Vercel Sandbox vs Northflank

Vercel Sandbox is built for short runs of untrusted code in Firecracker microVMs, with sessions that stop after 24 hours. Northflank is a broad cloud platform where sandboxes run as long as you like, next to GPUs, databases and your own cloud account.

The companies and products

Vercel Sandbox

Vercel is a cloud platform for web apps and agents, and Vercel Sandbox is its product for running untrusted or agent-written code in isolated Linux microVMs.

Each sandbox is a Firecracker microVM with its own kernel, started from a Vercel image, your own OCI image, or a saved snapshot. You drive it with the JavaScript or Python SDK or the sandbox CLI: run commands, move files, open ports, and set network rules. It is aimed at code interpreters, app builders, test runs and agent tools, and Vercel's docs say it is not meant to run nonstop.

Northflank

Northflank is a London developer platform for running services, jobs and databases on Kubernetes, and Sandboxes is its product for running untrusted code.

A Northflank sandbox is a container image deployed as a service: CPU sandboxes run in Kata microVMs with their own kernel, and GPU sandboxes run under gVisor. You create and drive them from the JavaScript or Python SDK, the CLI or the dashboard, on Northflank's cloud or inside your own. It suits teams that already want one platform for apps, databases, GPUs and sandboxes.

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. “22nd cheapest of 345” means 21 of the other 344 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 workloadVercel Sandbox: price rankNorthflank: price rankbochka: price rank
Software factory
100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month
Neither has a desktop or browser for an agent to check the app it built; Vercel stops each session at 24 hours and Northflank keeps only attached volumes, while bochka ranks 3rd with a full desktop and a whole disk that survives.
244th of 367, with limits
$3,351 a month
159th of 367, with limits
$1,477 a month
3rd cheapest of 367
$634 a month
Always-on AI employees
50 machines 24/7 · 2 vCPU / 4 GB
An assistant that lives on its machine for weeks is cut off every 24 hours on Vercel and loses files outside volumes on each Northflank pause; bochka ranks 5th here.
257th of 366, with limits
$3,657 a month
173rd of 366, with limits
$1,365 a month
5th cheapest of 366
$657 a month
Code interpreter bursts
100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month
Vercel fits short code bursts best of the three: it bills CPU only while code uses it and starts fast, while Northflank's small per-second plans are a cheaper compromise.
22nd cheapest of 345
$216 a month
84th of 345, with limits
$115 a month
186th 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
Vercel ranks 15th and fits sandboxes for your users at scale, Northflank does not publish how many can run at once, and bochka ranks first.
15th cheapest of 361
$1,267 a month
89th of 361, with limits
$815 a month
Cheapest of 361
$209 a month
RL rollouts
1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month
Vercel ranks 3rd for many short parallel rollouts thanks to high concurrency and fast starts; Northflank's unpublished concurrency makes it a poor fit.
3rd cheapest of 360
$15,761 a month
342nd of 360, with limits
$1,762 a month
34th 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
69th cheapest of 364
$1,508 a month
134th of 364, with limits
$689 a month
10th cheapest of 364
$200 a month
CI runners
40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month
Northflank ranks 38th and Vercel 107th for CI; bochka ranks 4th and adds /dev/kvm for Android emulators and VM builds.
107th cheapest of 377
$1,144 a month
38th cheapest of 377
$352 a month
4th cheapest of 377
$86 a month
24/7 dev boxes
1 machine 24/7 · 4 vCPU / 8 GB
276th of 368, with limits
$145 a month
151st of 368, with limits
$59 a month
18th cheapest of 368
$26 a month
Computer-use agents
50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month
Both are headless, so a computer-use agent needs a desktop you build and stream yourself; bochka ships a 60fps desktop with a computer-use API.
235th of 369, with limits
$952 a month
151st of 369, with limits
$371 a month
6th cheapest of 369
$200 a month
Browser-use agents
50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month
259th of 367, with limits
$486 a month
200th of 367, with limits
$194 a month
7th 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
277th of 346, with limits
$468 a month
196th of 346, with limits
$117 a month
6th 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
234th of 338, with limits
$9,960 a month
83rd of 338, with limits
$2,049 a month
71st 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
283rd of 360, with limits
$554 a month
141st of 360, with limits
$99 a month
2nd cheapest of 360
$200 a month
GPU eval jobs
10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month
Northflank has GPU sandboxes from L4 to H100 and up; Vercel Sandbox and bochka are CPU only.
Can't run it4th cheapest of 98
$693 a month
Can't run it
macOS agents
5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month
Can't run itCan't run itCan't run it

Key features

FeatureVercel SandboxNorthflankbochka
IsolationFirecracker microVM with its own kernelKata microVM with its own kernel for CPU; gVisor for GPUFull VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts
Longest session24 hours on Pro and Enterprise, 45 minutes on Hobby; resume starts a new sessionNo documented cap; runs until you pause or delete itNo cap on paid plans (--no-auto-stop); default TTL 1 hour, changeable
What survives a stopFilesystem snapshot on every stop, kept 30 days after last use by default, billed per GB a monthOnly attached volumes, which bill while paused; the rest of the disk resetsThe whole disk, saved every minute and on stop; a stopped sandbox costs nothing
Memory on resumeNot kept; processes start overNot kept; processes are lost on pauseNot kept; the machine reboots with its disk
Desktop, browser, computer useNone; headless Linux onlyNone; access by SDK, browser terminal or SSH60fps desktop and Chromium, computer-use and browser-control APIs, embeddable stream
Nested virtualizationNoNot documented inside a sandboxYes, /dev/kvm for Android emulators and VMs
Docker insideYes, with sudoYes, own kernel per sandboxYes, with root and systemd
Coding agentsClaude Code, Codex, OpenCode and pi preinstalled; you start them as commandsHarnesses with Claude Code, Codex, Cursor, OpenCode or Pi; you reach the agent through a terminal or SSHSeven agents preinstalled, with an API to prompt, steer and stream events
Keys kept outside the sandboxCredentials brokering injects headers for any HTTPS domain you allowHeader injection for listed hosts in the network policy APICredentials proxy for model providers and any HTTPS API; switch a credential to another account when credits run out
Inbound accessPublic URLs for up to 15 ports, no built-in auth; no raw TCP or public IPHTTPS ports on a code.run domain, private by default; public TCP and UDP need a separate load balancerHTTPS URLs for up to 50 ports, token-protected by default; raw TCP; dedicated IPv4 or IPv6
Outbound rulesAllow-all, deny-all or a domain and CIDR allowlist, changeable while runningNetwork policies with allow lists by IP, CIDR or domain, documented for your own clusterOpen internet, no provider-side allowlist; firewall inside the VM
Biggest sandbox8 vCPU / 16 GB on Pro, 32 vCPU / 64 GB on EnterpriseUp to 20 vCPU / 40 GB, more on request; a resize restarts the sandboxUp to 16 vCPU / 32 GB on the $100 plan and above
GPUsNoL4, A100, H100, RTX PRO 6000 and moreNo
Regions and your own cloud19 regions; Vercel's cloud onlyUS, Europe, Asia, Australia, South America and Africa; or your own AWS, GCP or Azure accountEU only (Germany, Finland, France), data kept in the EU
Every feature Battleships tracks (43)
FeatureVercel SandboxNorthflankbochka
Full VM (own kernel)YesYesYes
Docker insideYesYesYes
Nested virtualizationNoNoYes
RootYesNot publishedYes
SystemdNot publishedNot publishedYes
Desktop GUINoNoYes
BrowserNoNoYes
Computer-use APINoNoYes
SSHNoYesYes
Persistent diskYesYesYes
Pause/resumeYesYesYes
Automatic snapshotsYesYesYes
Fork/cloneYesYesYes
Live fork (no pause)YesYesYes
Memory forkNoNoNo
Custom image (Docker or snapshot)YesYesYes
≥24 h sessionsYesYesYes
Idle auto-stopNoNoNo
Wake on requestNoNoNo
HTTPS preview URLsYesYesYes
Raw TCP inboundNoYesYes
Public IPv4NoNoNo
Egress allowlistYesYesNo
Secret proxyYesYesYes
Secret proxy for any APIYesYesYes
Preinstalled agentsYesYesYes
Agent harness APINoNoYes
MCP serverYesNoNo
GPUNoYesNo
WindowsNoNoNo
Arm64NoYesNo
Self-hostingNoYesNo
BYOCNoYesNo
Open sourceNoNoNo
Scoped API keysNot publishedNot publishedYes
Spend limitsYesNot publishedYes
Usage per sandbox (API)NoYesYes
Audit logsNot publishedYesNo
SSOYesYesNo
SOC 2YesYesNo
HIPAAYesYesNo
EU data residencyYesYesYes
Zero data retentionNot publishedNot publishedYes

Performance

Vercel Sandbox12.96 runs/s
NorthflankNot measured
bochka26.69 runs/s

On the hpc-sandbox-benchmarks Node.js build-and-test loop at 4 vCPU / 8 GB, bochka did 26.69 runs per second and Vercel Sandbox did 12.96, so bochka finished about twice as many runs in the same time. Northflank has no result on that benchmark, so there is no comparable CPU measurement for it.

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

ScenarioVercel Sandbox, a monthNorthflank, a monthbochka, 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
$57 · Pro$32 · Developer Sandbox (free tier) · with limits$20 · $20/mo
One agent that never stops
4 vCPU / 8 GB, 50 GB disk, running 24/7 all month
$236 · Pro · with limits$108 · Developer Sandbox (free tier) · with limits$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
$2,884 · Pro$1,322 · Developer Sandbox (free tier) · with limits$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
$828 · Pro$361 · Developer Sandbox (free tier) · with limits$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
$3,152 · Pro$688 · Developer Sandbox (free tier) · with limits$2,000 · $2000/mo · with limits

Vercel Sandbox: Vercel charges for CPU only while your code uses it, but charges for memory for the whole time the sandbox runs, plus a small fee per sandbox created and a monthly fee per GB of snapshots. The $20 Pro credit is shared with your other Vercel usage.

Northflank: Northflank bills per second for the plan size you pick, with no monthly fee, until you pause the sandbox; there is no idle auto-pause. Volumes cost $0.15 per GB a month even while paused, and outbound traffic costs $0.06 per GB.

Northflank: bochka charges a flat price per size, per second, only while the machine runs, and a stopped sandbox costs nothing. Plans start at $20 a month, and the whole fee is usage credit.

Limits in those scenarios

For one agent that never stops, Vercel ends each session at 24 hours and the processes start over. On Northflank the limit in every scenario is what survives a stop: only attached volumes keep their files, and it does not publish how many sandboxes can run at once, so 300 or 1,000 at a time is not confirmed. Vercel fits the burst of 1,000 short jobs at once, while bochka does not: even its $2,000 plan queues starts above 90 a minute.

ScenarioVercel SandboxNorthflankbochka
One coding agent, workdaysNone that bite
  • files survive a stop only in attached volumes: packages installed elsewhere reset
None that bite
One agent that never stops
  • Pro: sessions capped at 24 h
  • files survive a stop only in attached volumes: packages installed elsewhere reset
None that bite
50 agents in parallelNone that bite
  • files survive a stop only in attached volumes: packages installed elsewhere reset
  • how many machines can run at once: limit not published, probably fine at 50 (the typical published limit is 50)
None that bite
Sandboxes for your usersNone that bite
  • not confirmed: how many machines can run at once (you need 300; the typical published limit is 50)
None that bite
1,000 short jobs at onceNone that bite
  • not confirmed: how many machines can run at once (you need 1,000; the typical published limit is 50)
  • 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 Vercel Sandbox when

you run many short bursts of untrusted code, like code interpreters or RL rollouts, and want a firewall you can change while the sandbox runs.

  • Fast, short bursts. For code interpreters and RL rollouts: Vercel starts sandboxes fast, at high concurrency, and bills CPU only while code uses it.
  • A firewall you change live. A running sandbox can be switched to deny-all, or to a list of allowed domains, without a restart.
  • Images from your Dockerfile. If you already deploy on Vercel, you can build an image from a Dockerfile and boot sandboxes straight into it.

Use Northflank when

you need GPUs next to the agent, sandboxes that run with no session cap, or sandboxes inside your own cloud account next to the rest of your apps.

  • GPU sandboxes. For agents that train or run models: Northflank has GPU sandboxes from L4 to H100 and up.
  • No session cap. A Northflank sandbox runs until you pause it, so long jobs are not cut off after a set number of hours.
  • Your own cloud, one platform. For teams that must run in their own AWS, GCP or Azure account, next to their apps, jobs and databases.

Where bochka fits

Both Vercel Sandbox and Northflank give each sandbox its own kernel, but both are headless and neither keeps the whole machine the way an agent left it. If your agents work for hours on one machine, building, testing, running Docker and checking their work in a browser, bochka gives them a full VM with a desktop, a disk that survives every stop, and an API to drive the agents inside from outside.

The whole machine comes back

Vercel stops each session at 24 hours and restarts every process; Northflank wipes everything outside attached volumes on a pause, so agents reinstall their tools unless you plan volumes for each path. bochka saves the whole disk every minute, and stops are free.

A desktop and a browser inside

Both are headless, so an agent cannot open the web app it just built to check it unless you build an image with a desktop, a browser and a way to stream it. bochka ships a 60fps desktop, Chromium and computer-use APIs.

Agents you drive from outside

On Vercel you run the agents as plain commands, and on Northflank you reach them through a terminal or SSH, so you build the layer that prompts, steers and reads them. bochka gives you one API for that.

Use bochka when your agents work on one machine for hours or days and should find it as they left it: installs, Docker, emulators, a desktop and a browser, with the agents inside driven from your code. Compare it one to one: bochka vs Vercel Sandbox and bochka vs Northflank.

Questions people ask

Is Vercel Sandbox cheaper than Northflank?

In our five scenarios Northflank costs less: one workday agent is $32 a month against $57, and 50 parallel agents are $1,322 against $2,884. But on Northflank only attached volumes survive a stop, and running 300 or 1,000 at once is not confirmed, while Vercel fits those at $828 and $3,152. bochka costs $20 and $317 for the first two.

Which is faster?

Vercel Sandbox did 12.96 runs per second on the hpc-sandbox-benchmarks Node.js loop at 4 vCPU / 8 GB, and bochka did 26.69. Northflank has no result on that benchmark, so we cannot compare its CPU speed.

Can Vercel Sandbox or Northflank run 24/7?

Northflank can: a sandbox runs until you pause it, though a pause wipes files outside attached volumes. Vercel sessions stop after 24 hours, and its docs say sandboxes are not designed to run continuously. bochka machines can run with no cap on paid plans and keep their whole disk.

Can both run Docker?

Yes. Both give each sandbox its own kernel, and Vercel documents Docker with sudo. bochka runs Docker with root and systemd, and also exposes /dev/kvm for Android emulators and VMs, which neither documents.

Does either have GPUs?

Northflank does, with GPU sandboxes under gVisor from L4 to H100 and more. Vercel Sandbox and bochka run on CPUs only.

Do they have a desktop for computer use?

No. Both are headless Linux sandboxes you drive with commands, files and ports. bochka includes a 60fps desktop, Chromium, and computer-use and browser-control APIs.

Try bochka

Install the CLI, then run bochka new.

Product preview

More comparisons

Bochka product preview · Pricing, benchmarks and commands are illustrative. Sandbox provisioning is not connected.