boat vs KakaoCloud
KakaoCloud Virtual Machine is a general cloud server in Seoul that you set up yourself. boat is a VM made for AI agents, with coding agents, a 60fps desktop and per-port HTTPS URLs ready, and it costs less in most of the scenarios below.
Where boat stands out
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. A KakaoCloud VM boots a plain OS image, so you install each agent, run it as a process and build the control layer yourself.
A desktop and browser for computer use
Every boat sandbox can stream a 60fps desktop with Chromium, and agents can click, type and take screenshots through a computer-use API. You can embed the stream in your own app. KakaoCloud gives you SSH, or RDP on Windows, and no managed desktop.
A sandbox lifecycle built in
boat snapshots the disk every minute and on stop, a stopped sandbox costs nothing, and you can fork a running sandbox in one call. Each port can get an HTTPS URL with a token. On KakaoCloud you take snapshots, make images, attach public IPs and set up TLS yourself.
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
KakaoCloud
KakaoCloud is the public cloud of Kakao Enterprise, a South Korean company; its Virtual Machine service rents general cloud servers, not agent sandboxes.
You pick an instance type (compute, general, memory, burstable or GPU), boot it from an Ubuntu, Rocky, Alma, CentOS Stream or Windows Server image, and connect over SSH or RDP. Networking, firewalls, disks and images are separate building blocks you put together through the console or the OpenAPI. It is built for companies running general workloads in Korea, including public institutions.
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 | KakaoCloud: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Both run full VMs, but on KakaoCloud you install the agents, desktop and preview URLs yourself; boat ships them in every sandbox. | 3rd cheapest of 367 $634 a month | 200th of 367, with limits $2,326 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB A KakaoCloud VM can run all month, but it has no built-in agent API, credentials proxy or per-minute snapshots, so you build that layer yourself. | 5th cheapest of 366 $657 a month | 168th of 366, with limits $1,279 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month KakaoCloud costs less for short bursts, with per-minute billing on small burstable instances and a quota of 500 instances per project. | 186th of 345, with limits $2,000 a month | 61st of 345, with limits $56 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month boat gives each user a sandbox with an HTTPS URL and free stops; on KakaoCloud you wire up public IPs, firewalls and TLS for every user yourself. | Cheapest of 361 $209 a month | 55th of 361, with limits $427 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month Both are a compromise for large rollouts: KakaoCloud's cheapest small instances share CPU, and boat's start limits queue big bursts. | 34th of 360, with limits $2,000 a month | 54th of 360, with limits $3,080 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 | 30th cheapest of 364 $617 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month | 4th cheapest of 377 $86 a month | 33rd cheapest of 377 $312 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB | 18th cheapest of 368 $26 a month | 71st cheapest of 368 $94 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month boat streams a 60fps desktop with Chromium and computer use; a KakaoCloud VM has no managed desktop, so you set up VNC or RDP yourself. | 6th cheapest of 369 $200 a month | 194th of 369, with limits $634 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 | 199th of 367, with limits $187 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 | 193rd of 346, with limits $115 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 | 79th of 338, with limits $2,016 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month boat runs many isolated sandboxes with fork and snapshots built in; on KakaoCloud you clone VMs from images by hand. | 2nd cheapest of 360 $200 a month | 136th of 360, with limits $93 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 | KakaoCloud |
|---|---|---|
| Kind of product | Sandbox VMs made for AI agents, created by API in seconds | General cloud servers you configure yourself |
| Isolation | Every sandbox is a full VM with its own kernel | Every instance is a full VM with its own OS |
| Docker and nested virtualization | Docker, root, systemd and /dev/kvm on every machine | Root on the VM, so you can install Docker; nested virtualization is not documented |
| Coding agents | Claude Code, Codex, OpenCode and more preinstalled, with a prompt, steer and events API | None; you install and run agents yourself |
| Desktop and computer use | 60fps desktop and Chromium, computer use and browser control, embeddable stream | No managed desktop; you connect over SSH, or RDP on Windows |
| Snapshots and stops | Snapshots every minute and on stop; a stopped sandbox costs nothing | Volume snapshots on demand or on a schedule; resize only while stopped |
| Fork | Fork a running sandbox in one call | Make an image of the root volume (running or stopped), then boot new instances from it |
| Reaching the sandbox | HTTPS URLs for up to 50 ports, token-protected; raw TCP; dedicated IPv6 or IPv4 | Attach a public IPv4 and open ports in security groups; TLS is yours to set up |
| Outbound rules | Open internet on every plan; no provider-side allowlist | Security groups with outbound rules by protocol, port and IP range (no domain rules) |
| 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 | Secrets Manager stores secrets that your workloads read or sync; no proxy that adds them to requests |
| GPUs | None | NVIDIA T4 and A100 instances, plus FuriosaAI NPUs |
| Windows | No Windows guests; Wine for Windows apps | Windows Server 2019 and 2022 images, billed per instance-hour on top |
| Regions | EU only: Germany, Finland, France | Seoul (kr-central-2), plus separate regions for Korean public institutions |
| 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 | KakaoCloud |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | Not published |
| Root | Yes | Not published |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | No |
| Browser | Yes | No |
| 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 | Yes |
| 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 | Yes |
| Egress allowlist | No | Yes |
| 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 | Yes |
| Windows | No | Yes |
| Arm64 | No | No |
| 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 | 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: KakaoCloud was not part of StarSling's public sandbox benchmark. In that run, at 4 vCPU / 8 GB, boat did 26.69 Node.js build-and-test runs per second on dedicated AMD Ryzen 9 9950X (Zen 5) hosts. KakaoCloud's c2a compute instances give full vCPUs, while its t1i burstable instances only guarantee a share of each vCPU, from 5% to 40% by size.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. KakaoCloud 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 | KakaoCloud, 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 | $26 · Pay as you go |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $94 · 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 | $1,153 · 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 | $187 · Pay as you go |
| 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 | $339 · 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.
KakaoCloud: KakaoCloud bills each instance per minute at an hourly rate in KRW, before VAT, with no plan fee. Disks, public IPs, outbound traffic and Windows images are billed on top, and a public IP you keep is billed even when unused.
Limits in those scenarios
KakaoCloud has no session cap and no plan limits, so it fits every scenario except the 1,000 short jobs. Its low price for sandboxes for your users and for those jobs comes from t1i burstable instances, which guarantee only a share of each vCPU (20% on the t1i.medium size, 40% on the largest), so heavy, steady CPU work slows down. A project also starts with a quota of 500 instances, so 1,000 jobs at once need a Helpdesk request. On boat the 1,000-job burst hits start limits per minute, hour and day, so it gets queued.
| Scenario | boat | KakaoCloud |
|---|---|---|
| One coding agent, workdays | None that bite | None that bite |
| One agent that never stops | None that bite | None that bite |
| 50 agents in parallel | None that bite | None that bite |
| Sandboxes for your users | None that bite | None that bite |
| 1,000 short jobs at once |
|
|
Which one should you use?
Use KakaoCloud when
you need GPUs or Windows Server, or your data must stay in South Korea.
- GPUs and NPUs. Useful if your agents train or test models: KakaoCloud rents NVIDIA T4 and A100 instances, and FuriosaAI NPUs.
- Windows Server. Useful if you need Windows Server: KakaoCloud boots Windows Server 2019 and 2022 images you reach over RDP.
- Data in South Korea. Useful if your users or data must stay in South Korea: KakaoCloud runs in Seoul, with separate regions for public institutions.
Use boat when
you run AI agents that need a ready sandbox: coding agents, a desktop, HTTPS URLs, per-minute snapshots and fork, without building that layer on top of plain VMs.
- Coding agents you drive from outside. On a plain VM you write and maintain the code that starts, prompts and watches every agent, and real AI keys sit on the disk. boat's credentials proxy keeps those keys outside the sandbox.
- A desktop and browser for computer use. Without it, each computer-use agent needs you to install a desktop, a VNC server and a browser on the VM, then build your own screen and input layer and keep it working.
- A sandbox lifecycle built in. Cloning a KakaoCloud VM means making an image, then booting a new instance from it, and its docs recommend stopping first. A kept public IP keeps billing even when unused.
Questions people ask
Is boat cheaper than KakaoCloud?
In most of the scenarios we priced, yes: one coding agent on workdays comes to $20 a month on boat and $26 on KakaoCloud, and one agent running all month comes to $26 versus $94. 50 agents in parallel come to $317 versus $1,153. For sandboxes for your users, KakaoCloud's shared-CPU burstable instances come to $187 against $200 on boat.
Is KakaoCloud a sandbox for AI agents?
No. KakaoCloud Virtual Machine is a general cloud server: you get a VM with an OS image and build everything else yourself. boat is made for agents, with coding agents, a desktop, HTTPS URLs and snapshots built into every sandbox.
Can I run Docker on KakaoCloud and boat?
Yes on both. A KakaoCloud VM gives you root, so you can install Docker yourself; nested virtualization is not documented there. On boat, Docker, systemd, root and /dev/kvm are ready in every sandbox.
Where does my data run?
KakaoCloud's general region is kr-central-2 in Seoul, South Korea. boat runs only in the EU (Germany, Finland and France), with EU data residency and a zero data retention option.
Does boat have GPUs or Windows like KakaoCloud?
No. boat has no GPUs and no Windows guests; it runs Linux VMs on AMD Ryzen 9 9950X hosts and has Wine docs for Windows apps. KakaoCloud has T4 and A100 GPU instances and Windows Server images.
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. KakaoCloud lists ISMS-P, CSAP, several ISO/IEC 27000 certificates and CSA STAR Gold.
Start a sandbox
Install the CLI, then run boat new.
7-day free trial