boat vs Isorun
Both give each agent its own Linux VM with its own kernel. boat gives it a whole machine that keeps its work and comes with coding agents built in, while Isorun is strongest at microVMs that start in milliseconds and hibernate with their memory.
Where boat stands out
A whole machine that keeps its work
A boat sandbox is a full VM with root, systemd, Docker, SSH and /dev/kvm for Android emulators. Its disk is saved every minute and when it stops, and a stopped sandbox keeps its files for as long as you want.
Coding agents built in
Claude Code, Codex and five other agents come preinstalled, and you drive agents inside from outside through one API to prompt, steer and follow their events. Isorun has no agent API of its own and points you to adapters for other frameworks.
Keys for any API stay outside
boat's credentials proxy keeps real keys outside the sandbox for model providers and any HTTPS API. When one account runs out of credits, a credential can switch to another account of the same type.
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
Isorun
Isorun, Inc. calls itself the agentic runtime: an API that boots a hardware-isolated microVM for each agent session.
Every Isorun sandbox is a KVM microVM with its own guest kernel, booted from any OCI image and driven by a TypeScript SDK or REST API through exec, WebSocket shells and file calls. It can hibernate a running VM with its memory, snapshot it, and fork it into copies. It is in public beta since its v1.0.0 launch in June 2026, and it is built for teams that start many short agent sessions.
- Compliance
- SOC 2 Type II listed as soon
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 | Isorun: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month A coding agent needs Docker, root and long installs; boat documents all of these and ships the agents preinstalled, while Isorun leaves root and Docker undocumented. | 3rd cheapest of 367 $634 a month | 24th of 367, with limits $2,684 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An assistant that lives on its machine all month needs to keep its state; Isorun destroys an idle sandbox after 300 seconds unless you turn the timer off or hibernate it. | 5th cheapest of 366 $657 a month | 17th cheapest of 366 $4,031 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Isorun is built for this: microVMs that start in milliseconds, run a short job and hibernate with their memory. | 186th of 345, with limits $2,000 a month | 106th of 345, with limits $180 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month boat publishes how many sandboxes each plan runs at once, while Isorun's concurrency limits are not published yet. | Cheapest of 361 $209 a month | 116th of 361, with limits $1,134 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month Isorun can fork a running VM with its memory, which suits many rollouts that branch from the same warm state. | 34th of 360, with limits $2,000 a month | 91st of 360, with limits $5,478 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 | 144th of 364, with limits $882 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month CI runners that build Android apps or VM images need nested virtualization, which boat has and Isorun does not document. | 4th cheapest of 377 $86 a month | 69th cheapest of 377 $530 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A dev box needs SSH and a disk that survives; boat has both, and Isorun has no SSH access. | 18th cheapest of 368 $26 a month | 281st of 368, with limits $162 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Both ship a Linux desktop with Chromium; boat streams it at 60fps, while Isorun serves noVNC on a port and needs at least 2 vCPU and 4 GiB. | 6th cheapest of 369 $200 a month | 16th cheapest of 369 $794 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 | 75th of 367, with limits $397 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 | 23rd cheapest of 346 $268 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 | 163rd of 338, with limits $5,365 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 | 25th of 360, with limits $262 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 | Isorun |
|---|---|---|
| Isolation | Full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts | KVM microVM with its own guest kernel |
| Start time | Seconds; the machine boots from its disk | Milliseconds; median cold start 16.83 ms in ComputeSDK's benchmark |
| Root, systemd and Docker | Root, systemd and Docker, all documented | Any OCI image; root, systemd and Docker are not documented |
| Nested virtualization | Yes, /dev/kvm for Android emulators and VMs | Not documented |
| Idle behavior | Runs until you stop it or its time to live ends; a stop keeps the disk | Destroyed after 300 seconds idle by default; set the timer to 0 to keep it, and idle time is billed |
| Saving state | Disk saved every minute and on stop; resume boots the machine again | Hibernate keeps memory, processes and files, free while paused; you call it yourself |
| Fork | Fork a running sandbox with its disk | Fork a running sandbox with its memory into many children |
| Shell access | SSH, plus exec and terminal APIs | exec and WebSocket terminal; no SSH |
| Inbound access | HTTPS URLs for up to 50 ports, token-protected by default; raw TCP; dedicated IPv4 or IPv6 | HTTPS and WebSocket proxy URL for any port under the API host; no raw TCP or public IP |
| Outbound rules | Open internet, no provider-side allowlist | Allow and deny lists by CIDR, IP, hostname or wildcard, enforced on the host |
| Keeping API keys out of the sandbox | Credentials proxy for model providers and any HTTPS API; switch a credential to another account when credits run out | Host-side proxy for five services: OpenAI, Anthropic, Gemini, GitHub API and Hugging Face |
| Coding agents | Seven agents preinstalled, with a prompt, steer and events API | No agent API of its own; adapters for Claude Managed Agents, OpenAI Agents, LangChain and CrewAI |
| Desktop and browser | Linux desktop with Chromium, streamed at 60fps, with computer use | Computer-use image with Xvfb, Chromium and noVNC on port 6080 |
| Regions | EU only (Germany, Finland, France) | US or EU, one per organization |
| 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 | Isorun |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | No |
| Root | Yes | Not published |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | Yes |
| Browser | Yes | Yes |
| Computer-use API | Yes | Yes |
| SSH | Yes | No |
| 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 | Yes |
| Custom image (Docker or snapshot) | Yes | Yes |
| ≥24 h sessions | Yes | Yes |
| Idle auto-stop | No | Yes |
| Wake on request | No | No |
| HTTPS preview URLs | Yes | Yes |
| Raw TCP inbound | Yes | No |
| 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 | No |
| MCP server | No | Yes |
| GPU | No | No |
| Windows | No | No |
| 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 | Yes |
| Audit logs | No | Yes |
| SSO | No | No |
| SOC 2 | No | No |
| HIPAA | No | No |
| EU data residency | Yes | Yes |
| Zero data retention | Yes | Not published |
Performance
Isorun has no result on the hpc-sandbox-benchmarks CPU test, so there is no comparable measurement of CPU speed. boat did 26.69 runs per second on that test at 4 vCPU / 8 GB. Isorun's measured strength is start time: ComputeSDK recorded a median cold start of 16.83 ms.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Isorun 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 | Isorun, 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 | $36 · Pay as you go (public beta) |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $162 · Pay as you go (public beta) |
| 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 | $1,758 · Pay as you go (public beta) |
| 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 | $496 · Pay as you go (public beta) · 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 | $1,096 · Pay as you go (public beta) · with limits |
boat: Plans start at $20 a month, and the whole fee is usage credit. You pay per second while a sandbox runs, and a stopped sandbox costs nothing and keeps its disk.
Isorun: Isorun has no plan fee and bills per second for the vCPU and RAM you set, from create until destroy or the idle timer, so idle time counts until then. A hibernated sandbox costs nothing, and new accounts get $50 of free credit.
Limits in those scenarios
Isorun does not publish how many sandboxes can run at once, so the 300 sandboxes for your users and the 1,000-job burst are not confirmed to fit, and 50 agents in parallel is probably fine but unstated. Its default 300-second idle timer destroys a sandbox, so an always-on agent needs the timer turned off. boat publishes its limits, and they bite at the burst: at 1,000 short jobs at once, the $2,000 plan queues starts above 90 a minute.
| Scenario | boat | Isorun |
|---|---|---|
| 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 |
|
| Sandboxes for your users | None that bite |
|
| 1,000 short jobs at once |
|
|
Which one should you use?
Use Isorun when
you start very many short sessions that must boot in milliseconds, hibernate or fork with their memory, run in the US, or reach only an allowlist of hosts.
- Millisecond starts and memory hibernation. Useful for many short sessions: microVMs start in milliseconds and hibernate with memory, paying nothing while paused.
- Outbound rules on the host. Isorun can limit which hosts a sandbox reaches, enforced outside the VM, with ready profiles for common agents.
- A US region. Teams that need their sandboxes in the US can pick a US region for their organization.
Use boat when
your agents do real work for hours: coding agents with Docker, emulators, SSH and a browser, on a machine that keeps its disk and keeps your keys outside.
- A whole machine that keeps its work. On Isorun an idle sandbox is destroyed after 300 seconds by default, so an agent that waits on a human loses its work unless your code hibernates it or turns the timer off and pays for the idle time.
- Coding agents built in. Without it you install the agent in each image and build the start, stream and follow-up plumbing yourself, or route everything through a third party's agent service.
- Keys for any API stay outside. Isorun's proxy covers five services, so keys for your database, Stripe or any other API still sit inside a machine the agent controls, where one prompt injection can leak them.
Questions people ask
Is boat cheaper than Isorun?
In the four scenarios where boat fits, yes: for example one agent that never stops is $26 a month on boat against $162 on Isorun. For 1,000 short jobs at once neither is confirmed to fit, since boat queues starts and Isorun does not publish its concurrency limit.
Does an Isorun sandbox keep running when idle?
Not by default. Isorun destroys a sandbox after 300 seconds idle; you can raise the timer, set it to 0 and keep paying for idle time, or hibernate it yourself. A boat sandbox keeps running until you stop it, and a stop keeps its disk.
Can boat hibernate a sandbox with its memory like Isorun?
No. boat saves the disk every minute and when the sandbox stops, so files survive, but resuming boots the machine again. Isorun keeps memory and processes across hibernate and resume.
Can I SSH into an Isorun sandbox?
Isorun's docs list exec, WebSocket terminals and an HTTPS port proxy, but no SSH. boat gives every sandbox SSH, raw TCP and HTTPS URLs for up to 50 ports.
Can I run Docker and Android emulators in each?
boat documents root, systemd, Docker and /dev/kvm for Android emulators and VMs. Isorun gives each sandbox its own kernel but does not document Docker inside or nested virtualization.
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.
Start a sandbox
Install the CLI, then run boat new.
7-day free trial