Google Agent Runtime vs Azure Container Apps Sandboxes
Google Agent Runtime hosts the agent code you build with ADK or LangGraph. Its Agent Sandbox runs tools in gVisor-isolated containers where commands have no root. Azure Container Apps Sandboxes gives each agent a microVM with its own kernel and root, which can suspend with its memory and wake on traffic.
The companies and products
Google Agent Runtime
Google Agent Runtime is the agent hosting part of Gemini Enterprise Agent Platform on Google Cloud, renamed from Vertex AI Agent Engine in April 2026.
It is a hyperscaler service, not a sandbox company. Agent Runtime deploys agents you build with ADK, LangGraph, LangChain, AG2 or LlamaIndex and bills them only while they work on a turn. Next to it, Agent Sandbox gives those agents a place to run code: a Code Execution sandbox, a shell or custom container sandbox, and a Computer Use browser. Each sandbox runs in a gVisor-isolated container in a Google-managed tenant project. Teams on GKE can also run the open-source GKE Agent Sandbox in their own cluster, with gVisor or Kata Containers, and manage it themselves.
Azure Container Apps Sandboxes
Azure Container Apps Sandboxes is a service inside Microsoft Azure that gives each agent or user an isolated microVM you create, suspend and resume through an SDK.
It is part of Azure Container Apps, the hyperscaler's serverless container platform, and lives in your Azure subscription as a sandbox group resource. Each sandbox boots from an OCI image or a snapshot and has its own kernel. A stop keeps the disk, or the disk and the memory, and a memory-mode sandbox resumes in under a second. It is built for teams already on Azure that want stateful code execution with Azure networking, identity and billing.
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. “34th cheapest of 361” means 33 of the other 360 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 | Google Agent Runtime: price rank | Azure Container Apps Sandboxes: price rank | boat: price rank |
|---|---|---|---|
| Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Google's sandboxes run without root and document no Docker. Azure runs Docker on its azure-dev image, but documents no nested virtualization or SSH. boat gives each agent root, Docker, /dev/kvm and SSH on a full VM. | 310th of 365, with limits $5,801 a month | 33rd cheapest of 365 $5,276 a month | 5th cheapest of 365 $634 a month |
| Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB Google fits this one: a sandbox lives up to 7 days, and a snapshot (Preview) carries it into a new sandbox. Azure's always-on price needs a 3-year commitment. boat also gives the assistant a desktop, a browser and SSH, with auto-stop turned off (default 1 hour). | 34th cheapest of 361 $7,519 a month | 52nd cheapest of 361 $6,780 a month | 6th cheapest of 361 $657 a month |
| Code interpreter bursts 100 at once · 1 vCPU / 2 GB · 200,000 sessions of 1 min a month Azure suits short code runs: its sandboxes suspend with their memory and resume in under a second. Google's Code Execution sandbox is a ready interpreter. On boat, each create counts against the plan's start limits. | 251st of 336, with limits $275 a month | 39th cheapest of 336 $354 a month | 189th of 336, 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 Azure can publish public HTTPS routes per port for each user sandbox, though 300 at once needs a quota raise. Google reaches a sandbox through its API or a private endpoint in your VPC. boat serves token-gated HTTPS on up to 50 ports per sandbox. | 275th of 349, with limits $1,916 a month | 32nd cheapest of 349 $2,260 a month | 3rd cheapest of 349 $209 a month |
| RL rollouts 1,000 at once · 2 vCPU / 4 GB · 1,000,000 sessions of 3 min a month | 132nd of 349, with limits $8,240 a month | 158th of 349, with limits $10,740 a month | 39th of 349, with limits $2,000 a month |
| Agent evals 250 at once · 4 vCPU / 8 GB · 15,000 sessions of 20 min a month | 90th cheapest of 359 $2,060 a month | 85th cheapest of 359 $1,728 a month | 10th cheapest of 359 $200 a month |
| CI runners 40 at once · 8 vCPU / 16 GB · 6,000 sessions of 12 min a month Jobs that build containers need Docker. Google documents none inside its sandboxes, and Azure has it in images like azure-dev. boat runs Docker and /dev/kvm on every VM, within its per-plan start limits. | 104th cheapest of 374 $989 a month | 108th cheapest of 374 $1,046 a month | 2nd cheapest of 374 $86 a month |
| 24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB Neither documents SSH: Google runs commands through its API, and Azure gives a shell through its own CLI, SDK and portal. boat gives you SSH and a desktop. | 317th of 371, with limits $241 a month | 327th of 371, with limits $262 a month | 19th cheapest of 371 $26 a month |
| Computer-use agents 50 at once · 4 vCPU / 8 GB · 20,000 sessions of 15 min a month Google gives a browser with a VNC live view and a click and type API. Azure's public images have no desktop; you build one into your own image and publish it on an HTTPS port. boat streams a full Linux desktop at 60fps. | 85th of 367, with limits $1,648 a month | 58th cheapest of 367 $1,573 a month | 7th cheapest of 367 $200 a month |
| Browser-use agents 50 at once · 2 vCPU / 4 GB · 30,000 sessions of 10 min a month Google ships a Chromium you can drive over CDP with Playwright. On Azure you build a browser into your own image. boat ships Chrome on a desktop in every sandbox. | 132nd of 356, with limits $824 a month | 129th of 356, with limits $795 a month | 7th cheapest of 356 $200 a month |
| Scraping & bots (startup scale) 50 at once · 1 vCPU / 2 GB · 100,000 sessions of 3 min a month | 56th cheapest of 343 $515 a month | 112th cheapest of 343 $561 a month | 6th cheapest of 343 $90 a month |
| Mass scraping & bots (undetectable) 1,000 at once · 1 vCPU / 2 GB · 3,000,000 sessions of 2 min a month | 212th of 327, with limits $8,240 a month | 240th of 327, with limits $10,956 a month | 73rd of 327, with limits $1,800 a month |
| Parallel red teaming 200 at once · 2 vCPU / 4 GB · 5,000 sessions of 30 min a month | 281st of 351, with limits $412 a month | 30th cheapest of 351 $517 a month | 3rd cheapest of 351 $200 a month |
| GPU eval jobs 10 at once · 8 vCPU / 32 GB · 500 sessions of 30 min a month | Not ranked: misses a requirement | Not ranked: misses a requirement | Not ranked: misses a requirement |
| macOS agents 5 at once · 4 vCPU / 8 GB · 1,000 sessions of 20 min a month | Not ranked: misses a requirement | Not ranked: misses a requirement | Not ranked: misses a requirement |
Key features
| Feature | Google Agent Runtime | Azure Container Apps Sandboxes | boat |
|---|---|---|---|
| Isolation | gVisor-isolated container in a Google-managed tenant project | MicroVM per sandbox with its own kernel | Full VM with its own kernel, on hosts dedicated to boat (AMD Ryzen 7 9700X and Ryzen 9 9950X); vCPUs are shared |
| Root, Docker and /dev/kvm | Runs as appuser with no sudo; images must not need root; Docker and /dev/kvm inside are not documented | Root; Docker depends on the image (azure-dev ships it on systemd, the default ubuntu image has only Buildah); nested virtualization not documented | Root, systemd, Docker and /dev/kvm on every sandbox |
| Getting a shell | Commands run through API calls; SSH is not documented | Interactive shell through the aca CLI, SDK or portal; SSH is not documented | SSH, plus CLI and API exec |
| Desktop, browser and computer use | A browser in a container with a VNC live view, a click, type and screenshot API, and CDP | No desktop or browser in the public images; you can build one into your own OCI image and publish it on an HTTPS port | Linux desktop and Chrome streamed at 60fps, with computer tools for agents; no GPU |
| Inbound access | Through Google's API or a private endpoint in your VPC; a public URL is not documented | Public HTTPS routes per published port; private endpoint from your VNet at extra cost; no public IP | Up to 50 HTTPS URLs, token-gated by default; SSH; raw ports over IPv6 only on sandboxes with no IPv4 of their own |
| Memory snapshots and forks | Snapshots (Preview) keep memory and files; start many sandboxes from one snapshot (same image and architecture) | Suspend with memory and resume in under a second; snapshot a running sandbox with its memory to seed new ones | Fork a running sandbox from its last 1-minute snapshot; memory is not saved, so a resume is like a reboot |
| Idle handling | Sandboxes are billed while they exist; a paused one keeps its files at a lower, unpublished rate | Auto-suspend after an idle timeout (5 minutes by default); traffic or a start command wakes it | Auto-stop defaults to 1 hour; disk saved every minute and at each stop; no compute charge while stopped, but extra saved snapshots and the AI credentials proxy still bill |
| How long a sandbox lives | TTL of up to 7 days for a custom container and 14 days for Code Execution; a snapshot (Preview) restores it into a new sandbox | No maximum runtime documented | No session cap once auto-stop is off on a paid plan |
| Outbound internet | Off unless the template turns it on; route through your VPC with Cloud NAT, a static IP, Cloud NGFW or Secure Web Proxy | Egress policy with allow, deny, rewrite and header rules by host, path and method; VNet integration | Open internet on every plan; no provider-side allowlist |
| Coding agents | Managed Agents API (Preview) runs Google's Antigravity harness | Copilot CLI and Claude Code images, run through the exec API | Seven harnesses installed, including Claude Code and Codex, with a prompt, steer and events API |
| Keeping AI keys out of the sandbox | You provision and scope the credentials the agent uses | You set it up: an egress rule adds a group secret or managed-identity token as a header, outside the sandbox; the built-in Copilot and Claude connections put tokens inside | Optional paid proxy ($1 a day plus $1 per 1,000 requests) for model providers and HTTPS APIs you register; other secrets still enter the sandbox |
| Hosting your agent code | Agent Runtime deploys ADK, LangGraph, LangChain, AG2 and LlamaIndex agents; idle time between turns is not billed | No separate agent hosting; you run your code inside sandboxes or other Container Apps | You run your own agent loop inside a sandbox |
| Scaling up a fleet | Default quota of 500 sandbox creates a minute per project and region | 300 concurrent active cores per region and subscription by default; more by Azure support request | 100 to 2,000 sandboxes at once by plan, self-serve; on the $2,000 plan, starts above 90 a minute cost extra and above 270 a minute fail |
| Regions | 23 locations across the US, Europe, the Middle East, Asia, Canada, Australia and Brazil | 40 Azure regions, 9 of them in the EU | Germany, Finland and France; you cannot pick a region; no US or Asia region |
Every feature Battleships tracks (43)
| Feature | Google Agent Runtime | Azure Container Apps Sandboxes | boat |
|---|---|---|---|
| Full VM (own kernel) | No documented way | Yes | Yes |
| Docker inside | No documented way | Yes | Yes |
| Nested virtualization | No documented way | Not published | Yes |
| Root | No documented way | Yes | Yes |
| Systemd | Not published | Yes | Yes |
| Desktop GUI | Not published | With your own setup | Yes |
| Browser | Yes | With your own setup | Yes |
| Computer-use API | Yes | No documented way | Yes |
| SSH | No documented way | No documented way | Yes |
| Persistent disk | Yes | Yes | Yes |
| Pause/resume | Yes | Yes | Yes |
| Automatic snapshots | Yes | Yes | Yes |
| Fork/clone | Yes | Yes | Yes |
| Live fork (no pause) | Yes | Yes | Yes |
| Memory fork | Yes | Yes | No documented way |
| Custom image (Docker or snapshot) | Yes | Yes | Yes |
| ≥24 h sessions | Yes | Yes | Yes |
| Idle auto-stop | No documented way | Yes | No documented way |
| Wake on request | No documented way | Yes | No documented way |
| HTTPS preview URLs | No documented way | Yes | Yes |
| Raw TCP inbound | Not published | No documented way | Yes |
| Public IPv4 | No documented way | No documented way | No documented way |
| Egress allowlist | Yes | Yes | No documented way |
| Secret proxy | No documented way | Yes | Yes |
| Secret proxy for any API | No documented way | Yes | Yes |
| Preinstalled agents | Yes | Yes | Yes |
| Agent harness API | Yes | No documented way | Yes |
| MCP server | No documented way | No documented way | No documented way |
| GPU | No documented way | No documented way | No documented way |
| Windows | No documented way | No documented way | No documented way |
| Arm64 | Yes | No documented way | No documented way |
| Self-hosting | No documented way | No documented way | No documented way |
| BYOC | No documented way | No documented way | No documented way |
| Open source | No documented way | No documented way | No documented way |
| Scoped API keys | Not published | Not published | Yes |
| Spend limits | Not published | Yes | Yes |
| Usage per sandbox (API) | Yes | Yes | Yes |
| Audit logs | Not published | Yes | No documented way |
| SSO | Yes | Yes | No documented way |
| SOC 2 | Yes | Yes | No documented way |
| HIPAA | Yes | Yes | No documented way |
| EU data residency | Yes | Yes | No documented way |
| Zero data retention | Not published | Not published | Yes |
Performance
Neither Google Agent Runtime nor Azure Container Apps Sandboxes is in the public benchmarks we use, so we cannot say which is faster. For reference, boat did 28.06 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB, second of 15 in StarSling's run of 3 October 2026 (run 37073680030). Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure.
V8 Web Tooling Benchmark at a 4 vCPU / 8 GB target, runs per second, higher is better. Source: StarSling's run of 3 October 2026 (run 37073680030), median of 5 sandboxes. Providers ran on different hardware in different regions. StarSling gives a shared rank when the run cannot separate two providers. Every boat sandbox in this run ran on an AMD Ryzen 9 9950X host. boat also places some sandboxes on AMD Ryzen 7 9700X hosts, which this run did not measure. Google Agent Runtime and Azure Container Apps Sandboxes are not in the public benchmarks we use, so we cannot say which is faster.
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. Start limits, per-start fees and plan minimums are in the price, boat's included.
| Scenario | Google Agent Runtime, a month | Azure Container Apps Sandboxes, a month | boat, 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 | $73 · Pay as you go | $68 · Pay-as-you-go | $20 · $20/mo |
| One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month | $301 · Pay as you go | $262 · Pay-as-you-go · 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 | $3,626 · Pay as you go · with limits | $3,421 · Pay-as-you-go | $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 | $1,030 · Pay as you go | $989 · Pay-as-you-go | $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 | $2,060 · Pay as you go | $2,148 · Pay-as-you-go | $8,800 · $2000/mo · with limits |
Google Agent Runtime: Google bills $0.085 per vCPU-hour and $0.009 per GiB-hour allocated, rounded to the second. The pricing page lists a monthly free amount but does not say it covers sandboxes, so our totals do not subtract it. Runtime skips idle time between turns, but sandboxes are billed while they exist. A paused one costs less at an unpublished rate, and disk, snapshot and egress rates are not published.
Azure Container Apps Sandboxes: Azure bills per second on the Container Apps Consumption meters for the vCPU and memory of the sandbox's tier while it runs, idle time before auto-suspend included. Stopped sandboxes pay no CPU or memory, and there is no plan fee. Some regions such as West Europe and UK South charge more, and storage for images, snapshots and volumes is announced as billed per GiB later.
Azure Container Apps Sandboxes: Plans cost $20, $100, $500 or $2,000 a month. The whole plan price comes back as VM time for that month, and it does not carry over. A 4 vCPU / 8 GB sandbox costs $0.036 an hour, billed per second while it runs, busy or idle. The $20 plan includes 200 starts a day; extra starts cost $0.03 to $0.07. A stopped sandbox has no compute charge; saved snapshots above the plan's count and the AI credentials proxy still charge. A card is required, also for the 7-day trial.
Limits in those scenarios
Each of the two fits four of the five scenarios. Google does not fit the 50-agent fleet, because it documents no Docker inside its sandboxes, and every Google row leaves disk and egress unpriced. A Google sandbox is deleted when its TTL ends, 7 days at most for a custom container; a snapshot (Preview) carries it into a new sandbox. Azure's always-on price ($262 a month) needs a 3-year savings plan, and Azure caps active cores at 300 per region and subscription by default, so 300 user sandboxes and 1,000 short jobs need a support request first. boat fits all but the burst: the $2,000 plan includes 90 starts a minute, charges for extra starts and refuses starts above 270 a minute. boat's always-on agent needs auto-stop turned off (default 1 hour).
| Scenario | Google Agent Runtime | Azure Container Apps Sandboxes | boat |
|---|---|---|---|
| One coding agent, workdays |
| None for this scenario | None for this scenario |
| One agent that never stops |
|
| None for this scenario |
| 50 agents in parallel |
| None for this scenario | None for this scenario |
| Sandboxes for your users |
| None for this scenario | None for this scenario |
| 1,000 short jobs at once |
| None for this scenario |
|
Which one should you use?
Use Google Agent Runtime when
you build agents with ADK or LangGraph on Google Cloud and want them hosted in your own project, with short code runs and a browser in a gVisor sandbox behind your network rules.
- A managed home for your agent code. If you build agents with ADK or LangGraph, Agent Runtime hosts them and does not bill idle time between turns.
- A browser agents can drive. Its Computer Use sandbox gives agents a browser with a VNC live view, a click and type API, and CDP for Playwright.
- Egress locked down by default. Outbound internet is off unless a template turns it on, and it can leave through your VPC with a static IP.
Use Azure Container Apps Sandboxes when
your team already runs on Azure and wants microVMs from its own images with root, that suspend with their memory, wake on traffic and sit behind fine-grained egress rules.
- A microVM with root. Each sandbox has its own kernel and root, and images like azure-dev ship Docker on systemd.
- Memory suspend and wake on traffic. Idle sandboxes suspend with their memory and resume in under a second when traffic or a start command arrives.
- Your own image and public ports. A sandbox boots from any OCI image you import, runs until you stop it or it goes idle, and can publish HTTPS routes per port.
Where boat fits
Google and Azure both run agents inside a hyperscaler account, and both are built more for tool calls than for agents that work like a person at a computer. If your agents install tools, run Docker and emulators, open a browser on a screen or serve what they build, boat gives each one a full VM that keeps its disk. boat has no pause with memory and no US region.
Root, Docker and /dev/kvm on every VM
Google runs commands as appuser with no sudo and documents no Docker, so the 50-agent fleet does not fit there. On Azure, Docker means picking an image like azure-dev, and nested virtualization for Android emulators is not documented. boat has all three on every sandbox.
A screen, SSH and shareable URLs
Google gives a browser view and documents no public URL; Azure publishes HTTPS routes and documents no SSH or desktop. You set up a VPC endpoint or tunnel on Google, and a display and streaming stack in your own image on Azure. boat ships SSH, a 60fps desktop and token-gated HTTPS URLs.
Coding agents driven from outside
Google's Managed Agents API is in Preview and runs Antigravity. Azure's Claude Code image runs through raw exec, and its built-in connections put tokens inside the sandbox. boat drives seven harnesses through one API, and its optional paid credentials proxy keeps real keys of supported providers out.
Use boat when your agents need a whole computer: root, Docker, /dev/kvm, SSH, a 60fps desktop and public HTTPS URLs on a full VM that keeps its disk, with coding agents installed. boat runs only in Germany, Finland and France, and caps starts per plan. Compare it one to one: boat vs Google Agent Runtime and boat vs Azure Container Apps Sandboxes.
boat's own limits
- Plans start at $20 a month. The plan price comes back as VM time for that month. Unused time does not carry over. Source
- Starts (create, fork, resume): the $20 plan includes 12 a minute, 60 an hour and 200 a day. The $2,000 plan includes 90 a minute, 600 an hour and 2,400 a day. Extra starts cost $0.03 to $0.07. A start fails with 429 above 3 times the included starts per minute, or 6 times per hour or per day. Source
- A stopped sandbox has no compute charge. boat keeps its disk for at least 14 days. A saved snapshot keeps it longer; extra saved snapshots and the AI credentials proxy are paid. Source
- Auto-stop defaults to 1 hour. You can turn it off on a paid plan. The trial allows 2 hours at most. Source
- boat saves the disk every minute and at each stop. Memory, running programs and /tmp are not saved: a resume is like a reboot. A machine failure loses writes after the last good snapshot. Source
- Machines run in Germany, Finland and France only. Snapshot copies can be stored outside the EU. Source
- Public IPv6 and outbound IPv4. No dedicated IPv4. The IP changes at each resume. Source
- Hosts are AMD Ryzen 7 9700X and Ryzen 9 9950X. vCPUs are shared. No GPU. You cannot boot your own Docker image. Source
Questions people ask
Is Google Agent Runtime cheaper than Azure Container Apps Sandboxes?
They are close. Azure comes out a little lower in two of our scenarios: one coding agent on workdays is $68 a month on Azure and $73 on Google, and sandboxes for your users are $989 against $1,030. For 1,000 short jobs at once Google is lower, $2,060 against $2,148. boat comes to $20 and $200 a month for the first two, and does not fit the burst.
Which is faster?
Neither Google Agent Runtime nor Azure Container Apps Sandboxes is in the public benchmarks we use, so we cannot say. For reference, boat did 28.06 runs/s on the V8 Web Tooling Benchmark at 4 vCPU / 8 GB in StarSling's run of 3 October 2026, on an AMD Ryzen 9 9950X host.
Can Google Agent Runtime or Azure Container Apps Sandboxes run Docker?
Google's sandboxes run commands as appuser with no sudo, images must not need root, and Google documents no Docker inside; you build images outside and push them to Artifact Registry. On Azure it depends on the image: azure-dev ships Docker on systemd, while the default ubuntu image has only Buildah. boat runs Docker and /dev/kvm on every sandbox.
Can I SSH into either one?
Neither documents SSH. Google runs commands through its API, and Azure gives an interactive shell through the aca CLI, the SDK or the portal. Every boat sandbox has SSH.
Can a sandbox keep its memory when it stops?
Both can. Google snapshots (Preview) keep memory and files, and Azure suspends a sandbox with its memory and resumes it in under a second, waking on traffic. boat saves the disk every minute and at each stop, but a resume is like a reboot.
Where do they run?
Google runs in 23 locations across the US, Europe, the Middle East, Asia, Canada, Australia and Brazil, and Azure in 40 regions, 9 of them in the EU. boat runs on machines in Germany, Finland and France; you cannot pick a region, and snapshot copies can be stored outside the EU.
Try boat
Install the CLI, then run boat new.
7-day free trial
More comparisons
- boat vs Google Agent Runtime
- boat vs Azure Container Apps Sandboxes
- E2B vs Google Agent Runtime
- Daytona vs Google Agent Runtime
- Modal vs Google Agent Runtime
- Vercel Sandbox vs Google Agent Runtime
- Cloudflare Sandbox vs Google Agent Runtime
- Blaxel vs Google Agent Runtime
- Runloop vs Google Agent Runtime
- Tensorlake vs Google Agent Runtime
- Morph Cloud vs Google Agent Runtime
- CodeSandbox SDK vs Google Agent Runtime