, /dev/kvm for Android emulators and VMs | Not documented |
| Desktop and computer use | Linux desktop streamed at 60fps, Chromium, and a computer-use API | Linux and Windows desktops, driven by the open-source Cua Driver over MCP or CLI |
| Inbound access | HTTPS URLs for up to 50 ports, token-protected by default; raw TCP; dedicated IPv4 or IPv6 | Named services through an authenticated gateway, plus signed URLs |
| Coding agents | Seven agents preinstalled, with a prompt, steer and events API | None preinstalled; you bring your own agent loop |
| 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 | Secrets delivered into the running sandbox as a file |
| Run on your own machines | No | Yes: your Mac, your own boxes, or your AWS, GCP or Modal account |
| Open source | No | Driver, Lume and SDK under MIT; Cua Spaces source-available |
| Regions | EU only (Germany, Finland, France) | US (the SDK defaults to us-east-1); your own cloud's regions when you run it there |
| 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 | Cua Fleet |
|---|---|---|
| Full VM (own kernel) | Yes | No |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | Not published |
| Root | Yes | Yes |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | Yes |
| Browser | Yes | Yes |
| Computer-use API | Yes | Yes |
| SSH | Yes | Yes |
| 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 | 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 | Yes |
| GPU | No | No |
| Windows | No | Yes |
| Arm64 | No | No |
| Self-hosting | No | Yes |
| BYOC | No | Yes |
| 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 | Yes |
| HIPAA | No | No |
| EU data residency | Yes | No |
| Zero data retention | Yes | Not published |
Performance
There is no comparable measurement for Cua Fleets: it was not part of the hpc-sandbox-benchmarks run. boat scored 26.69 runs per second on that Node.js build-and-test loop at 4 vCPU / 8 GB, on dedicated AMD Ryzen 9 9950X hosts.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Cua Fleet 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 | Cua Fleet, 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 | Can't run it |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | Can't run it |
| 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 | Can't run it |
| 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 | Can't run it |
| 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 | Can't run it |
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.
Cua Fleet: Cua no longer publishes a Fleets rate, so the table has no Cua price. Its pricing page shows only Cua Spaces plans, priced per seat and relay hours, while the Fleets docs say usage is billed while a sandbox runs and that warm replicas bill whether or not anything claims them.
Limits in those scenarios
The table cannot place Cua in any of the five scenarios because Cua no longer publishes a Fleets rate. Past price, nothing on a Cua disk survives: there is no suspend in the cloud and a released sandbox may be recycled. Cua does not publish how many sandboxes an account can run at once, and a managed pool runs at most 10 per image unless you raise it, so the 300 and 1,000 sandbox scenarios are unconfirmed. boat's limit that bites is starts: for 1,000 short jobs at once, the $2,000 plan allows 600 starts an hour and queues bursts above 90 a minute.
| Scenario | boat | Cua Fleet |
|---|---|---|
| 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 Cua Fleet when
your agents must work in Windows desktops, or you want open-source computer-use tooling that runs on your own machines or cloud account.
- Windows desktops and drivers for every OS. Useful when your agent must use Windows apps, or native apps on macOS and Linux through one open-source driver.
- Runs on machines you already have. Teams that want agents on their own Macs, boxes or AWS, GCP and Modal accounts can run Cua there and pay that cloud directly.
- Open-source computer-use tooling. Its driver, Lume and SDK are MIT-licensed in one public repo, so you can read, change and self-host them.
Use boat when
your agents do real work that has to last: coding, builds, Docker, emulators and a browser, on a full VM that keeps its disk, forks, and comes with agents and a keys proxy built in.
- A disk that stays, and forks of a running machine. On Cua Fleets there is no suspend in the cloud and a released sandbox may be recycled, so an agent's work is lost unless your code copies it out first, and every new session starts from the image again.
- A whole VM, not a container. Cua's default Linux sandbox is a gVisor container, which runs on a user-space kernel; its VM images must be a prebuilt bootable disk, and its docs say nothing about nested virtualization.
- Coding agents and keys built in. Cua gives you drivers, not an agent: you build and host the agent loop yourself, and secrets land inside the sandbox as a file the agent can read, where one prompt injection can leak them.
Questions people ask
Is boat cheaper than Cua Fleet?
There is no current Cua price to compare: Cua no longer publishes a Fleets rate, and its pricing page lists only Cua Spaces plans. The table shows what boat costs in each scenario, from $20 a month for a workday agent, so ask Cua for a quote and compare it with those numbers.
Does a Cua sandbox keep its files after it stops?
Cua's docs say suspend and resume are local only, and a released cloud sandbox may be recycled, so you copy results out first. A boat sandbox saves its disk every minute and on stop, and starts again with every file in place.
Can I run Windows or macOS agents?
Cua Fleets runs Windows images in the cloud, and Lume runs macOS VMs locally on Apple silicon; Cua's docs say there is no macOS in its cloud. boat runs Linux only.
Can I run Docker and Android emulators in each?
boat gives every sandbox a full VM with Docker and /dev/kvm, so Docker, Android emulators and nested VMs all run. Cua's default Linux sandbox is a gVisor container, and its Fleets docs do not mention nested virtualization.
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