boat vs Omnara

Omnara is an open-source platform that runs agents you configure and sends their commands to machines from Blaxel and eleven other providers. boat is the machine itself: a full VM with Docker, a desktop, public HTTPS URLs and Claude Code or Codex ready to drive from outside.

Where boat stands out

A full VM with a real disk

On boat every sandbox has its own kernel, with root, systemd, Docker and /dev/kvm ready. Disk is fixed per size (51 GB on the default size), and boat saves it every minute and at each stop. Omnara's default machines are Blaxel microVMs whose files live in memory, and Docker is not documented there. Files kept on the machine count against its billed memory, and if a pool deletes idle machines, the replacement starts from the base image.

Ports you can open

boat gives each sandbox token-gated HTTPS URLs on up to 50 ports, SSH and raw TCP/UDP. The IP changes at each resume, and some sandboxes have no IPv4 of their own, so a raw port there works over IPv6 only. Omnara machines connect out to its control plane only. That is good for safety, but nothing outside can reach a server on them.

Claude Code and Codex installed

boat installs seven coding-agent harnesses in each sandbox, including Claude Code, Codex, pi and OpenCode. You drive them from outside with a prompt, steer and events API. Omnara runs its own agent loop, so you write the instructions and tools for each agent yourself. Claude Code and Codex can call Omnara as an MCP server, but they do not run as harnesses inside its machines.

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.

Omnara

Omnara is an open-source managed agent platform, a self-hostable alternative to Claude Managed Agents and OpenAI's Agents API.

You write an agent as a small YAML config: a model, tools, approvals and machine sources. Omnara runs the agent loop in its control plane and sends tool calls like run_command to machines. A machine is either a computer you connect with its omnarad daemon or a sandbox from a pool on Blaxel, Daytona, E2B, Modal, Unikraft Cloud or seven other providers. Claude Code and Codex can also use Omnara as an MCP server. It is built for teams shipping Slack bots, code review agents and agents inside their own product. It is a different kind of product from boat: the two overlap where an agent needs a machine to work on.

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 workloadboat: price rankOmnara: price rank
Software factory
100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month
On Omnara's default pool, files live in memory and Docker is not documented. You set it up: create a Freestyle pool, which runs Docker through nested virtualization and bills through Freestyle. boat gives each agent a full VM with a 51 GB disk on the default size, Docker and public HTTPS URLs.
5th cheapest of 365
$634 a month
276th of 365, with limits
$4,044 a month
Always-on AI employees
50 machines 24/7 · 2 vCPU / 4 GB
Omnara is strong at the agent side: Slack, Discord, approvals and scheduled launches. An always-on Omnara machine bills its memory every hour it is online, and idle machines sleep to stop runtime charges. A running boat sandbox bills at full rate, busy or idle.
6th cheapest of 361
$657 a month
27th cheapest of 361
$5,198 a month
Code interpreter bursts
100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month
Omnara fits short bursts better: idle machines sleep and wake in milliseconds on Blaxel with memory kept. boat's $20 plan includes 12 starts per minute and 200 per day, and extra starts cost $0.03 to $0.07.
189th of 336, with limits
$2,000 a month
132nd of 336, with limits
$271 a month
Sandboxes for your users
300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month
Omnara does not publish how many hosted machines an organization can run at once, so 300 sandboxes for your users is not confirmed there. boat's $100 plan runs 300 at once and includes 30 starts per minute and 840 per day.
3rd cheapest of 349
$209 a month
142nd of 349, with limits
$1,725 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
130th of 349, with limits
$8,234 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
165th of 359, with limits
$1,325 a month
CI runners
40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month
Both fit. On Omnara you pick a pool whose provider gives the CPU and image your build needs; the default pool sizes by memory and does not publish CPU. boat runs Docker on every size, up to 8 vCPU on the $20 plan, with 200 included starts a day.
2nd cheapest of 374
$86 a month
91st cheapest of 374
$795 a month
24/7 dev boxes
1 machine 24/7 · 4 vCPU / 8 GB
Omnara machines connect out to the control plane only, with no SSH or open ports. For direct access, you connect your own computer with the omnarad daemon. boat gives SSH and HTTPS URLs, and the IP changes at each resume.
19th cheapest of 371
$26 a month
306th of 371, with limits
$208 a month
Computer-use agents
50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month
Omnara documents no desktop, live screen or computer-use tools, and its machines take no inbound connections. boat streams a 1920x1080, 60 fps desktop with Chrome and computer tools. boat has no GPU: graphics run on the CPU.
7th cheapest of 367
$200 a month
282nd of 367, with limits
$1,192 a month
Browser-use agents
50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month
Omnara's default machines include Playwright with headless Chromium, so browser agents work. A live view of the browser is not documented.
7th cheapest of 356
$200 a month
112th of 356, with limits
$596 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
43rd cheapest of 343
$404 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
204th of 327, with limits
$8,073 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
61st of 351, with limits
$397 a month
GPU eval jobs
10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month
Not ranked: misses a requirementNot 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 requirementNot ranked: misses a requirement

Key features

FeatureboatOmnara
What you getA machine for your agent: a full VM you control by API, SSH or a coding agentAn agent platform: model, tools, approvals and history, plus machines from sandbox providers
IsolationEvery sandbox is a full VM with its own kernelDefault pool: Blaxel microVM with its own kernel; other pools depend on the provider, from containers to VMs
Docker and nested virtualizationDocker, root, systemd and /dev/kvm on every sandboxRoot on Ubuntu 26.04 in the default pool; Docker and /dev/kvm are not documented there. You set it up: a Freestyle pool (billed by Freestyle) runs Docker through nested virtualization
DiskDisk fixed per size (51 GB on the default size), saved every minute and at each stop. If a machine fails, writes after the last good snapshot are lostDefault Blaxel machines keep files in memory, billed as memory. Sleeping machines keep their files. If idle deletion is on (configurable, 0 disables it), a replacement machine starts from the pool image. Memory stores keep files in Omnara outside the machine
Sleep and wakeStop and resume from disk, like a reboot: memory and running programs are not kept. A stopped sandbox has no compute charge, and boat keeps its disk for at least 14 daysIdle machines sleep and wake on the next tool call with memory and running processes kept (about 25 ms on Blaxel)
ForkFork a running sandbox into new ones; the fork uses the last 1-minute snapshot, so the last minute of writes can be missingNo fork tool; create_machine starts from the pool image. Freestyle pools boot every new machine from a whole-machine snapshot (memory included) that you build with the Freestyle CLI
Reaching the machineHTTPS URLs for up to 50 ports, token-protected; raw TCP/UDP on the sandbox IP (changes at each resume; IPv6 only on sandboxes without their own IPv4); SSHNo inbound path by design: the daemon connects out to the control plane only. You set it up: connect your own computer, where you control the network
Desktop and browser1920x1080, 60 fps desktop with Chrome, computer use and browser control, embeddable stream; no GPUPlaywright with headless Chromium preinstalled; a desktop, live view or computer-use tools are not documented
Coding agentsSeven harnesses installed, including Claude Code, Codex, pi and OpenCode, with a prompt, steer and events APIIts own agent loop on any model you configure, with a REST API, SDK, CLI, event stream and approvals; Claude Code and Codex can call Omnara as an MCP server
Keeping keys out of the machineOptional paid AI credentials proxy ($1 per day on, plus $1 per 1,000 requests): real keys of supported providers stay out of the sandbox. Other secrets still enter it. Switch a credential to another account when credits run outThe agent loop runs in the control plane, so model keys never need to be on the machine. Secrets the machine needs are injected as environment variables
Chat apps, schedules and approvalsNot built in; you set it up on boat's agent APISlack, Discord and GitHub launches, scheduled launches, webhooks and tool approvals built in
Self-hosting and your own machinesHosted only; you start from boat's image or a saved snapshotApache 2.0 control plane you can self-host; connect laptops, VPC servers or GPU boxes with the omnarad daemon
Scale and starts100 to 2,000 sandboxes at once by plan, with included starts per minute, hour and day. Extra starts are paid; boat never queues a startMachines at once per organization are not published; each pool has a max_total_machines cap you set
RegionsGermany, Finland and France; you cannot pick a region, and there is no US or Asia regionDefault pool region not published; Blaxel pools offer Oregon, Virginia, London and Frankfurt
Compliance and transparencySOC 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)
FeatureboatOmnara
Full VM (own kernel)YesYes
Docker insideYesWith your own setup
Nested virtualizationYesWith your own setup
RootYesYes
SystemdYesNot published
Desktop GUIYesNo documented way
BrowserYesYes
Computer-use APIYesNo documented way
SSHYesNo documented way
Persistent diskYesYes
Pause/resumeYesYes
Automatic snapshotsYesYes
Fork/cloneYesNo documented way
Live fork (no pause)YesNo documented way
Memory forkNo documented wayNo documented way
Custom image (Docker or snapshot)YesYes
≥24 h sessionsYesYes
Idle auto-stopNo documented wayYes
Wake on requestNo documented wayYes
HTTPS preview URLsYesNo documented way
Raw TCP inboundYesNot published
Public IPv4No documented wayNo documented way
Egress allowlistNo documented wayNo documented way
Secret proxyYesNo documented way
Secret proxy for any APIYesNo documented way
Preinstalled agentsYesNo documented way
Agent harness APIYesNo documented way
MCP serverNo documented wayYes
GPUNo documented wayNo documented way
WindowsNo documented wayNo documented way
Arm64No documented wayYes
Self-hostingNo documented wayYes
BYOCNo documented wayYes
Open sourceNo documented wayYes
Scoped API keysYesNot published
Spend limitsYesNot published
Usage per sandbox (API)YesNo documented way
Audit logsNo documented wayYes
SSONo documented wayYes
SOC 2No documented wayNo documented way
HIPAANo documented wayNo documented way
EU data residencyNo documented wayNo documented way
Zero data retentionYesNot published

Performance

boat28.06 runs/s
OmnaraNot measured

Omnara is not in the public benchmarks we use, so we cannot say which is faster. Its default pool runs on Blaxel, and it does not publish how much CPU a managed machine gets. boat did 28.06 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB, second of 15 in StarSling's run of 3 October 2026 (run 37073680030). Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. The boat fleet also has Ryzen 7 9700X hosts, and boat does not promise a CPU model.

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. Omnara 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.

Scenarioboat, a monthOmnara, 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$52 · Self-serve
One agent that never stops
4 vCPU / 8 GB, 50 GB disk, running 24/7 all month
$26 · $20/mo$218 · Self-serve
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$2,623 · Self-serve
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$745 · Self-serve · 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$1,647 · Self-serve · 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 unused time does not carry over. A default sandbox (4 vCPU / 8 GB) costs $0.036 per hour, billed per second while it runs, busy or idle. A stopped sandbox has no compute charge; saved snapshots above the included count and the AI credentials proxy still charge. Extra starts above the plan's included starts cost $0.03 to $0.07 each. A card is required, also for the 7-day free trial.

Omnara: The platform is free. Omnara-managed machines cost $0.0414 per GiB of memory per hour while online, with no separate CPU charge, plus $0.20016 per GiB of memory for every 30 days a machine is kept. Idle machines sleep and stop runtime charges. Omnara Cloud is prepaid with credits that are not refundable. Your own machines and model keys cost nothing on Omnara, but you still pay their providers.

Limits in those scenarios

Omnara does not publish how many managed machines an organization can run at once, so 50 agents in parallel, 300 sandboxes for your users and 1,000 jobs at once are all unconfirmed there. Docker is not documented on the default machines. Every pool sets a max_total_machines cap you raise yourself, and the API allows 100,000 reads and 50,000 writes per hour. Disk is not priced separately because the default machines keep files in memory. On boat, plans run 100 to 2,000 sandboxes at once. Each plan includes a fixed number of starts: 12 per minute and 200 per day on $20, 90 per minute and 2,400 per day on $2,000. Extra starts are paid, and a start above 3 times the included per-minute number fails with 429. boat never queues a start. So the 1,000-job burst pays for extra starts and hits the per-minute cap even on the $2,000 plan. A stopped sandbox keeps its disk for at least 14 days.

ScenarioboatOmnara
One coding agent, workdaysNone for this scenario
  • disk beyond 0 GiB: price unknown
  • egress price not published
One agent that never stopsNone for this scenario
  • disk beyond 0 GiB: price unknown
  • egress price not published
50 agents in parallelNone for this scenario
  • disk beyond 0 GiB: price unknown
  • egress price not published
  • how many machines can run at once: limit not published, probably fine at 50 (the typical published limit is 50)
Sandboxes for your usersNone for this scenario
  • not confirmed: how many machines can run at once (you need 300; the typical published limit is 50)
  • disk beyond 0 GiB: price unknown
  • egress price not published
1,000 short jobs at once
  • 2,400 starts/day included, paid above that (needs ~6,667)
  • 600 starts/hour included, paid above that (needs ~1,000)
  • starts above 90/min are paid, and starts above 270/min fail with 429
  • not confirmed: how many machines can run at once (you need 1,000; the typical published limit is 50)
  • disk beyond 0 GiB: price unknown
  • egress price not published
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 Omnara when

you want a ready agent platform (any model, approvals, Slack and scheduled launches) that you can self-host or point at your own machines, or when your agents mostly wait and you want them to sleep with memory kept.

  • A whole agent platform. Useful if you are building agents for Slack, Discord or your product: Omnara gives you approvals, schedules, event history and any model out of the box.
  • Open source and your own machines. Useful if you must self-host or run agents on your own servers, laptops or GPU boxes: Omnara is Apache 2.0 and connects any computer.
  • Sleep with memory kept, many providers. Useful if your agents mostly wait: idle machines sleep and wake in milliseconds with memory kept, on US or EU providers you choose.

Use boat when

your agent needs a full computer to work on: a VM with Docker, a disk, public URLs, a desktop and Claude Code or Codex installed. You can also point an agent platform at it as the machine.

  • A full VM with a real disk. An agent that runs docker compose, a database and a test suite needs Docker and room on disk. On Omnara's default pool you size up memory for files, or you set up a Freestyle pool for Docker. On boat, a machine failure loses the writes after the last good snapshot.
  • Ports you can open. When an agent builds a web app, someone has to open the preview, and a webhook needs somewhere to land. On Omnara you set it up: connect your own computer with its daemon, where you control the network.
  • Claude Code and Codex installed. Teams that already trust Claude Code or Codex want those exact agents, not a new loop to tune. boat's optional paid credentials proxy also keeps real keys of supported providers out of the sandbox.

Questions people ask

Is Omnara a sandbox?

Not by itself. Omnara is a control plane that runs agents and sends their commands to machines. The machines come from sandbox providers like Blaxel, Daytona, E2B and Modal, or from computers you connect. boat is the sandbox: each one is a full VM.

Is boat cheaper than Omnara?

It depends on the workload. In the scenarios we priced at list prices: one coding agent on workdays costs $20 a month on boat and $52 on Omnara. One agent running all month costs $26 versus $218. 50 agents in parallel cost $317 versus $2,623. 300 sandboxes for your users cost $200 versus $745, but Omnara does not confirm 300 at once. For 1,000 short jobs at once, Omnara comes to $1,647 and boat to $8,800, because boat charges for extra starts. boat's per-minute start cap refuses part of that burst, and Omnara does not confirm 1,000 at once. Omnara bills memory only and does not publish CPU per machine. Its idle machines sleep, so agents that mostly wait cost less there. boat's smallest plan is $20 a month.

Can I run Claude Code on Omnara?

Omnara runs its own agent loop on any model you pick, including Claude. Claude Code or Codex can call Omnara as an MCP server, but Omnara does not ship them as harnesses inside its machines. On boat, Claude Code, Codex and OpenCode are installed in every sandbox, and you drive them with a prompt, steer and events API.

Can I self-host Omnara or boat?

Omnara's control plane is open source under Apache 2.0 and free to self-host, and it can use machines you own. boat is a hosted service only. Its sandboxes run on AMD Ryzen 7 9700X and Ryzen 9 9950X hosts in Germany, Finland and France.

Where does my data run?

boat runs sandboxes in Germany, Finland and France. You cannot pick a region, and snapshot copies can be stored outside the EU. Omnara's Trust Center scopes its program to North American operations, and its machines run wherever the chosen provider runs them.

Is boat SOC 2 compliant?

No SOC 2 report today: 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. Omnara's Trust Center lists security controls monitored with Vanta.

Start a sandbox

Install the CLI, then run boat new.

7-day free trial

More comparisons