Agent 37 vs Instabox

Agent 37 starts each agent in a gVisor sandbox that sleeps with its memory kept and wakes on the next request, with a full VM type you can opt into and per-minute billing. Instabox, a free invite-only beta that launched today, gives every machine its own Firecracker microVM with root and Docker, in one 4 vCPU / 8 GiB size and up to 5 machines per account.

The companies and products

Agent 37

Agent 37 hosts AI agents, one isolated, always-on sandbox per user, behind one API.

You create an instance from a template (Hermes, OpenClaw, Claude Code, Codex, OpenCode, Pi, Grok or your own Docker image) and talk to it through a chat API, files, exec, SSH and HTTPS URLs. The default instance is a container enforced by gVisor; a VM type gives a full Linux machine with its own kernel for Docker, Kubernetes and Android emulators. It is built for companies that give each of their customers an agent.

Instabox

Instabox, made by InsForge, gives each AI agent its own Linux computer in the cloud. It launched on 2026-10-14 as an invite-only beta.

Each Instabox machine is a Firecracker microVM with its own kernel, root and Docker, started from one of four templates (agent-browser, universal, herdr or orca) or from your own snapshot. Every machine is 4 vCPU / 8 GiB with 20 GB of storage, and you drive it from a CLI, a TypeScript SDK or an HTTP API, with SSH, mosh, public port URLs and a desktop for computer use. It is aimed at developers who want a cloud box for a coding agent.

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. “2nd cheapest of 367” means 1 of the other 366 providers runs 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 workloadAgent 37: price rankInstabox: price rankboat: price rank
Software factory
100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month
Docker works on every Instabox machine but only on Agent 37's VM type; neither forks a running machine, and boat does, with sizes up to 16 vCPU / 32 GB.
2nd cheapest of 367
$551 a month
No published price3rd cheapest of 367
$634 a month
Always-on AI employees
50 machines 24/7 · 2 vCPU / 4 GB
Agent 37 is built for one always-on assistant per user that sleeps when idle and wakes on the next message; Instabox also pauses with memory, but a visit to a port URL does not wake it.
Cheapest of 366
$279 a month
No published price5th cheapest of 366
$657 a month
Code interpreter bursts
100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month
Cheapest of 345
$22 a month
No published price186th of 345, with limits
$2,000 a month
Sandboxes for your users
300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month
Agent 37 holds up to 200 instances on its top self-serve tier, and Instabox allows 5 machines per account, both counting stopped ones; boat plans go up to 2,000 at once.
30th of 361, with limits
$79 a month
No published priceCheapest of 361
$209 a month
RL rollouts
1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month
6th of 360, with limits
$363 a month
No published price34th of 360, with limits
$2,000 a month
Agent evals
250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month
Instabox's 5-machine cap limits how many evals run at once; Agent 37 goes to 200 after top-ups, and boat runs each eval on its own full VM.
90th of 364, with limits
$58 a month
No published price10th cheapest of 364
$200 a month
CI runners
40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month
CI jobs that build Docker images need Agent 37's VM type, while Instabox runs Docker on every machine; jobs that need /dev/kvm get it on boat and on Agent 37's VM type.
2nd cheapest of 377
$84 a month
No published price4th cheapest of 377
$86 a month
24/7 dev boxes
1 machine 24/7 · 4 vCPU / 8 GB
Both keep a machine and its disk; Instabox adds mosh and paste-anywhere SSH tokens, Agent 37 has a published price, and boat adds systemd and raw TCP.
3rd cheapest of 368
$12 a month
No published price18th cheapest of 368
$26 a month
Computer-use agents
50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month
Agent 37 adds a noVNC desktop through a template build, Instabox has a desktop template with a screenshot and click API, and boat streams a 60fps desktop in every sandbox.
Cheapest of 369
$53 a month
No published price6th cheapest of 369
$200 a month
Browser-use agents
50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month
48th of 367, with limits
$29 a month
No published price7th cheapest of 367
$200 a month
Scraping & bots (startup scale)
50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month
Cheapest of 346
$32 a month
No published price6th cheapest of 346
$90 a month
Mass scraping & bots (undetectable)
1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month
36th of 338, with limits
$648 a month
No published price71st of 338, with limits
$1,800 a month
Parallel red teaming
200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month
Neither Agent 37 nor Instabox can fork a running machine, so each attempt starts from scratch; boat forks a running sandbox to try many paths at once.
Cheapest of 360
$43 a month
No published price2nd cheapest of 360
$200 a month
GPU eval jobs
10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month
Can't run itNo published priceCan't run it
macOS agents
5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month
Can't run itNo published priceCan't run it

Key features

FeatureAgent 37Instaboxboat
IsolationgVisor sandbox by default; full VM with its own kernel as an opt-in typeFirecracker microVM with its own kernel and root for every machineEvery sandbox is a full VM with its own kernel, on dedicated AMD Ryzen 9 9950X hosts
Docker and systemdDocker only on the VM type, at 2.5x the default compute rate; no init system, even on the VM typeDocker and Docker Compose run inside every machine; systemd not documentedDocker, Compose and systemd in every sandbox
Nested virtualization/dev/kvm on the VM type, with a ready Android templateNot documented/dev/kvm in every sandbox, Android emulator works
Machine sizesShapes from 2 vCPU / 4 GB up to 16 vCPU / 32 GB after a first top-upOne size for everyone: 4 vCPU / 8 GiB with 20 GB storage; other sizes are refusedSeveral sizes, up to 8 vCPU / 16 GB on the $20 plan and 16 vCPU / 32 GB on $100 and above
Machines per account1 before a top-up, then 10, 50 and 200 as top-ups grow; sleeping and stopped ones count; email for more5, in any state: paused and stopped ones countPlans go up to 2,000 sandboxes at once
Sleep and pauseOpt-in auto-sleep after a quiet period, memory checkpointed, wakes on the next requestPause keeps memory and running programs; auto-pause after 30 idle minutes; a visit does not wake itStop keeps the disk and costs nothing; start boots the machine again
Fork and snapshotsNo fork; daily backups plus one on-demand backup, restored in place onlyFork copies a stopped machine only; up to 10 snapshots saved by hand, and saving stops a running machine unless the server allows a live saveFork a running sandbox; disk saved every minute; named snapshots as shareable templates
Custom imagesYour own linux/amd64 Docker image as a templateFour templates (agent-browser, universal, herdr, orca) or your own snapshotTemplates from named snapshots, no Dockerfile builds
Coding agentsHermes, OpenClaw, Claude Code, Codex, Grok, OpenCode, Pi, with a chat API per instanceTemplates and saved agent sign-ins in snapshots; you run the agent yourselfSeven agents preinstalled, driven from outside with a prompt, steer and events API
Keeping API keys out of the sandboxManaged LLM, search and Composio through a per-instance token; your own keys go in as env varsVault writes your secrets into each machine as environment variables or filesCredentials proxy for model providers and any HTTPS API; real keys never enter the sandbox
Desktop and computer usenoVNC desktop through a template build or the Ubuntu VM templateLinux desktop with a browser on one template, screenshot and click API, shareable view or control links60fps Linux desktop with Chromium and a computer-use agent in every sandbox
Port URLsHTTPS URL per port, signed URLs, custom domains; no public IPHTTPS on up to 16 ports, open to anyone with the URL; offline while the machine is pausedHTTPS on up to 50 ports, token-protected by default; raw TCP too
SSHSSH through a WebSocket tunnel, root by defaultSSH gateway with short-lived certificates, paste-anywhere SSH tokens and moshSSH into every sandbox
RegionsDedicated servers in the United StatesNot publishedEU only (Germany, Finland, France), EU data residency
Every feature Battleships tracks (43)
FeatureAgent 37Instaboxboat
Full VM (own kernel)NoYesYes
Docker insideNoYesYes
Nested virtualizationNoNot publishedYes
RootYesYesYes
SystemdNot publishedNot publishedYes
Desktop GUIYesYesYes
BrowserYesYesYes
Computer-use APIYesYesYes
SSHYesYesYes
Persistent diskYesYesYes
Pause/resumeYesYesYes
Automatic snapshotsYesNoYes
Fork/cloneYesYesYes
Live fork (no pause)YesNoYes
Memory forkNoNoNo
Custom image (Docker or snapshot)YesYesYes
≥24 h sessionsYesYesYes
Idle auto-stopYesYesNo
Wake on requestYesNoNo
HTTPS preview URLsYesYesYes
Raw TCP inboundNot publishedNot publishedYes
Public IPv4NoNoNo
Egress allowlistNoNot publishedNo
Secret proxyYesNoYes
Secret proxy for any APINoNoYes
Preinstalled agentsYesNot publishedYes
Agent harness APIYesNot publishedYes
MCP serverNoNot publishedNo
GPUNoNoNo
WindowsNoNoNo
Arm64NoNot publishedNo
Self-hostingNoNot publishedNo
BYOCNoNot publishedNo
Open sourceNoNoNo
Scoped API keysNot publishedYesYes
Spend limitsYesNot publishedYes
Usage per sandbox (API)YesYesYes
Audit logsNot publishedNot publishedNo
SSONoNoNo
SOC 2YesNot publishedNo
HIPAANoNot publishedNo
EU data residencyNoNot publishedYes
Zero data retentionNot publishedNot publishedYes

Performance

Agent 37Not measured
InstaboxNot measured
boat26.69 runs/s

Neither Agent 37 nor Instabox has a result on the hpc-sandbox-benchmarks Node.js build-and-test loop, so there is no comparable measurement for either. boat did 26.69 runs per second there at 4 vCPU / 8 GB. Agent 37 says its default instances share host resources and sells dedicated cores on its Performance type; Instabox says a new machine's shell is ready in under a second, which has not been measured.

Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Agent 37 and Instabox have 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.

ScenarioAgent 37, a monthInstabox, a monthboat, 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
$6 · Wallet (any top-up)No published price$20 · $20/mo
One agent that never stops
4 vCPU / 8 GB, 50 GB disk, running 24/7 all month
$12 · Wallet (any top-up)No published price$26 · $20/mo
50 agents in parallel
50 machines of 4 vCPU / 8 GB at once, 8-hour sessions, 1,100 sessions a month, Docker inside
$329 · Wallet ($100 lifetime top-ups)No published price$317 · $20/mo × 3 seats
Sandboxes for your users
2 vCPU / 4 GB, 30,000 sessions of 10 minutes a month, up to 300 at once
$33 · Wallet ($250 lifetime top-ups) · with limitsNo published price$200 · $100/mo × 2 seats
1,000 short jobs at once
2 vCPU / 4 GB, 1,000 at once, 3 minutes each, 200,000 jobs a month
$73 · Wallet ($250 lifetime top-ups) · with limitsNo published price$2,000 · $2000/mo · with limits

Agent 37: A prepaid wallet with no plan fee, metered per minute: $0.80 per vCPU and $0.70 per GB of RAM per month while running, and $0.09 per GB of disk per month even while stopped or asleep. Performance compute costs 4x and VM compute 2.5x the default rate.

Instabox: Instabox is free during the beta ("Free until we figure it out"), with 5 free machines per user, and has no published price, so no bill can be compared with Agent 37's yet.

Instabox: 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.

Limits in those scenarios

Agent 37 counts sleeping and stopped instances toward a cap of 200 on its top self-serve tier, so the scenarios with 300 or 1,000 machines at once do not fit it, and a 4 vCPU / 8 GB instance takes at most 20 GB of disk. Instabox allows 5 machines per account in any state, one 4 vCPU / 8 GiB size with 20 GB storage, and API exec calls of at most 10 minutes, 4 at once; with no published rate, none of the scenarios can be priced on it. boat's limit in the table is its cap on starts, which stops it in the 1,000-at-once burst.

ScenarioAgent 37Instaboxboat
One coding agent, workdays
  • egress price not published
  • no published per-resource rate
None that bite
One agent that never stops
  • egress price not published
  • no published per-resource rate
None that bite
50 agents in parallel
  • egress price not published
  • no published per-resource rate
None that bite
Sandboxes for your users
  • max 200 concurrent
  • egress price not published
  • no published per-resource rate
None that bite
1,000 short jobs at once
  • max 200 concurrent
  • shared vCPUs on this tier: heavy, sustained CPU use can be slowed by other tenants
  • egress price not published
  • no published per-resource rate
  • max 2,400 starts/day (needs ~6,667)
  • max 600 starts/hour (needs ~1,000)
  • bursts above 90 starts/min are queued (a burst of 333 takes ~4 min to start)

Which one should you use?

Use Agent 37 when

you host one mostly idle assistant per customer in the US that should sleep with its memory kept, wake on the next message and bill per minute at published rates.

  • Sleeps and wakes on the next request. For assistants that sit idle most of the day, an instance sleeps with its memory kept and wakes when the next message arrives.
  • Room to grow, with a published price. Agent 37 bills per minute from a wallet at published rates, offers sizes up to 16 vCPU / 32 GB and holds up to 200 instances.
  • Bring a Dockerfile, pick dedicated cores. Agent 37 builds templates from your own Docker image, and sells a Performance type on dedicated cores for steady speed.

Use Instabox when

you want to try a few free 4 vCPU / 8 GiB microVMs with Docker during its invite-only beta, and like mosh and memory pause.

  • A microVM with Docker on every machine. Every Instabox machine is a Firecracker microVM with its own kernel, root and Docker, with no special type to opt into.
  • Free to try during the beta. During the invite-only beta, each account can run 5 machines of 4 vCPU / 8 GiB for free; no later price is published.
  • SSH gateway with mosh. Its SSH gateway relays mosh and gives paste-anywhere SSH tokens, so a terminal session survives sleep and network changes.

Where boat fits

If your agents do hours of real work, like installs, builds, Docker and emulators, look at boat. Every sandbox is a full VM with systemd and /dev/kvm, you can fork it while it runs, and real API keys stay outside it. boat publishes its sizes, plans and EU regions today.

Fork work that is already running

Agent 37 has no fork, and its backups restore only in place. Instabox forks only a stopped machine, so its running programs end first. boat copies a running sandbox and the original keeps working.

A full Linux server by default

Agent 37's default type cannot run Docker, and its VM type costs 2.5x and has no init system. Instabox has one 4 vCPU / 8 GiB size and does not document /dev/kvm. boat has systemd, Docker and /dev/kvm in every size.

Keys the agent cannot read

On Agent 37 your own keys sit inside the instance as env vars, and Instabox's Vault writes secrets into the machine. One prompt injection can leak them. boat's proxy keeps real keys outside for any HTTPS API.

Use boat when your agents need a full VM with systemd and /dev/kvm, forks of running work, private port URLs and keys kept outside, at more than a handful of machines. Compare it one to one: boat vs Agent 37 and boat vs Instabox.

Questions people ask

Is Agent 37 cheaper than Instabox?

It can't be compared yet. Instabox is free during its beta and publishes no price, so there is no bill to set against Agent 37's per-minute wallet. Agent 37's and boat's prices for each scenario are in the table above.

Which is faster, Agent 37 or Instabox?

Neither has a measured CPU score on the benchmark used here, so there is no fair answer. Instabox says a new shell is ready in under a second, and its docs say a paused machine usually resumes in about a second. Agent 37 wakes from a memory checkpoint in a few seconds.

Can Agent 37 and Instabox run Docker?

Instabox runs Docker and Docker Compose on every machine. Agent 37 runs Docker only on its VM type, which bills compute at 2.5x; its default gVisor type cannot run a Docker daemon. boat runs Docker in every sandbox.

Can I fork a running machine on Agent 37 or Instabox?

No. Agent 37 has no fork, and its backups restore only onto the same instance. Instabox's fork copies a stopped machine. boat forks a running sandbox and keeps the original working.

How many machines can I run?

Agent 37 allows 1 instance before a top-up and up to 200 after $250 of top-ups, counting sleeping and stopped ones; beyond that you email them. Instabox allows 5 machines per account, paused and stopped ones included.

Where do Agent 37 and Instabox run?

Agent 37 runs on dedicated servers in the United States. Instabox does not publish its regions. boat runs only in the EU, in Germany, Finland and France.

Try boat

Install the CLI, then run boat new.

7-day free trial

More comparisons