boat vs OpenPond
OpenPond sells AI model post-training, evaluation and dedicated inference. It also ships an open-source agent workspace with cloud sandboxes that you drive through its TypeScript SDK, up to 10 at a time on Pro. boat gives every agent a full Linux VM with SSH, root, Docker, /dev/kvm, a desktop and seven coding agents installed. boat plans run 100 to 2,000 sandboxes at once, and starts per minute and per day are limited per plan.
Where boat stands out
A whole Linux machine, not just an SDK
On boat each agent gets a VM you can SSH into, with root, systemd, Docker and /dev/kvm, plus a public IPv6 address with raw ports. OpenPond sandboxes are isolated, and you reach them through its SDK, CLI or MCP: exec, processes, PTY sessions and HTTPS preview ports. SSH, systemd and /dev/kvm are not documented there.
Fleets bigger than ten
OpenPond's Free plan runs 1 cloud sandbox at a time, and Pro runs up to 10. More needs an Enterprise contract. boat's $20 plan runs 100 at once, and 50 agents in parallel and 300 sandboxes for your users fit on self-serve plans. On boat the limit is starts: each plan includes a number of starts per minute, hour and day, and extra starts cost $0.03 to $0.07.
Coding agents and a desktop installed
boat installs seven coding-agent harnesses in each sandbox, among them Claude Code, Codex and OpenCode. You drive them from outside with a prompt, steer and events API. Each VM also streams a 1920x1080, 60 fps desktop with Chrome for computer use. OpenPond's Work runs its own hosted model, and its cloud sandboxes document no desktop.
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
OpenPond
OpenPond sells AI model post-training, evaluation and dedicated inference, plus an open-source agent workspace with cloud sandboxes.
The company leads with model post-training, evaluation and dedicated inference (Ponder Training, Ponder Inference). Its MIT-licensed workspace covers desktop, web, CLI and chat, with an agent harness and evals. Its cloud sandboxes are isolated Linux sandboxes that your server code creates and controls through the openpond-sdk TypeScript SDK: run commands, manage files, open preview ports, snapshot and fork. OpenPond does not publish the isolation technology. Use it when a hosted model must finish a task (Work) or when your own code drives raw sandboxes.
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. “5th cheapest of 365” means 4 of the other 364 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 | OpenPond: 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 and services at once. OpenPond Pro stops at 10 sandboxes, and more needs Enterprise. boat adds SSH and /dev/kvm on each machine. | 5th cheapest of 365 $634 a month | 308th of 365, with limits $5,758 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB Always-on assistants need machines that keep running. OpenPond Pro runs up to 10 at once. On boat, you turn off auto-stop (default 1 hour) on a paid plan. | 6th cheapest of 361 $657 a month | 332nd of 361, with limits $8,760 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month OpenPond can cost less for short code runs: it bills only running seconds and its Free plan costs $0. boat's start limits push a large burst of short runs onto a bigger plan. | 189th of 336, with limits $2,000 a month | 150th of 336, with limits $391 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month 300 sandboxes at once for your users need an Enterprise contract on OpenPond, where Pro stops at 10. On boat they fit on 2 seats of the $100 plan. | 3rd cheapest of 349 $209 a month | 163rd of 349, with limits $2,458 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month | 39th of 349, with limits $2,000 a month | 161st of 349, with limits $11,874 a month |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month | 10th cheapest of 359 $200 a month | 189th of 359, with limits $1,910 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month OpenPond documents no CI runner integration or SSH, so CI jobs run through its SDK or CLI. boat machines take a runner like any Linux server. | 2nd cheapest of 374 $86 a month | 192nd of 374, with limits $1,152 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB | 19th cheapest of 371 $26 a month | 343rd of 371, with limits $351 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month OpenPond documents no desktop in its cloud sandboxes; you set it up: add a desktop to your Dockerfile or OCI image. Its built-in browser tools drive the browser in the local desktop app. boat streams a 60 fps desktop with Chrome from each VM. | 7th cheapest of 367 $200 a month | 311th of 367, with limits $1,719 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month Browser agents need a browser inside the sandbox. On OpenPond you set it up: add a browser to your Dockerfile or OCI image. boat ships Chrome with browser control in each VM. | 7th cheapest of 356 $200 a month | 321st of 356, with limits $860 a month |
| Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month | 6th cheapest of 343 $90 a month | 313th of 343, with limits $582 a month |
| Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month | 73rd of 327, with limits $1,800 a month | 243rd of 327, with limits $11,642 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month | 3rd cheapest of 351 $200 a month | 298th of 351, with limits $570 a month |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month | Not ranked: misses a requirement | Not ranked: misses a requirement |
| macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month | Not ranked: misses a requirement | Not ranked: misses a requirement |
Key features
| Feature | boat | OpenPond |
|---|---|---|
| Isolation | Every sandbox is a full VM with its own kernel | Isolated sandbox; OpenPond does not publish the isolation technology |
| Root, systemd and nested virtualization | Root via sudo, real systemd, Docker and /dev/kvm on every machine | Docker commands are recognized; root, systemd and /dev/kvm are not documented |
| How you get in | SSH and SCP, plus an API and a CLI | SDK, CLI and MCP: exec, processes and PTY sessions; SSH is not documented |
| Sandboxes at once | 100 at once on $20, up to 2,000 on $2,000; starts per minute and per day are limited per plan | 1 on Free, up to 10 on Pro ($20/month), contract-defined on Enterprise |
| Machine size | 2 to 16 vCPU and 4 to 32 GB; disk fixed per size (51 to 251 GB); 16 vCPU needs the $100 plan | Tiers from 0.25 vCPU / 0.5 GiB to 4 vCPU / 8 GiB; templates allow up to 8 vCPU / 32 GiB / 100 GiB disk |
| Files after a stop | Disk saved every minute and at stop; no compute charge while stopped; disk kept at least 14 days, longer with a saved snapshot | Stopping ends compute billing and keeps the filesystem until you delete it; retained disk and snapshots keep billing storage |
| Snapshot and fork | Fork a running sandbox from its last 1-minute snapshot; saved snapshots as templates (disk, not memory) | Filesystem snapshots you can fork, restore or publish as templates (disk, not memory) |
| Custom images | Start from boat's image or a saved snapshot; you cannot boot your own Docker or OCI image | Start from a base, a snapshot, an OCI image or a Dockerfile |
| Reaching the sandbox | Up to 50 HTTPS URLs per sandbox, token-gated by default; public IPv6 address with raw ports (raw ports over IPv6 only on sandboxes without their own IPv4) | HTTPS preview ports, private with tokens or public; raw TCP and a public IP are not documented |
| Outbound internet | Open internet; egress not billed; no provider-side allowlist | Egress set to restricted, allow or block, with allowed hosts, in the sandbox network policy; egress price not published |
| Desktop and browser | 1920x1080, 60 fps desktop and Chrome in each VM, with computer tools; no GPU, graphics run on the CPU | None documented in cloud sandboxes; you set it up: add a browser or desktop to your Dockerfile. Built-in browser tools drive the local desktop app |
| Coding agents | Seven harnesses installed, among them Claude Code, Codex and OpenCode, with a prompt, steer and events API | Work runs an OpenPond-hosted model with file and command tools; you set up Claude Code or Codex yourself: install it in your image and drive it through exec |
| Keeping keys out of the sandbox | Optional paid AI credentials proxy ($1 per day on plus $1 per 1,000 requests) for model providers and HTTPS APIs you register; other secrets still enter the sandbox | OpenPond Connect links Google, GitHub and X accounts without handing raw tokens to agents |
| Spend caps | Plan price plus a wallet; with an empty wallet, sandboxes stop after a 24-hour grace period; no per-sandbox budget | Per-sandbox budget, spend, duration and idle limits set in code |
| 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 | OpenPond |
|---|---|---|
| Full VM (own kernel) | Yes | Not published |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | Not published |
| Root | Yes | Not published |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | No documented way |
| Browser | Yes | No documented way |
| Computer-use API | Yes | No documented way |
| SSH | Yes | No documented way |
| Persistent disk | Yes | Yes |
| Pause/resume | Yes | Yes |
| Automatic snapshots | Yes | No documented way |
| Fork/clone | Yes | Yes |
| Live fork (no pause) | Yes | Yes |
| Memory fork | No documented way | No documented way |
| Custom image (Docker or snapshot) | Yes | Yes |
| ≥24 h sessions | Yes | Yes |
| Idle auto-stop | No documented way | Yes |
| Wake on request | No documented way | No documented way |
| HTTPS preview URLs | Yes | Yes |
| Raw TCP inbound | Yes | Not published |
| Public IPv4 | No documented way | No documented way |
| Egress allowlist | No documented way | Yes |
| Secret proxy | Yes | No documented way |
| Secret proxy for any API | Yes | No documented way |
| Preinstalled agents | Yes | No documented way |
| Agent harness API | Yes | No documented way |
| MCP server | No documented way | Yes |
| GPU | No documented way | No documented way |
| Windows | No documented way | No documented way |
| Arm64 | No documented way | No documented way |
| Self-hosting | No documented way | No documented way |
| BYOC | No documented way | No documented way |
| Open source | No documented way | Yes |
| Scoped API keys | Yes | Yes |
| Spend limits | Yes | Yes |
| Usage per sandbox (API) | Yes | No documented way |
| Audit logs | No documented way | Not published |
| SSO | No documented way | Yes |
| SOC 2 | No documented way | Not published |
| HIPAA | No documented way | Not published |
| EU data residency | No documented way | Not published |
| Zero data retention | Yes | Not published |
Performance
OpenPond is not in the public benchmarks we use, so we cannot say which is faster. In StarSling's run of 3 October 2026 (run 37073680030), boat did 28.06 V8 Web Tooling Benchmark runs/s at 4 vCPU / 8 GB, second of 15. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure.
V8 Web Tooling Benchmark at a 4 vCPU / 8 GB target, runs per second, higher is better. Source: StarSling's run of 3 October 2026 (run 37073680030), median of 5 sandboxes. Providers ran on different hardware in different regions. StarSling gives a shared rank when the run cannot separate two providers. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. OpenPond is not in the public benchmarks we use, so we cannot say which is faster.
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. Start limits, per-start fees and plan minimums are in the price, boat's included.
| Scenario | boat, a month | OpenPond, 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 | $76 · Free |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $351 · Free |
| 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 | $3,783 · Pro · 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 | $1,075 · Pro · with limits |
| 1,000 short jobs at once 2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month | $8,800 · $2000/mo · with limits | $2,375 · Pro · with limits |
boat: Plans cost $20, $100, $500 or $2,000 a month. The whole plan price comes back as VM time for that month and does not carry over. A default 4 vCPU / 8 GB sandbox costs $0.036 an hour, billed per second while it runs. A stopped sandbox has no compute charge; saved snapshots above the included count and the AI credentials proxy still charge. Starts above the plan's included number cost $0.03 to $0.07 each. A card is required, also for the 7-day free trial.
OpenPond: OpenPond bills per second while a sandbox runs: $0.072 per vCPU-hour, $0.0234 per GiB-hour of memory and $0.00012 per GiB-hour of VM disk. A stopped sandbox keeps its root disk, and that disk bills storage at a rate OpenPond does not publish. Free costs $0 and includes limited monthly usage. Pro ($20/month) includes $20 of monthly usage; beyond that you add prepaid credit or turn on metered overage. The docs say the account's Billing page is authoritative for the rate card.
Limits in those scenarios
On OpenPond the limit is sandboxes at once: 1 on Free and 10 on Pro. So 50 agents in parallel, 300 sandboxes for your users and 1,000 short jobs at once need an Enterprise contract. OpenPond does not publish an egress price or the storage rate of a stopped sandbox's disk. On boat the limit is starts. The $20 plan includes 12 starts a minute and 200 a day. The 1,000-job burst needs more starts than the $2,000 plan includes (90 a minute, 2,400 a day): extra starts cost $0.03 to $0.07, starts above 270 a minute fail with 429, and boat never queues a start. Auto-stop defaults to 1 hour. A stopped boat sandbox keeps its disk for at least 14 days unless you save a snapshot. A machine failure loses what was written after the last 1-minute snapshot.
| Scenario | boat | OpenPond |
|---|---|---|
| One coding agent, workdays | None for this scenario |
|
| One agent that never stops | None for this scenario |
|
| 50 agents in parallel | None for this scenario |
|
| Sandboxes for your users | None for this scenario |
|
| 1,000 short jobs at once |
|
|
boat's own limits
- Plans start at $20 a month. The plan price comes back as VM time for that month. Unused time does not carry over. Source
- Starts (create, fork, resume): the $20 plan includes 12 a minute, 60 an hour and 200 a day. The $2,000 plan includes 90 a minute, 600 an hour and 2,400 a day. Extra starts cost $0.03 to $0.07. A start fails with 429 above 3 times the included starts per minute, or 6 times per hour or per day. Source
- A stopped sandbox has no compute charge. boat keeps its disk for at least 14 days. A saved snapshot keeps it longer; extra saved snapshots and the AI credentials proxy are paid. Source
- Auto-stop defaults to 1 hour. You can turn it off on a paid plan. The trial allows 2 hours at most. Source
- boat saves the disk every minute and at each stop. Memory, running programs and /tmp are not saved: a resume is like a reboot. A machine failure loses writes after the last good snapshot. Source
- Machines run in Germany, Finland and France only. Snapshot copies can be stored outside the EU. Source
- Public IPv6 and outbound IPv4. No dedicated IPv4. The IP changes at each resume. Source
- Hosts are AMD Ryzen 7 9700X and Ryzen 9 9950X. vCPUs are shared. No GPU. You cannot boot your own Docker image. Source
Which one should you use?
Use OpenPond when
you want an open-source agent workspace with a hosted Work API, sandboxes that boot from your Dockerfiles or OCI images, egress rules and per-sandbox spend caps, or light use on a $0 plan.
- Egress rules and spend caps per sandbox. Useful if untrusted code must stay offline or inside a host list, with a hard dollar cap on each sandbox.
- Build from a Dockerfile or OCI image. Useful if your environments already live as Dockerfiles or registry images: OpenPond boots a sandbox straight from them.
- A $0 plan and small sizes. Useful for light or small jobs: Free costs $0, and sizes start at 0.25 vCPU.
- An open-source workspace around the sandbox. Useful if you want an MIT-licensed agent workspace, harness and evals that you can read, change and run locally.
Use boat when
your agents need full Linux VMs in numbers: SSH, root, Docker and /dev/kvm, a desktop and browser, preinstalled coding agents you drive by API, and more than 10 running at once.
- A whole Linux machine, not just an SDK. Tools that expect a normal server, like remote editors, CI runners or emulators that need /dev/kvm, must be wrapped in SDK calls on OpenPond, or OpenPond does not document a way to run them.
- Fleets bigger than ten. An agent team, an eval run or a product that gives each user a sandbox reaches OpenPond's 10-sandbox limit fast. The extra jobs wait or need a sales call.
- Coding agents and a desktop installed. On OpenPond you set it up: install Claude Code or Codex in your image, then write the code that starts, prompts and watches it. A browser or desktop agent needs a browser and display that you add to the image yourself.
Questions people ask
Is boat cheaper than OpenPond?
In the 4 vCPU / 8 GB scenarios we priced from list rates on 2026-11-08, boat costs less. One coding agent on workdays is $20 a month on boat and $76 on OpenPond. One agent running all month is $26 versus $351. 50 agents in parallel are $317 versus $3,783, and OpenPond needs Enterprise above 10 at once. OpenPond's Free plan costs $0 and its sizes start at 0.25 vCPU, so light or small jobs can cost less there.
Can I SSH into an OpenPond sandbox?
OpenPond does not document SSH. You run commands, start processes and open PTY sessions through its SDK, CLI or MCP server. boat machines take SSH like any Linux server.
How many OpenPond sandboxes can run at once?
One on Free and up to 10 on Pro ($20/month). Enterprise limits are contract-defined.
Do files survive a stop?
Yes, on both. A stopped OpenPond sandbox keeps its filesystem until you delete it, and that disk bills storage at a rate OpenPond does not publish. A stopped boat sandbox has no compute charge and keeps its disk for at least 14 days; a saved snapshot keeps it longer.
Does OpenPond have a desktop or browser in the sandbox?
OpenPond documents none in its cloud sandboxes. You can add a browser or desktop to your own Dockerfile or OCI image. Its built-in browser tools control the browser in the local OpenPond desktop app. boat streams a 60 fps desktop with Chrome from each VM, with computer tools.
Is OpenPond open source?
The workspace, CLI, SDK, harness and evals are MIT licensed on GitHub. The cloud sandboxes run on OpenPond's hosted service and need an API key. boat is a hosted service too.
Is boat SOC 2 compliant?
SOC 2 is in progress, and boat has no SOC 2 report today. Live status is at status.ascii.dev and the data retention policy is at docs.boat.dev/data-retention.
Start a sandbox
Install the CLI, then run boat new.
7-day free trial