boat vs microsandbox
Both give every sandbox its own kernel. microsandbox is an open-source microVM runtime for your own machine with a cloud in private beta, while boat is a cloud of full VMs where fork, hosting, a desktop and coding agents already work.
Where boat stands out
The whole machine works in the cloud today
boat is a self-serve cloud where every sandbox is a full VM with root, systemd, Docker and /dev/kvm, a disk saved every minute, and fork of a running sandbox. On microsandbox cloud, which is still a private beta, forking, full snapshots, live resize and the flat root disk its Docker recipe uses are all local-only for now.
Show what the agent built
boat gives each sandbox HTTPS URLs on up to 50 ports, token-protected by default, plus raw TCP and its own IP. microsandbox cloud does not accept published ports, so the only way in is SSH through its hostname.
Agents and a desktop built in, on a faster CPU
Claude Code, Codex and five other agents come preinstalled, and you drive agents inside from outside with a prompt, steer and events API, next to a 60fps desktop with computer use. On the same benchmark at 4 vCPU / 8 GB, boat did 26.69 runs per second and microsandbox did 21.14.
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
microsandbox
microsandbox is an open-source microVM runtime for running untrusted and AI-generated code, with a managed cloud in private beta.
Each sandbox is a small virtual machine with its own Linux kernel, started from an OCI image, built on libkrun and driven from Rust, TypeScript, Python, Go or Ruby SDKs, a CLI or an MCP server. The local runtime is free and runs on macOS, Linux and Windows hosts; the cloud runs the same sandboxes on its hardware, but access is by request and many features stay local-only for now. It is built for developers who want hardware isolation they can read the code of and run anywhere.
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 | microsandbox: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month A software factory needs Docker, preview URLs and long builds; microsandbox cloud has no published ports yet, and boat ran the CPU benchmark about a quarter faster. | 3rd cheapest of 367 $634 a month | 261st of 367, with limits $4,027 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB Both keep a disk between runs, but boat snapshots it every minute and has built-in agents you drive from outside, while microsandbox's cloud is still a private beta. | 5th cheapest of 366 $657 a month | 56th of 366, with limits $6,080 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Short code runs fit microsandbox well: Builder lets 1,000 sandboxes run at once with no start rate published, while boat's start limits queue big bursts. | 186th of 345, with limits $2,000 a month | 126th of 345, with limits $263 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Sandboxes for your users usually need a URL to show what they built, and microsandbox cloud does not publish ports yet, while boat gives HTTPS on up to 50 ports. | Cheapest of 361 $209 a month | 141st of 361, with limits $1,713 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month | 34th of 360, with limits $2,000 a month | 353rd of 360, with limits $8,236 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 | 161st of 364, with limits $1,318 a month |
| 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, which boat exposes and microsandbox does not document. | 4th cheapest of 377 $86 a month | 177th of 377, with limits $792 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A dev box that runs all month fits both, but microsandbox caps persistent sandboxes at 50 on Builder and gives no public IP or inbound ports. | 18th cheapest of 368 $26 a month | 128th of 368, with limits $236 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month boat has a 60fps desktop with computer use built in; microsandbox's VNC desktop is a do-it-yourself recipe that runs only locally. | 6th cheapest of 369 $200 a month | 251st of 369, with limits $1,185 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month Both run headless Playwright and Browser Use inside the VM; boat adds a visible Chromium on a desktop stream you can embed. | 7th cheapest of 367 $200 a month | 90th of 367, with limits $588 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 | 98th of 346, with limits $396 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 | 205th of 338, with limits $8,074 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 | 34th of 360, with limits $391 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 | microsandbox |
|---|---|---|
| Isolation | Full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts | libkrun microVM with its own guest kernel |
| CPU benchmark (4 vCPU / 8 GB) | 26.69 runs/s | 21.14 runs/s |
| Availability | Self-serve cloud, sign up and start | Cloud in private beta, access by request; local runtime free for anyone |
| Snapshots | Disk saved every minute and on stop, plus named snapshots you restore from | Cloud: disk snapshots of a stopped persistent sandbox, restored into a new one; full memory snapshots are local-only |
| Fork | Fork a running sandbox with its disk | Forking is local-only; not yet on the cloud |
| Pause and resume | Stop keeps the disk; start boots it again | Local runtime pauses in memory; cloud does not support pause |
| Docker and nested virtualization | Docker, root, systemd and /dev/kvm on every machine | Docker inside the VM is documented, but the main recipe needs a flat root disk that is local-only; no /dev/kvm inside the sandbox |
| Inbound access | HTTPS URLs for up to 50 ports, token-protected by default; raw TCP; dedicated IPv6 or IPv4 | Cloud: SSH through a platform hostname only; published ports are local-only |
| Outbound rules | Open internet, no provider-side allowlist | Allow or deny by group, IP, CIDR, domain and port; private and metadata addresses blocked by default |
| Keeping keys out of the sandbox | Credentials proxy for model providers and any HTTPS API; switch a credential to another account when credits run out | Placeholder secrets swapped outside the guest for allowed hosts, in headers by default |
| Coding agents | Claude Code, Codex and five more preinstalled, with a prompt, steer and events API | Recipes to install agents yourself; MCP server and Agent Skills so an outside agent can drive sandboxes |
| Desktop and computer use | 60fps desktop and Chromium with computer use, embeddable stream | Headless browsers in the VM; VNC desktop recipe is local-only, no computer-use API |
| Resize a sandbox | Recreate or restore from a snapshot at a new size | Live resize is local-only; on cloud you recreate the sandbox |
| Run it yourself | No; managed cloud only | Open-source runtime on your laptop or servers, plus BYOC by evaluation |
| 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 | microsandbox |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | No |
| Root | Yes | Not published |
| Systemd | Yes | Yes |
| Desktop GUI | Yes | No |
| Browser | Yes | Yes |
| Computer-use API | Yes | No |
| SSH | Yes | Yes |
| Persistent disk | Yes | Yes |
| Pause/resume | Yes | No |
| Automatic snapshots | Yes | No |
| Fork/clone | Yes | No |
| Live fork (no pause) | Yes | No |
| Memory fork | No | No |
| 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 | No |
| Raw TCP inbound | Yes | No |
| Public IPv4 | No | No |
| Egress allowlist | No | Yes |
| Secret proxy | Yes | Yes |
| Secret proxy for any API | Yes | Yes |
| 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 | Yes |
| BYOC | No | Yes |
| Open source | No | Yes |
| Scoped API keys | Yes | Not published |
| Spend limits | Yes | Yes |
| Usage per sandbox (API) | Yes | Yes |
| Audit logs | No | Yes |
| SSO | No | Yes |
| SOC 2 | No | No |
| HIPAA | No | No |
| EU data residency | Yes | No |
| Zero data retention | Yes | Not published |
Performance
Both were measured in the same StarSling hpc-sandbox-benchmarks run of 2026-09-29, a Node.js build-and-test loop at 4 vCPU / 8 GB. boat did 26.69 runs per second and microsandbox did 21.14, so boat got through about a quarter more CPU work in the same time. This measures CPU speed only, not start time.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite.
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 | microsandbox, 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 | $49 · Builder · with limits |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $236 · Builder · with limits |
| 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 | $2,618 · Builder · 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 | $738 · Builder · 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 | $1,640 · Builder · with limits |
boat: Plans start at $20 a month, and the whole fee is usage credit spent per second while a sandbox runs. A stopped sandbox costs nothing and keeps its disk.
microsandbox: microsandbox bills per second for the vCPU and memory you configure while a sandbox runs, plus a little for disk. Builder costs $49 a month and includes separate pools of vCPU-hours, memory and disk that do not swap for each other, then overage at the same rates; all prices are labelled private beta pricing and egress has no published price.
Limits in those scenarios
None of the microsandbox scenarios count as a clean fit, because the cloud is a private beta with prices that may change and no published egress price. Builder caps persistent sandboxes at 50 running at once, so the 50-agent fleet uses every slot, and sandboxes for your users that must stay around need Teams. boat has no persistent cap below its plan limit, but at 1,000 short jobs at once its $2,000 plan queues the starts, while microsandbox Builder lets 1,000 run at once.
| Scenario | boat | microsandbox |
|---|---|---|
| One coding agent, workdays | None that bite |
|
| One agent that never stops | 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 microsandbox when
you want open-source microVMs on your own machines or cloud, or strict per-domain egress rules, and can wait for cloud access.
- Open source that runs on your laptop. Useful if you want sandboxes on your own Mac, Linux or Windows machine: the Apache-2.0 runtime is free and needs no account.
- Fine-grained network rules. Useful if an agent must reach only chosen hosts: rules allow or deny by domain, IP range and port, and private addresses are blocked by default.
- Many sandboxes at once on Builder. Useful for big bursts of short code runs: Builder allows 1,000 sandboxes running at once with no published start rate.
Use boat when
your agents need a full cloud VM now: Docker, emulators, fork of a running sandbox, URLs for what they build, and coding agents you drive from outside.
- The whole machine works in the cloud today. An agent that has to branch its work, run Docker builds or start an Android emulator hits a wall on microsandbox cloud. The workaround is running the open-source runtime on hardware you manage yourself.
- Show what the agent built. An agent that builds a web app or an API needs a URL so a person or a test can reach it. On microsandbox cloud you would have to tunnel out over SSH yourself for every preview.
- Agents and a desktop built in, on a faster CPU. On microsandbox you install each agent from a recipe and write the code that starts and watches it, and every build or test takes about a quarter longer in that benchmark.
Questions people ask
Is boat cheaper than microsandbox?
In four of the five scenarios on this page, yes: one agent running all month comes to $26 on boat and $236 on microsandbox, and 50 agents in parallel come to $317 versus $2,618. For 1,000 short jobs at once, microsandbox Builder comes to $1,640 and boat to $2,000, and boat queues the starts.
Is microsandbox cloud available to everyone?
Not yet. Its docs say microsandbox cloud is in private beta and you have to request access. The open-source runtime is free for anyone to run on their own machine.
Can I open a port or preview URL on microsandbox cloud?
No. Published ports are local-only, and cloud sandboxes are reached over SSH through a platform hostname. boat gives HTTPS URLs on up to 50 ports, raw TCP and its own IP.
Can microsandbox fork a running sandbox?
On your own machine, yes. On the cloud, forking and full memory snapshots are not available yet; you can only snapshot the disk of a stopped sandbox and restore it into a new one. boat forks a running sandbox in the cloud.
Is boat faster than microsandbox?
On the hpc-sandbox-benchmarks CPU test at 4 vCPU / 8 GB, boat did 26.69 runs per second and microsandbox did 21.14. That is about a quarter more CPU work in the same time.
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