boat vs AgentComputer
Both give each AI agent its own cloud computer with a desktop, a browser, SSH and Claude Code. boat adds a fuller machine you can fork while it runs and agents you drive from outside, while AgentComputer also offers macOS desktops and bigger memory sizes.
Where boat stands out
Fork a machine while it runs
boat can fork a running sandbox in one call, and it saves the disk every minute and on stop without you asking. On AgentComputer you queue a snapshot, wait for it, and restore it into a new computer, and nothing is saved unless you call it.
Agents you drive from outside
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 a credential can switch to another account when one runs out of credits.
A whole machine, not just a desktop
A boat sandbox gives you root with systemd, Docker, and /dev/kvm for Android emulators and VMs. It can host services on HTTPS for up to 50 ports, take raw TCP, and have its own IP, and its desktop streams at 60fps with a computer-use 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
- Regions
- EU: Germany, Finland, France
AgentComputer
AgentComputer sells cloud computers for AI agents: Firecracker microVMs on its own hosts, billed by the resources they use.
Each Linux computer is a Firecracker microVM that boots from AgentComputer's guest image, with an Xfce desktop over VNC, Chrome, SSH, Node, Python, Claude Code and Codex preinstalled. You create and drive computers from a CLI and a REST API that cover exec, files, published ports, snapshots and share links, and you can also start a macOS desktop that runs as a full QEMU/KVM VM. It is built for developers who want a machine an agent can live on, paid for by CPU time and memory with no plan fee.
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 | AgentComputer: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month A software factory needs Docker, emulators and copies of a working machine; boat forks a running sandbox in one call and exposes /dev/kvm, which AgentComputer's guest kernel does not. | 3rd cheapest of 367 $634 a month | 31st of 367, with limits $7,325 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB Both keep a machine running with no session cap; boat also lets you prompt and steer the agent inside through an API, while AgentComputer gives you exec and a shell. | 5th cheapest of 366 $657 a month | 27th cheapest of 366 $7,182 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month | 186th of 345, with limits $2,000 a month | 145th of 345, with limits $344 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month AgentComputer does not publish how many computers can run at once, so a few hundred user sandboxes is not confirmed there. | Cheapest of 361 $209 a month | 159th of 361, with limits $2,252 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 | 168th of 360, with limits $15,392 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 | 195th of 364, with limits $2,168 a month |
| 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 nested virtualization, which boat has and AgentComputer does not. | 4th cheapest of 377 $86 a month | 116th cheapest of 377 $1,309 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB Both make a good always-on dev box with SSH and a desktop; boat adds a dedicated IP, raw TCP and a disk snapshot every minute. | 18th cheapest of 368 $26 a month | 107th cheapest of 368 $292 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Both ship a Linux desktop with a browser; boat streams it at 60fps with a computer-use API, while AgentComputer gives VNC access that you connect your own agent to. | 6th cheapest of 369 $200 a month | 37th cheapest of 369 $1,865 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month Both have Chrome on a desktop; boat adds Chromium control and a credentials proxy so site and model keys stay out of the machine. | 7th cheapest of 367 $200 a month | 108th of 367, with limits $932 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 | 62nd cheapest of 346 $595 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 | 246th of 338, with limits $11,892 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 | 46th of 360, with limits $647 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 AgentComputer can start a macOS desktop VM, which boat does not offer. | Can't run it | 18th of 36, with limits $186 a month |
Key features
| Feature | boat | AgentComputer |
|---|---|---|
| Isolation | Full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts | Firecracker microVM with its own kernel |
| Operating systems | Linux | Linux, plus macOS desktops (no exec, ports or snapshots on macOS yet) |
| Biggest machine | Up to 16 vCPU / 32 GB on the $100 plan and above | Up to 64 GiB RAM and 250 GiB storage, subject to host capacity; vCPU count not published |
| Root and init | Root, with systemd as the init system | Passwordless sudo for the default user; a custom init, not systemd |
| Nested virtualization | Yes, /dev/kvm for Android emulators and VMs | No; the guest kernel has no KVM module |
| Snapshots | Disk saved every minute and on stop, automatically | Snapshots you create by CLI or API; restore goes into a new computer |
| Fork | Fork a running sandbox with its disk in one call | Take a snapshot, wait for it, then restore it into a new computer |
| Session length | No cap on paid plans; you set the time to live | No cap documented; a computer runs until you stop it |
| Desktop and computer use | Linux desktop streamed at 60fps, Chromium, and a computer-use API | Xfce desktop over VNC with Chrome; no first-party computer-use API |
| Inbound access | HTTPS URLs for up to 50 ports, token-protected by default; raw TCP; dedicated IPv4 or IPv6 | Published ports on a provider URL, public by default or private with an access session; shared NAT, no dedicated IP |
| Coding agents | Seven agents preinstalled, with a prompt, steer and events API | Claude Code and Codex preinstalled; you run them through exec or a shell, no agent API |
| 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 | claude-login and codex-login copy credentials onto the computer |
| Regions | EU only (Germany, Finland, France), with EU data residency | Not published; no region setting in the API |
| 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 | AgentComputer |
|---|---|---|
| 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 | No |
| SSH | Yes | Yes |
| Persistent disk | Yes | Yes |
| Pause/resume | Yes | Yes |
| Automatic snapshots | Yes | No |
| Fork/clone | Yes | Yes |
| 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 | Yes |
| Agent harness API | Yes | No |
| MCP server | No | No |
| GPU | No | No |
| Windows | No | No |
| Arm64 | No | No |
| Self-hosting | No | No |
| BYOC | No | No |
| Open source | No | No |
| Scoped API keys | Yes | Not published |
| Spend limits | Yes | Yes |
| Usage per sandbox (API) | Yes | Yes |
| Audit logs | No | Not published |
| SSO | No | Not published |
| SOC 2 | No | Not published |
| HIPAA | No | Not published |
| EU data residency | Yes | No |
| Zero data retention | Yes | Not published |
Performance
There is no comparable measurement. boat did 26.69 runs per second on the hpc-sandbox-benchmarks Node.js CPU test at 4 vCPU / 8 GB, and AgentComputer has not been measured on the same test. AgentComputer does not publish how many vCPUs each size gets, so a direct CPU comparison would need a test run.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. AgentComputer 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 | AgentComputer, 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 | $101 · Pay as you go |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $342 · Pay as you go |
| 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 | $4,318 · Pay as you go |
| 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 | $1,119 · Pay as you go · 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 | $3,078 · Pay as you go · 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.
AgentComputer: AgentComputer has no plan fee: it bills $0.07 per CPU-hour of CPU time, $0.04375 per GB-hour of memory, plus hot storage while running and cold storage while stopped. It does not say whether memory is billed on what you set or what you use, so the scenarios count what you set; egress prices are not published.
Limits in those scenarios
AgentComputer does not publish how many computers can run at once, so 300 user sandboxes or 1,000 jobs at once are not confirmed, and 50 agents is likely but not promised. Its egress price is not published either, so heavy downloads could add to the bill. boat's limits are published: the $2,000 plan starts at most 90 machines a minute and 600 an hour, so a burst of 1,000 short jobs is queued.
| Scenario | boat | AgentComputer |
|---|---|---|
| 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 AgentComputer when
your agent needs a macOS desktop, more than 32 GB of memory on one machine, or you only run a computer now and then and want no monthly plan.
- macOS desktops. Useful when an agent has to work in macOS apps: it can start a macOS desktop VM next to its Linux computers.
- Bigger memory and disk. A computer can have up to 64 GiB of RAM and 250 GiB of storage, set in 1 GiB steps.
- No plan fee. Pay as you go with no monthly fee, billed on CPU time, memory and storage, which suits light or occasional use.
Use boat when
your agents do real work on Linux for hours: Docker, emulators, builds and a browser, on a machine you can fork while it runs, with agents you drive by API and keys that never enter the sandbox.
- Fork a machine while it runs. An agent that wants to try three fixes from the same half-built state needs copies fast. With manual snapshots, a crash between snapshots loses the work, and each copy waits on a queued snapshot.
- Agents you drive from outside. AgentComputer also preinstalls Claude Code and Codex, but you drive them through a shell or exec, and claude-login copies your credentials onto a machine the agent controls, where a prompt injection can read them.
- A whole machine, not just a desktop. AgentComputer's guest kernel has no KVM module, so emulators and VMs cannot start, and with shared NAT and no dedicated IP, anything that is not HTTPS on a published port is hard to reach.
Questions people ask
Is boat cheaper than AgentComputer?
In the scenarios on this page, yes: one agent that never stops is $26 a month on boat against $342 on AgentComputer, and 50 agents in parallel is $317 against $4,318. Those AgentComputer numbers count memory as set, not as used; if it bills only used memory, the gap shrinks.
Does AgentComputer come with Claude Code?
Yes. Its default image has Claude Code and Codex preinstalled, and the CLI can copy your credentials onto the computer. boat also preinstalls them, plus five other agents, and adds an API to prompt and steer them while keeping real keys outside the sandbox.
Can I run Docker and Android emulators in each?
Docker can run in both, though AgentComputer does not preinstall it. Android emulators and VMs need /dev/kvm, which boat exposes and AgentComputer's guest kernel does not include.
Can AgentComputer run macOS?
Yes. It can start a macOS computer as a full QEMU/KVM VM with a VNC desktop, though exec, ports and snapshots are Linux-only for now. boat runs Linux only.
Can I clone a running machine?
On boat, you fork a running sandbox in one call. On AgentComputer, you queue a snapshot of the computer and then restore it into a new one.
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