bochka vs Mosaic Sandbox
Mosaic Sandbox runs headless Firecracker microVMs that hibernate after three quiet seconds and fork with their memory. bochka gives each agent a full VM with a desktop, Chromium, /dev/kvm and a disk that stays, hosted in the EU.
Where bochka stands out
A desktop and a browser in every sandbox
Every bochka sandbox has a 60fps desktop and Chromium, with computer use and browser control built in, and you can embed the stream in your own app. Mosaic sandboxes are headless, and its docs list computer use as not supported.
A full machine, not a task box
bochka machines run systemd, Docker and /dev/kvm on every template, so services start on boot and Android emulators or nested VMs work. Mosaic runs Docker only on its base template, maps systemd services to its own durable processes, and documents no /dev/kvm.
A disk that stays, in the EU
A bochka sandbox keeps its disk until you delete it, with snapshots every minute and on stop, and a stopped one costs nothing. It runs in Germany, Finland or France with EU data residency and a zero data retention option. Mosaic reaps a sandbox at its deadline unless you set persist: true, keeps an archive for 7 days, and has no EU region today.
The companies and products
bochka
bochka.xyz 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
Mosaic Sandbox
Mosaic Sandbox is an execution API for AI agents, run by Mosaic OS: you create Linux microVMs through REST, a CLI, SDKs, MCP or SSH.
Each sandbox is a Firecracker microVM on KVM with its own guest kernel. A quiet sandbox hibernates after three seconds and wakes on the next request, and you pay only for awake time. It is built for coding agents and repository tasks: clone, install, test, fork into branches, and run Claude Code, Codex or other agents from one prompt call.
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 | bochka: price rank | Mosaic Sandbox: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Both run Docker in a VM; bochka adds /dev/kvm for emulators, systemd and a disk that keeps the project between sessions, while Mosaic reaps a sandbox at its deadline unless you set persist: true. | 3rd cheapest of 367 $634 a month | 137th of 367, with limits $1,206 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB An assistant that lives for weeks fits bochka's always-there disk; on Mosaic an archived sandbox is kept for 7 days unresumed. | 5th cheapest of 366 $657 a month | 181st of 366, with limits $1,570 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Mosaic fits short runs well: a one-shot run creates, executes and destroys a sandbox in about a second and bills only awake seconds. | 186th of 345, with limits $2,000 a month | 17th cheapest of 345 $164 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month 300 sandboxes at once is more than Mosaic's published cap of 200 per organization, so it needs a raised limit there. | Cheapest of 361 $209 a month | 212th of 361, with limits $515 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 | 343rd of 360, with limits $2,486 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 | 254th of 364, with limits $400 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month | 4th cheapest of 377 $86 a month | 17th cheapest of 377 $240 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB | 18th cheapest of 368 $26 a month | 48th cheapest of 368 $63 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Only bochka ships a desktop, Chromium and computer use; Mosaic lists computer use as not supported. | 6th cheapest of 369 $200 a month | 145th of 369, with limits $360 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month bochka has Chromium and browser control on every machine; Mosaic has no browser template or browser API. | 7th cheapest of 367 $200 a month | 197th of 367, with limits $180 a month |
| Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month bochka gives each machine a dedicated IPv6 or IPv4 and a real browser; Mosaic sandboxes are headless and get no public IP of their own. | 6th cheapest of 346 $90 a month | 242nd of 346, with limits $244 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 | 156th of 338, with limits $4,875 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month Both isolate with a VM; bochka adds a desktop and raw TCP for the tools a red team runs, while Mosaic adds tight egress allowlists. | 2nd cheapest of 360 $200 a month | 150th of 360, with limits $119 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
Every feature Battleships tracks (43)
| Feature | bochka | Mosaic Sandbox |
|---|---|---|
| Full VM (own kernel) | Yes | Yes |
| Docker inside | Yes | Yes |
| Nested virtualization | Yes | No |
| Root | Yes | Yes |
| Systemd | Yes | Not published |
| Desktop GUI | Yes | No |
| Browser | Yes | No |
| Computer-use API | Yes | No |
| SSH | Yes | Yes |
| Persistent disk | Yes | Yes |
| Pause/resume | Yes | Yes |
| Automatic snapshots | Yes | Yes |
| 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 | Yes |
| 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 | Yes |
| Preinstalled agents | Yes | Yes |
| Agent harness API | Yes | Yes |
| 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 | Yes |
| Spend limits | Yes | No |
| 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
There is no comparable CPU measurement for Mosaic Sandbox: it was not part of the StarSling hpc-sandbox-benchmarks run where bochka did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB, so we make no speed claim either way.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Mosaic Sandbox 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 | bochka, a month | Mosaic Sandbox, 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 | $16 · Metered (checkout required) |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $66 · Metered (checkout required) |
| 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 | $792 · Metered (checkout required) |
| 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 | $225 · Metered (checkout required) · 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 | $497 · Metered (checkout required) · with limits |
bochka: You pick a monthly plan from $20, and all of it becomes usage credit spent per second on running sandboxes. A stopped sandbox costs nothing and keeps its disk.
Mosaic Sandbox: Mosaic bills $0.05 per awake hour for each 2 vCPU / 4 GB unit, per second, with RAM, storage and egress included; a 4 vCPU sandbox bills as two units. Hibernated time is free, but a running process or an open SSH session keeps the sandbox awake, and you need a card at checkout before you start.
Limits in those scenarios
Mosaic allows up to 200 sandboxes alive per organization, and hibernated ones count, so 300 sandboxes for your users and 1,000 short jobs at once both need a raised limit; 50 agents in parallel fit. Its default shape is 2 vCPU / 4 GB and only that shape is kept warm, so the 4 vCPU agents in these scenarios cold boot. On bochka the limit is a sandbox count (100 to 2,000 by plan) plus starts per minute, so the 1,000-job burst gets queued and needs more starts per day than the $2,000 plan allows.
| Scenario | bochka | Mosaic Sandbox |
|---|---|---|
| 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 | None that bite |
| Sandboxes for your users | None that bite |
|
| 1,000 short jobs at once |
|
|
Which one should you use?
Use Mosaic Sandbox when
your agents run short, bursty repository tasks in the US, wait a lot between steps, and need tight egress rules or memory forks.
- Hibernate in seconds, fork with memory. Useful if your agents wait a lot: Mosaic pauses a quiet sandbox after three seconds and wakes it on the next request, memory intact.
- Fine-grained network rules. Useful for running untrusted code: each sandbox can be airgapped or limited to chosen hosts, and the rules can change while it runs.
- Dockerfile images and US regions. Useful if you start from your own Dockerfile or need machines in the US: Mosaic builds images and runs in Silicon Valley and Virginia.
Use bochka when
your agents need a whole computer that stays: a desktop and browser, Docker, systemd and /dev/kvm on every machine, a disk that keeps their work, and EU hosting.
- A desktop and a browser in every sandbox. An agent that has to click through a web app, log in to a site or test a UI needs a screen and a browser. On Mosaic you would install and stream your own desktop inside the microVM, and keep it working yourself.
- A full machine, not a task box. Agents that build images, run emulators or start a database as a service need the parts a normal Linux server has. On Mosaic they must pick the base template and still go without KVM.
- A disk that stays, in the EU. An assistant that comes back next month expects its repo and tools where it left them. On Mosaic you set persist: true and resume within 7 days, or copy work to /mnt/data, and EU-resident data has nowhere to go.
Questions people ask
Is bochka cheaper than Mosaic Sandbox?
It depends on how long the sandbox is awake. One coding agent on workdays comes to $20 a month on bochka and $16 on Mosaic, but one agent running all month comes to $26 versus $66, and 50 agents in parallel come to $317 versus $792. For 1,000 short jobs at once Mosaic comes to $497 and bochka to $2,000, and neither fits without higher limits.
Does Mosaic Sandbox have a desktop or computer use?
No. Mosaic sandboxes are headless, and its docs list computer use as not supported. bochka streams a 60fps desktop with Chromium on every machine, with computer use and browser control built in.
Do Mosaic sandboxes keep their files?
Through hibernation and pause, yes, with memory too. At its inactivity deadline a sandbox is destroyed unless you created it with persist: true, which archives it for 7 days, and shared /mnt/data volumes outlive sandboxes. On bochka the disk stays until you delete the sandbox, with snapshots every minute.
Can I run Docker and emulators in Mosaic Sandbox?
Docker runs on Mosaic's base template; the node and python templates list it as unavailable. Mosaic documents no /dev/kvm, so Android emulators and nested VMs are not an option there. bochka has Docker, systemd and /dev/kvm on every machine.
Where does my data run?
bochka runs only in the EU (Germany, Finland and France), with EU data residency and a zero data retention option. Mosaic runs in Silicon Valley and Northern Virginia and says it has no EU region today.
Is bochka SOC 2 compliant?
SOC 2 is in progress. We're open about where we stand: live status at status.bochka.xyz and our data retention policy at docs.bochka.xyz/data-retention. Mosaic's roadmap also lists SOC 2 as in progress.
Start a sandbox
Install the CLI, then run bochka new.
Product preview