bochka vs Alibaba Cloud ACK
Alibaba Cloud ACK is managed Kubernetes: you run a cluster, and each sandbox is a pod you schedule on it, often through an eval tool like Harbor. bochka hands each agent its own VM with Docker, a desktop and coding agents built in, with nothing to operate.
Where bochka stands out
A full VM, not a pod
On bochka every sandbox has its own kernel, with root, systemd, Docker and /dev/kvm ready. On ACK a sandbox is a pod: on ECS nodes it shares the node's kernel, and the runtime is containerd, so Docker inside needs a privileged pod. Neither kind of pod has a documented /dev/kvm.
Disk that keeps itself, fork that copies a live machine
bochka snapshots each sandbox's disk every minute and on stop, a stopped sandbox costs nothing, and you can fork a running sandbox into a copy. On ACK a pod's files survive only on a cloud-disk volume you attach, snapshots are made when you ask, and there is no pause or fork.
Coding agents and keys handled for you
bochka ships Claude Code, Codex, OpenCode and other agents inside each sandbox, and you drive them from outside with a prompt, steer and events API. The credentials proxy keeps real AI keys out of the sandbox and can switch to another account when one runs out of credits. On ACK you bake the agent into an image and pass keys in as Kubernetes Secrets the pod can read.
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
Alibaba Cloud ACK
Alibaba Cloud ACK (Container Service for Kubernetes) is the managed Kubernetes service of Alibaba Cloud, a hyperscaler. It is not a sandbox product.
You create a Kubernetes cluster in your Alibaba Cloud account, then run pods on ECS worker nodes you pay for, or on serverless ECI virtual nodes billed per second for the vCPU and memory you request. Tools such as the Harbor eval framework use it as a backend: each task becomes a pod built from your container image. It is built for teams that already run containers on Kubernetes and want to manage their own fleet.
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 | Alibaba Cloud ACK: price rank |
|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month A software factory needs Docker builds and long sessions; bochka has Docker and root on every VM, while ACK pods run on containerd and Docker inside needs privileged pods you set up yourself. | 3rd cheapest of 367 $634 a month | 207th of 367, with limits $2,438 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB Both can run all month, but on ACK each assistant is a pod in a cluster you keep running, and you add NAT and an elastic IP before it reaches the internet. | 5th cheapest of 366 $657 a month | 220th of 366, with limits $2,528 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Short bursts suit ECI well: pods bill per second from Pending to release with no plan fee, while bochka's start limits make big bursts a compromise. | 186th of 345, with limits $2,000 a month | 344th of 345, with limits $115 a month |
| Sandboxes for your users 300 at once · 2 vCPU / 4 GB · 30,000 sessions of 23 min a month Giving each of your users a sandbox on ACK means running a Pro cluster and building the isolation, routing and cleanup yourself; bochka gives each user a VM with HTTPS URLs out of the box. | Cheapest of 361 $209 a month | 356th of 361, with limits $805 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 | 346th of 360, with limits $3,463 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 | 353rd of 364, with limits $693 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 | 376th of 377, with limits $332 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB A dev box on ACK is a pod with no pause or fork; bochka stops for free, keeps the disk and resumes when you call it. | 18th cheapest of 368 $26 a month | 257th of 368, with limits $101 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month ACK ships no desktop, browser or computer-use API, so you would build and stream one yourself; bochka has a 60fps desktop and Chromium in every VM. | 6th cheapest of 369 $200 a month | 207th of 369, with limits $693 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 | 236th of 367, with limits $346 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 | 215th of 346, with limits $173 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 | 119th of 338, with limits $3,463 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month ECI pods are cheap for parallel attack runs, but they reach the internet only through a NAT gateway you add, and a running pod cannot be forked. | 2nd cheapest of 360 $200 a month | 186th of 360, with limits $173 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 | bochka | Alibaba Cloud ACK |
|---|---|---|
| What you get | A ready sandbox VM per call; nothing to operate | A Kubernetes cluster you run; each sandbox is a pod you schedule |
| Isolation | Every sandbox is a full VM with its own kernel | ECI pods each run in their own CIPU-based sandbox with VM-class isolation; pods on ECS nodes share the node kernel |
| Docker and nested virtualization | Docker, root, systemd and /dev/kvm on every machine | containerd runtime; Docker inside needs privileged pods; no documented /dev/kvm for pods |
| Files after a stop | Snapshots every minute and on stop; a stopped sandbox costs nothing | Cloud-disk volumes survive pod restarts; volume snapshots are made on request, not automatically |
| Pause and fork | Stop and resume from disk; fork a running sandbox | No pause or fork of a pod; you scale to zero or delete and recreate it |
| Custom images | Named snapshots (templates), shareable by code; no Dockerfile builds | Any OCI image from Container Registry or another registry |
| Desktop and computer use | 60fps desktop and Chromium, computer use and browser control, embeddable stream | None built in |
| Coding agents | Claude Code, Codex, OpenCode and more preinstalled, with a prompt, steer and events API | None built in; you bake an agent into your image and run it with kubectl exec |
| Keeping keys out of the sandbox | AI credentials proxy for model providers and any HTTPS API you register; switch a credential to another account when credits run out | Secrets reach the pod as Kubernetes Secrets, synced from KMS |
| Outbound internet | Open internet on every plan; no provider-side allowlist | None until you add a NAT gateway and elastic IP; then IP and port network policies and a static egress IP |
| Reaching the sandbox | HTTPS URLs for up to 50 ports, token-protected; raw TCP; dedicated IPv6 or IPv4 | Load balancers, Ingress and elastic IPs you set up per service; SSH to ECS nodes |
| Wake on request | No; you start a stopped sandbox through the API | Knative on ACK scales a service to zero and back up when a request arrives |
| GPUs, Windows and Arm | None; Linux x86 VMs only, with Wine for Windows apps | GPU instances, Windows Server node pools and Arm64 nodes |
| Regions | EU only: Germany, Finland, France | About 30 regions across Asia, the Middle East, the Americas and Europe |
| Compliance and transparency | SOC 2 is in progress. We're open about where we stand: live status and data retention policy. | SOC 2 |
Every feature Battleships tracks (43)
| Feature | bochka | Alibaba Cloud ACK |
|---|---|---|
| Full VM (own kernel) | Yes | No |
| Docker inside | Yes | No |
| Nested virtualization | Yes | No |
| Root | Yes | Not published |
| 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 | 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 | No |
| Wake on request | No | Yes |
| HTTPS preview URLs | Yes | Yes |
| Raw TCP inbound | Yes | Not published |
| Public IPv4 | No | Yes |
| Egress allowlist | No | Yes |
| Secret proxy | Yes | No |
| Secret proxy for any API | Yes | No |
| Preinstalled agents | Yes | No |
| Agent harness API | Yes | No |
| MCP server | No | Yes |
| GPU | No | Yes |
| Windows | No | Yes |
| Arm64 | No | Yes |
| Self-hosting | No | No |
| BYOC | No | Yes |
| Open source | No | No |
| Scoped API keys | Yes | Not published |
| Spend limits | Yes | Not published |
| Usage per sandbox (API) | Yes | Yes |
| Audit logs | No | Not published |
| SSO | No | Yes |
| SOC 2 | No | Yes |
| HIPAA | No | Yes |
| EU data residency | Yes | Yes |
| Zero data retention | Yes | Not published |
Performance
There is no comparable measurement: the public sandbox benchmark we use did not test ACK pods, and an ACK pod's speed depends on which ECS or ECI instance type you choose. bochka did 26.69 Node.js build-and-test runs per second at 4 vCPU / 8 GB on dedicated AMD Ryzen 9 9950X (Zen 5) hosts.
Node.js build-and-test loop at 4 vCPU / 8 GB, higher is better. Measured with the open hpc-sandbox-benchmarks suite. Alibaba Cloud ACK 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 | Alibaba Cloud ACK, 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 | $24 · ACK managed Basic · with limits |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $26 · $20/mo | $101 · ACK managed Basic · 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 | $1,219 · ACK managed Basic · 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 | $346 · ACK managed Basic · 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 | $693 · ACK managed Basic · 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.
Alibaba Cloud ACK: On ECI virtual nodes you pay per second for the vCPU and memory a pod requests, from Pending until it is released. On top come the cluster fee (Pro, per cluster per hour, even with no pods), the NAT gateway and elastic IP for internet, disks beyond the free 30 GiB and egress, and the scenario prices above leave out the ones with no published price.
Limits in those scenarios
None of the ACK scenarios count as a full fit, because pods have no internet until you add a NAT gateway and elastic IP, and Alibaba does not publish a flat egress price. The prices assume the free Basic edition, which Alibaba says is for personal testing or learning only and caps a cluster at 10 nodes and 1,000 serverless pods; production use means Pro and its hourly cluster fee. 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. ECI vCPU quotas also vary by account history, and raising them takes a ticket.
| Scenario | bochka | Alibaba Cloud ACK |
|---|---|---|
| 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 Alibaba Cloud ACK when
you already run Kubernetes on Alibaba Cloud, need GPUs, Windows or regions in Asia, or want short ECI bursts billed per second inside your own account.
- GPUs, Windows and Arm nodes. Useful if your agents need GPUs, Windows Server or Arm64: ACK node pools offer all three.
- Regions around the world. Useful if your data or users are in Asia, the Middle East or the Americas: ACK runs in about 30 regions.
- Your own Kubernetes, with its controls. Useful if your team already runs Kubernetes and wants network policies, a static egress IP and scale to zero in its own cloud account.
Use bochka when
you want each agent on a ready VM with Docker, /dev/kvm, a desktop, coding agents and a disk that saves itself, without building and running a cluster.
- A full VM, not a pod. Agents install packages, run docker compose and start emulators. On ACK you get Docker only through privileged pods, which on ECS nodes hand the agent's code the node, and emulators have no /dev/kvm.
- Disk that keeps itself, fork that copies a live machine. When an agent's pod is evicted or deleted, work outside an attached volume is gone. To try two paths from one state on ACK you write the snapshot, restore and reschedule steps yourself.
- Coding agents and keys handled for you. A key in a Secret is readable by any code the agent runs in that pod. Without a driving API you also write and maintain the code that starts, prompts and watches each agent.
Questions people ask
Is bochka cheaper than Alibaba Cloud ACK?
In most of the scenarios we priced, yes: one agent running all month comes to $26 on bochka against $101 of ECI compute, and 50 agents in parallel come to $317 against $1,219. For 1,000 short jobs at once ACK is cheaper, $693 against $2,000. The ACK figures leave out the NAT gateway, elastic IP and egress, which have no published flat price.
Is ACK a sandbox for AI agents?
No. ACK is managed Kubernetes; tools like Harbor use it as a backend by starting one pod per task in a cluster you run. Alibaba sells sandboxes for agents as separate products. bochka is built as a sandbox: each call gives you a VM ready to use.
Can I run Docker inside an ACK pod?
ACK nodes use containerd, and running a Docker daemon inside a pod needs a privileged pod that Alibaba does not document as a supported setup. On bochka every VM has Docker, root, systemd and /dev/kvm.
Can ACK pods reach the internet?
Only after you create an Internet NAT gateway, attach an elastic IP and add SNAT entries for the cluster's vSwitches, all billed separately. bochka sandboxes have open internet on every plan.
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. ACK runs in about 30 regions, including Frankfurt, London and Paris in Europe.
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. Alibaba Cloud's trust center lists a SOC 2 report for its services.
Start a sandbox
Install the CLI, then run bochka new.
Product preview