boat vs OmniRun
Both give each agent its own small virtual machine on machines in Europe. boat gives it a full machine whose disk is kept after a stop and can be forked. OmniRun is a sandbox service that you can also run on your own Linux host.
Where boat stands out
A disk that stays, with snapshots and forks
A boat sandbox keeps its disk for at least 14 days after a stop, saves it every minute while it runs, and can be forked from its last 1-minute snapshot. OmniRun deletes a sandbox's disk when it is killed, and lists user snapshots and fork as planned for v0.2.
A whole machine you can size and SSH into
You pick the size per sandbox, from 2 vCPU / 4 GB to 16 vCPU / 32 GB (16 vCPU needs the $100 plan or higher). You get root, Docker, systemd, SSH and /dev/kvm for Android emulators and VMs. OmniRun sizes are fixed per template, it has no SSH, and it does not document nested virtualization.
Agents you drive from outside, keys kept outside
Claude Code, Codex and five other agent harnesses are installed, and you prompt, steer and follow them through one API. With the optional paid credentials proxy on, real keys for supported model providers and custom HTTPS APIs stay outside the sandbox.
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
OmniRun
OmniRun is a sandbox service from A14A B.V. that runs each agent session in its own Firecracker microVM.
Each OmniRun sandbox is a Firecracker microVM restored from a template snapshot. Templates exist for Python, Node, Rust, an XFCE desktop and agents like Claude Code and Codex. You drive it through a TypeScript SDK, REST API or CLI, on the managed service in Finland or on your own Linux host with KVM. Managed OmniRun is arranged per team: limits and price are agreed in a conversation, with no public plan.
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 | OmniRun: 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 outlive one run and room for big builds. OmniRun deletes the disk when a sandbox ends, so your code copies work out first, and sizes are fixed per template. boat keeps the disk for at least 14 days after a stop. | 5th cheapest of 365 $634 a month | No published price |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An assistant that lives on its machine all month needs its disk to survive. An OmniRun sandbox can be kept alive, but its disk is deleted when it ends. boat keeps the disk after a stop; auto-stop is 1 hour by default and you can turn it off on a paid plan. | 6th cheapest of 361 $657 a month | No published price |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Short code runs are what OmniRun is shaped for: a sandbox restores from a snapshot in about 250ms and is thrown away after. On boat, each start counts against the plan's start limits (200 a day included on the $20 plan). | 189th of 336, with limits $2,000 a month | No published price |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Both can give each of your users a VM. OmniRun can also run on your own hardware and has Claude Managed Agents, LangChain and LlamaIndex integrations. boat adds a kept disk, forks and up to 50 HTTPS URLs per sandbox. | 3rd cheapest of 349 $209 a month | No published price |
| 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 | No published price |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month | 10th cheapest of 359 $200 a month | No published price |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month CI that builds Android apps or VM images needs /dev/kvm inside the sandbox. boat exposes it on every sandbox; OmniRun does not document it. | 2nd cheapest of 374 $86 a month | No published price |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A dev box needs SSH and a disk that stays. OmniRun has no SSH and deletes the disk when a sandbox ends; you work through its commands and terminal API. boat has SSH and saves the disk every minute while the sandbox runs. | 19th cheapest of 371 $26 a month | No published price |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Both ship a Linux desktop you can click and type into from code. boat streams it at 1920x1080 and 60 fps with Chrome; OmniRun uses XFCE with noVNC and Firefox. | 7th cheapest of 367 $200 a month | No published price |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month | 7th cheapest of 356 $200 a month | No published price |
| 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 | No published price |
| 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 | No published price |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month Parallel red teaming branches from one prepared state, which needs fork. boat forks a running sandbox from its last 1-minute snapshot; OmniRun lists fork as planned for v0.2. | 3rd cheapest of 351 $200 a month | No published price |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month | Not ranked: misses a requirement | No published price |
| macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month | Not ranked: misses a requirement | No published price |
Key features
| Feature | boat | OmniRun |
|---|---|---|
| Isolation | Full VM with its own kernel, on AMD Ryzen bare-metal hosts dedicated to boat; vCPUs are shared | Firecracker microVM with its own kernel |
| Where it runs | Machines in Germany, Finland and France; you cannot pick a region; snapshot copies can be stored outside the EU | Managed hosts in Finland, or your own host |
| Disk after the sandbox ends | Kept for at least 14 days after a stop, longer with a saved snapshot; saved every minute while running and at stop; writes after the last snapshot can be lost if a machine fails | Copy-on-write disk deleted when the sandbox is killed; no persistent volumes yet. You set it up: copy files out through the commands API before it ends |
| Snapshots and fork | Snapshots every minute; fork a running sandbox from its last 1-minute snapshot | No user snapshots or fork yet; both planned for v0.2 |
| Pause | Stop keeps the disk; resume boots the machine again, like a reboot | Pause in memory; it does not survive a host restart |
| Lifetime | Auto-stop after 1 hour by default; set up to 30 days, or turn it off on a paid plan | 300 second default lifetime you can raise; killed after 15 minutes idle unless created with vault injection |
| Sandboxes at once and starts | $20 plan: 100 at once, 12 starts a minute and 200 a day included; extra starts cost $0.03 to $0.07, up to 1,200 a day, then refused | 3 sandboxes at once per user by default, more by agreement; no start rate published |
| Machine size | 2 vCPU / 4 GB up to 16 vCPU / 32 GB, picked per sandbox; 16 vCPU needs the $100 plan or higher; disk is fixed per size | Fixed per template; custom templates are built by the operator |
| SSH | Yes | No; access through the commands and terminal API |
| Nested virtualization | Yes, /dev/kvm for Android emulators and VMs | Not documented |
| Desktop and computer use | Built-in Linux desktop and Chrome streamed at 1920x1080, 60 fps, with computer tools for agents | XFCE desktop with noVNC, Firefox and a mouse, keyboard and screenshot API |
| Inbound access | Up to 50 HTTPS URLs per sandbox, token-gated by default; public IPv6 address; SSH over IPv4 through a shared host:port; no dedicated IPv4 | HTTPS preview URL per port, public in the docs (the SDK also has a private option with a token); no dedicated IP |
| Outbound rules | Open internet, no provider-side allowlist | Internet off for base templates, or an opt-in hostname allowlist for HTTPS |
| Coding agents | Seven agent harnesses installed, with a prompt, steer and events API | Claude Code, Codex, Gemini CLI and OpenClaw templates run through the commands API, plus a Claude Managed Agents template and self-hosted worker |
| Keeping API keys out of the sandbox | Optional paid credentials proxy ($1 a day plus $1 per 1,000 requests): real keys for supported model providers and custom HTTPS APIs stay outside; other secrets enter the sandbox | OpenAI-compatible LLM proxy called with an OmniRun key; vault secrets are written into the VM at /tmp/.omnirun-env |
| Self-hosting | No | Yes, on one Linux host with KVM through its installer; source announced as Apache-2.0, not yet public |
| 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 | OmniRun |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | No documented way |
| Root | Yes | Not published |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | Yes |
| Browser | Yes | Yes |
| Computer-use API | Yes | Yes |
| SSH | Yes | No documented way |
| Persistent disk | Yes | No documented way |
| Pause/resume | Yes | Yes |
| Automatic snapshots | Yes | No documented way |
| Fork/clone | Yes | No documented way |
| Live fork (no pause) | Yes | No documented way |
| 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 | Yes |
| Secret proxy for any API | Yes | No documented way |
| Preinstalled agents | Yes | Yes |
| Agent harness API | Yes | No documented way |
| MCP server | No documented way | No documented way |
| 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 | Yes |
| BYOC | No documented way | Yes |
| Open source | No documented way | No documented way |
| Scoped API keys | Yes | Not published |
| Spend limits | Yes | Yes |
| Usage per sandbox (API) | Yes | No documented way |
| Audit logs | No documented way | Yes |
| SSO | No documented way | Not published |
| SOC 2 | No documented way | No documented way |
| HIPAA | No documented way | No documented way |
| EU data residency | No documented way | Yes |
| Zero data retention | Yes | Not published |
Performance
OmniRun 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 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB, second of 15. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. The boat fleet also has Ryzen 7 9700X hosts, which this run did not measure. OmniRun says a sandbox restores from a snapshot in about 250ms, which is start time, not CPU speed.
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. OmniRun 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 | OmniRun, 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 | No published price |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | No published price |
| 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 | No published price |
| 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 | No published price |
| 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 | No published price |
boat: Plans cost $20, $100, $500 or $2,000 a month. The whole plan price comes back as VM time for that month; unused time does not carry over. A running sandbox bills per second, $0.036 an hour at 4 vCPU / 8 GB. A stopped sandbox has no compute charge; extra saved snapshots and the optional AI credentials proxy still charge. The $20 plan runs up to 100 sandboxes at once and includes 200 starts a day; extra starts cost $0.03 to $0.07 each, up to 1,200 a day. A card is required, also for the 7-day free trial.
OmniRun: OmniRun has no public price list or self-serve plan: managed limits and price are agreed with each team, so the table shows no OmniRun price. Self-hosting has no OmniRun fee; you pay for your own hardware.
Limits in those scenarios
OmniRun publishes no price, so the table has no OmniRun figure in any scenario. Its disk is deleted when a sandbox ends, so work that must survive a stop needs your own copy step. Each user can run 3 sandboxes at once by default, more only by agreement. boat has no session time cap once auto-stop is off (default 1 hour). Each boat plan caps sandboxes at once (100 on $20) and charges for starts above its included rate; the $2,000 plan includes 90 a minute and refuses above 270. A stopped boat sandbox keeps its disk for at least 14 days; save a snapshot to keep it longer.
| Scenario | boat | OmniRun |
|---|---|---|
| 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 OmniRun when
you must run sandboxes on your own hardware, or need short throwaway runs with outbound internet switched off.
- Run it on your own hardware. Teams that must keep sandboxes on their own hardware can run the same release on one Linux host with KVM, with no OmniRun fee.
- Outbound traffic you can switch off. Useful for running untrusted code that should reach nothing: base templates can start with no route out at all.
- Fast throwaway starts. Good for short code runs: a sandbox restores from a template snapshot in about 250ms and is gone when it ends.
Use boat when
your agents do work that has to be kept: installs, builds, Docker, emulators and a browser on a machine whose disk is kept after a stop, forks, and is driven from outside by API.
- A disk that stays, with snapshots and forks. On OmniRun your code copies work out before a sandbox ends. A sandbox idle for 15 minutes is killed unless created with vault injection. Branching a run means rebuilding the state from the template.
- A whole machine you can size and SSH into. On OmniRun a build that needs more memory than its template has needs a custom template from the operator. Emulators and VMs need /dev/kvm, which its docs do not mention.
- Agents you drive from outside, keys kept outside. OmniRun's agent templates run through its generic commands API, so you build the prompt-and-follow harness yourself. Its vault writes secrets into the VM, where code in the sandbox can read them.
Questions people ask
Is boat cheaper than OmniRun?
There is no public OmniRun price to compare: managed OmniRun is priced per team by agreement, so the table shows only boat. boat plans start at $20 a month, all of it VM time for that month, billed per second. Self-hosted OmniRun has no OmniRun fee, but you pay for your own hardware.
Does an OmniRun sandbox keep its files?
Its files last while the sandbox runs. Its copy-on-write disk is deleted when the sandbox is killed or times out. An idle sandbox is killed after 15 minutes unless created with vault injection. To keep work, copy it out first. A boat sandbox keeps its disk for at least 14 days after a stop and saves it every minute while it runs; writes after the last snapshot can be lost if a machine fails.
Can I self-host OmniRun or boat?
OmniRun can run on one x86_64 Linux host with KVM through its installer; its source is announced as Apache-2.0 but is not public yet. boat is a managed service on machines in Germany, Finland and France.
Can I run Docker and Android emulators in each?
boat runs Docker and exposes /dev/kvm, so Android emulators and VMs work. OmniRun gives each sandbox its own kernel with root, but its docs do not mention Docker or nested virtualization.
Which one has a desktop for computer use?
Both. OmniRun has an XFCE desktop with noVNC and an API for clicks, typing and screenshots. boat streams a Linux desktop and Chrome at 1920x1080 and 60 fps, with computer tools for agents.
Is boat SOC 2 compliant?
SOC 2 is in progress; there is no SOC 2 report today. No HIPAA BAA is documented. 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