boat vs Huawei Cloud AgentArts
Huawei Cloud AgentArts is an agent platform: it hosts your agent image, starts a microVM per session and throws it away when the session ends. boat gives each agent a full VM it keeps, with root, Docker, a desktop and coding agents built in.
Where boat stands out
A machine that stays
A boat sandbox keeps its whole disk, saves a snapshot every minute and on stop, and costs nothing while stopped. An AgentArts sandbox is destroyed when its session stops or times out, and only folders you mount from session storage, SFS Turbo or OBS come back.
Root, SSH, Docker and /dev/kvm
Every boat sandbox is a full VM where your agent is root, with systemd, Docker and nested virtualization ready, and you can SSH in. AgentArts reaches its sandbox through API calls for commands and files, and documents no Docker inside it and no /dev/kvm.
Coding agents you drive from outside
boat ships Claude Code, Codex, OpenCode and other agents inside each sandbox, and you drive them from outside with a prompt, steer and events API. AgentArts offers its own Managed Agents loop instead. boat's credentials proxy keeps keys for any HTTPS API out of the sandbox and 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
Huawei Cloud AgentArts
AgentArts is Huawei Cloud's platform for building, hosting, running and evaluating AI agents, sold on its mainland China site.
You package your agent as a Docker image and AgentArts runs it as a runtime: a gateway routes each session ID to its own microVM and starts a new one when none is live. Around that it adds a code interpreter tool, a cloud browser tool, Managed Agents (Huawei's own hosted agent loop driven over WebSocket), memory, model routing and evaluation. It is built for companies that want a managed agent platform inside Huawei Cloud, with low-code and visual builders next to the SDK.
- Kind of product
- A full agent platform (runtime, tools, memory, gateway, evaluation), not only a sandbox
- Session lifetime
- Idle timeout and maximum lifetime from 1 minute to 7 days (defaults 15 minutes and 24 hours); stopping a session destroys the sandbox
- CPU architecture
- arm64 (the default) or x86_64 runtimes
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 | Huawei Cloud AgentArts: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month boat fits: agents keep a whole disk with Docker and root, while an AgentArts session loses everything outside its mounted storage when it ends. | 3rd cheapest of 367 $634 a month | 128th of 367, with limits $1,143 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB AgentArts ends a session after at most 7 days, so an assistant that runs for weeks needs restarts there; a paid boat sandbox has no session cap. | 5th cheapest of 366 $657 a month | 66th of 366, with limits $1,455 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month AgentArts fits short code runs better: a runtime wakes a fresh microVM per session and bills by the second, while boat's start limits make big bursts a compromise. | 186th of 345, with limits $2,000 a month | 72nd of 345, with limits $91 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month AgentArts does not publish how many sessions can run at once, so 300 user sandboxes is unconfirmed there; boat's $100 plan allows 300 per seat. | Cheapest of 361 $209 a month | 211th of 361, with limits $508 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 | 45th of 360, with limits $2,691 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 | 115th of 364, with limits $438 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month Both fit; on boat a CI job also gets Docker and /dev/kvm, which AgentArts does not document inside its sandboxes. | 4th cheapest of 377 $86 a month | 22nd cheapest of 377 $282 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A dev box needs SSH and a disk that stays; AgentArts reaches a sandbox only through its API and resets it when the session ends. | 18th cheapest of 368 $26 a month | 235th of 368, with limits $63 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month AgentArts has a cloud browser with a watch-only noVNC view and human takeover, but no full desktop; boat streams a 60fps desktop. | 6th cheapest of 369 $200 a month | 146th of 369, with limits $365 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month Both fit browser agents; AgentArts adds saved browser profiles and per-session domain allow and block lists. | 7th cheapest of 367 $200 a month | 60th of 367, with limits $200 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 | 13th cheapest of 346 $189 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 | 117th of 338, with limits $3,203 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 | 52nd of 360, with limits $121 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 | Huawei Cloud AgentArts |
|---|---|---|
| Isolation | Every sandbox, browser included, is a full VM with its own kernel | A microVM per runtime session; the browser tool runs in a sandbox container |
| Getting inside | Root shell over SSH, plus an API for commands and files | API only: run commands, upload and download files, invoke the agent; no SSH |
| Docker and nested virtualization | Docker, root, systemd and /dev/kvm on every machine | Your agent ships as a Docker image; Docker inside the sandbox and /dev/kvm are not documented |
| Files after a stop | The whole disk stays, with snapshots every minute and on stop; a stopped sandbox costs nothing | Stopping a session destroys the sandbox; only mounted session storage, SFS Turbo or OBS folders survive |
| Session length | No session cap on paid plans | Idle timeout and maximum lifetime of 1 minute to 7 days (default 24 hours) |
| Snapshots and fork | Named snapshots and fork of a running sandbox | No snapshot, pause or fork operation in the API |
| Desktop and browser | 60fps desktop and Chromium, computer use and browser control, embeddable stream | Cloud Chrome with a watch-only noVNC view, human takeover, saved browser profiles and proxies; no full desktop |
| Waking on a request | You start a sandbox through the API before sending it work | The gateway starts a fresh microVM for any session ID with no live sandbox |
| Coding agents | Claude Code, Codex, OpenCode and more preinstalled, with a prompt, steer and events API | Managed Agents: Huawei's own agent loop with bash and file tools, driven over WebSocket, which you can redirect or stop |
| 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 | Model providers: the gateway stores the key and adds it to inference requests |
| Reaching the sandbox | HTTPS URLs for up to 50 ports, token-protected; raw TCP | The runtime's HTTP or MCP endpoints through a gateway with IAM, API key or OAuth sign-in; no raw ports |
| Outbound network | Open internet on every plan; no provider-side allowlist | Open internet or private VPC only, with security groups; browser sessions take domain allow and block lists |
| CPU architecture | x86_64 only | arm64 or x86_64 |
| Regions | EU only: Germany, Finland, France | Huawei Cloud's mainland China site |
| 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 | Huawei Cloud AgentArts |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | No |
| Nested virtualization | Yes | No |
| Root | Yes | Not published |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | No |
| Browser | Yes | Yes |
| Computer-use API | Yes | Yes |
| SSH | Yes | No |
| Persistent disk | Yes | Yes |
| 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 | Yes |
| Secret proxy | Yes | Yes |
| Secret proxy for any API | Yes | No |
| Preinstalled agents | Yes | No |
| Agent harness API | Yes | Yes |
| MCP server | No | No |
| GPU | No | No |
| Windows | No | No |
| Arm64 | No | Yes |
| Self-hosting | No | No |
| 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 | Not published |
| SSO | No | Yes |
| SOC 2 | No | No |
| HIPAA | No | No |
| EU data residency | Yes | No |
| Zero data retention | Yes | Not published |
Performance
There is no comparable measurement: AgentArts was not part of StarSling's public sandbox benchmark. In that run of 2026-09-29, boat did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB, 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. Huawei Cloud AgentArts 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 | Huawei Cloud AgentArts, 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 | $17 · Pay as you go · with limits |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $66 · Pay as you go · 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 | $740 · Pay as you go · 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 | $241 · 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 | $463 · Pay as you go · 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.
Huawei Cloud AgentArts: AgentArts bills compute units in CNY: 1 CU per vCPU-hour and 0.1 CU per GiB-hour of memory while a sandbox runs, with Team and Enterprise packages of monthly CU bought by ticket. Its public sizing for a 4 vCPU / 8 GB sandbox is unconfirmed, so the scenario table prices Huawei on pay-per-use ECS T6 burstable servers in Singapore, billed by the second, with disk and public IP extra.
Limits in those scenarios
On Huawei the CPU in the priced servers is burstable: 20% of each vCPU is guaranteed and work above that is throttled once credits run out. A session ends after 168 hours, so an agent that never stops needs restarts, and files outside mounted session storage reset each time. Huawei does not publish how many sandboxes can run at once, so 300 user sandboxes and 1,000 short jobs are unconfirmed. On boat the 1,000-job burst does not fit either: the $2,000 plan allows 2,400 starts a day where the burst needs about 6,667, and starts above 90 a minute are queued.
| Scenario | boat | Huawei Cloud AgentArts |
|---|---|---|
| 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 Huawei Cloud AgentArts when
your agents must run in mainland China or inside a Huawei Cloud VPC, or when you want a sandbox woken per chat session on a managed agent platform.
- A sandbox per request, started for you. Useful for chat-style agents: the gateway starts a fresh microVM for each new session ID and bills it by the second.
- Network rules and Huawei Cloud wiring. Useful for teams already on Huawei Cloud: sandboxes can run inside your VPC with security groups and mount SFS Turbo or OBS.
- China region and Arm. Useful if your users and data must stay in mainland China, or your agent image is built for Arm.
Use boat when
your agents do real work for hours or weeks: installs, builds, tests, Docker and emulators on a full VM that keeps its disk, with a desktop and coding agents built in.
- A machine that stays. An agent that installs packages, builds caches and edits a repo loses all of it on AgentArts when the session ends, unless every path lives in a mounted folder. That means reinstalling on each new session.
- Root, SSH, Docker and /dev/kvm. Agents that run docker compose, test containers or start an Android emulator need a container engine and /dev/kvm. Without them on AgentArts, that work has to move to separate ECS servers you manage yourself.
- Coding agents you drive from outside. If your team already uses Claude Code or Codex, AgentArts means rebuilding the agent on Huawei's loop or packaging the CLI into your own image. Its key proxy covers model providers only, so other API keys still sit in the sandbox.
Questions people ask
Is boat cheaper than Huawei Cloud AgentArts?
In most of the scenarios we priced, yes: one agent running all month comes to $26 a month on boat against $66 on Huawei, and 50 agents in parallel come to $317 against $740. For one agent on workdays Huawei's burstable server is a little cheaper but loses installed packages between sessions. For 1,000 short jobs at once Huawei comes to $463 against $2,000, though neither fits without higher limits.
Are AgentArts sandboxes VMs or containers?
Runtime sessions each run in their own microVM, and the browser tool runs each session in a sandbox container. Either way the sandbox lives only as long as its session. On boat every sandbox is a full VM that keeps its disk until you delete it.
Can I SSH into an AgentArts sandbox?
No SSH is documented. You run shell commands and move files through the runtime API, or let Managed Agents use its bash and file tools. On boat you get a root shell over SSH as well as an API.
Can I run Claude Code or Codex on AgentArts?
Not out of the box. AgentArts offers its own Managed Agents loop, and a memory plugin for Claude Code, Codex and OpenCode, but you would have to package a coding CLI into your own runtime image. boat ships these agents preinstalled with an API to drive them.
Where does my data run?
AgentArts is sold on Huawei Cloud's mainland China site. boat runs only in the EU (Germany, Finland and France), with EU data residency and a zero data retention option.
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. Huawei Cloud lists ISO 27001, ISO 27017 and CSA STAR certifications on its trust center.
Start a sandbox
Install the CLI, then run boat new.
7-day free trial