Claude Managed Agents vs smol machines

Claude Managed Agents is Anthropic's hosted agent: you send Claude a task and it works in a small container that Anthropic runs. smol machines rents you Linux microVMs with their own kernel, built on an open-source engine, and you bring the agent.

The companies and products

Claude Managed Agents

Claude Managed Agents is Anthropic's ready-made agent harness: Claude runs as an autonomous agent in a cloud sandbox that Anthropic manages, and you talk to it through an events API.

This is an agent platform, not a sandbox you rent. You define an agent (model, prompt, tools, MCP servers, skills) and an environment, then start sessions, send messages as events and stream back what Claude does; the harness handles tool calls, prompt caching and compaction. Each session gets its own Ubuntu container, or runs its tools on a worker you host yourself. It is built for teams that want Claude doing long tasks without building their own agent loop or sandbox.

Isolation: Each session gets its own isolated Linux container on Anthropic-managed infrastructure
Sandbox size: Ubuntu 24.04, x86_64, up to 8 GB memory and 10 GB disk; vCPU count not published
Models: Claude models only, through Anthropic's own harness
Billing: Model tokens at API rates, plus a session runtime fee that accrues only while a session is running
Self-hosted sandboxes: Tools can run on a worker in your own infrastructure while Anthropic keeps the orchestration
Data location: Workspace geo, which sets where data is stored and code runs, is US only today; inference can be global or US
Status: Beta, behind the managed-agents-2026-04-01 header; not eligible for Zero Data Retention or a HIPAA BAA

smol machines

smol machines makes fast, isolated Linux microVMs that run the same way on your laptop, in their hosted cloud, or on your own servers.

The core is smolvm, an open-source microVM engine built on libkrun, where each machine has its own guest kernel. smol cloud is the paid hosted service that runs fleets of those machines behind a REST API and Node, Python and Rust SDKs. It is built for developers who want one VM model from local development to production, with live branching of running machines.

Isolation: Hardware-isolated microVM with its own guest kernel (libkrun)
Company: Founded 2026, Y Combinator Spring 2026, San Francisco, team of one listed
Open source: smolvm engine and smol SDK are open source and run on macOS, Linux and Windows hosts
Guests: Linux only; no Windows or macOS guests
Regions: No region field in the cloud create API
Uptime SLA: 99.95% on Pro, Startup and Enterprise

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. “106th cheapest of 367” means 105 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	Claude Managed Agents: price rank	smol machines: price rank	bochka: price rank
Software factory 100 at once · 4 vCPU / 8 GB · 2,200 sessions of 8 h a month Managed Agents gives a coding agent a container with up to 8 GB of memory and Docker only in limited availability; smol machines runs Docker in a full microVM, and bochka adds /dev/kvm, SSH and URLs to open the app.	106th of 367, with limits $964 a month	73rd of 367, with limits $604 a month	3rd cheapest of 367 $634 a month
Always-on AI employees 50 machines 24/7 · 2 vCPU / 4 GB Managed Agents sessions sleep for free between messages, but sandbox files are kept only 30 days; smol cloud has no desktop or browser and its secrets are readable inside the VM.	219th of 366, with limits $2,511 a month	148th of 366, with limits $1,038 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 smol machines fits short code runs well, with a one-minute minimum and a /code endpoint, while Managed Agents runtime only exists inside a Claude agent session with tokens on top.	163rd of 345, with limits $756 a month	12th cheapest of 345 $140 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 Your users cannot open what a Managed Agents sandbox runs because it has no inbound ports; smol machines publishes up to 16 ports and bochka up to 50, with a desktop.	231st of 361, with limits $823 a month	7th cheapest of 361 $531 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 smol machines can branch a running machine with its memory and processes, so many rollouts start from one state; Managed Agents has no fork at all.	252nd of 360, with limits $3,978 a month	111th of 360, with limits $6,978 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	244th of 364, with limits $320 a month	36th cheapest of 364 $670 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 CI jobs that build Android apps or boot VMs need /dev/kvm, which bochka has, smol cloud does not expose and Managed Agents does not document.	227th of 377, with limits $96 a month	66th cheapest of 377 $514 a month	4th cheapest of 377 $86 a month
24/7 dev boxes 1 machine 24/7 · 4 vCPU / 8 GB Neither Managed Agents nor smol machines offers SSH, and a stopped smol machine keeps paying for its disk; bochka has SSH and a stopped sandbox costs nothing.	224th of 368, with limits $48 a month	216th of 368, with limits $29 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 Managed Agents ships headless Chromium with Playwright and smol machines has a guide to run your own Chromium; bochka streams a full Linux desktop at 60fps with computer use.	127th of 369, with limits $288 a month	90th of 369, with limits $156 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	126th of 367, with limits $288 a month	176th of 367, with limits $136 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	127th of 346, with limits $390 a month	241st of 346, with limits $242 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	155th of 338, with limits $4,798 a month	148th of 338, with limits $4,344 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	179th of 360, with limits $160 a month	252nd of 360, with limits $377 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	Can't run it	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	Can't run it

Key features

Feature	Claude Managed Agents	smol machines	bochka
What it is	A hosted Claude agent: Anthropic's harness runs Claude in a sandbox and you send it events	Linux microVMs behind a REST API and SDKs, from an open-source engine	A sandbox platform: each agent gets its own full Linux machine, with coding agents built in
Isolation	Isolated Linux container per session	libkrun microVM with its own guest kernel	Full VM with its own kernel on dedicated AMD Ryzen 9 9950X hosts
Machine size	Up to 8 GB memory and 10 GB disk; vCPU not published; not configurable	Up to 32 vCPU / 32 GB on Startup, 16 vCPU / 24 GB on Pro	Up to 16 vCPU / 32 GB on the $100 plan and above
Docker and /dev/kvm	Docker listed as limited availability; root, systemd and /dev/kvm not documented	Docker inside the VM; nested virtualization not exposed on hosted machines	Root, Docker, systemd and /dev/kvm for Android emulators and VMs
What persists	Sandbox is checkpointed when a session goes idle; state is kept only 30 days after the sandbox is created	Filesystem survives a stop, /tmp is emptied; a stopped machine's disk keeps billing	The whole disk, saved every minute and on stop; a stopped sandbox costs nothing
Fork	No fork or clone of a session or sandbox	Branch a running machine with its memory and processes; checkpoints capture RAM to a file	Fork a running sandbox with its disk
Wake on next message	An idle session resumes from its checkpoint on the next message, and idle time is not billed	A command or terminal starts a stopped machine; idle apps wake on the next request	No; a stopped sandbox is started by API
Inbound access and shell	No inbound ports or SSH; reached only through the sessions and events API	Up to 16 published ports, a command API and a WebSocket terminal; no SSH or dedicated IP	SSH, HTTPS URLs on up to 50 ports, raw TCP, dedicated IPv4 or IPv6
Outbound rules	Unrestricted, or limited to listed hosts plus optional package registries and MCP servers	Blocked, open, or an allowlist of hosts and CIDR ranges	Open internet, no provider-side allowlist
Browser and desktop	Headless Chromium with Playwright; no desktop or computer-use tool	No hosted desktop; a guide shows running your own Chromium inside a machine	Linux desktop streamed at 60fps, with Chromium and computer use built in
Agents	Anthropic's harness with Claude models only, with steering, multiagent threads, memory stores and webhooks	Registry images with the Codex CLI and the Harbor eval harness, run through the command API	Claude Code, Codex and five more preinstalled, driven from outside with a prompt, steer and events API
Keeping API keys out of the sandbox	Vault secrets live in the sandbox as placeholders and are swapped in at egress, per host	Cloud secrets are environment variables the guest can read; placeholder swapping only in self-hosted smolvm	Credentials proxy for model providers and any HTTPS API; switch a credential to another account when credits run out
Self-hosting	Self-hosted sandboxes run tools on your own workers while Anthropic keeps the orchestration	Open-source smolvm runs on your laptop or your own servers, with the same .smolmachine artifact	No
Where data lives	US workspace geo only	No region choice in the cloud API	EU only (Germany, Finland, France), with zero data retention

Every feature Battleships tracks (43)

Feature	Claude Managed Agents	smol machines	bochka
Full VM (own kernel)	No	Yes	Yes
Docker inside	Not published	Yes	Yes
Nested virtualization	No	No	Yes
Root	Not published	Not published	Yes
Systemd	Not published	Not published	Yes
Desktop GUI	No	No	Yes
Browser	Yes	No	Yes
Computer-use API	No	No	Yes
SSH	No	No	Yes
Persistent disk	Yes	Yes	Yes
Pause/resume	Yes	Yes	Yes
Automatic snapshots	Yes	No	Yes
Fork/clone	No	Yes	Yes
Live fork (no pause)	No	Yes	Yes
Memory fork	No	Yes	No
Custom image (Docker or snapshot)	Yes	Yes	Yes
≥24 h sessions	Yes	Yes	Yes
Idle auto-stop	Yes	Yes	No
Wake on request	Yes	Yes	No
HTTPS preview URLs	No	Yes	Yes
Raw TCP inbound	No	Not published	Yes
Public IPv4	No	No	No
Egress allowlist	Yes	Yes	No
Secret proxy	Yes	No	Yes
Secret proxy for any API	Yes	No	Yes
Preinstalled agents	Yes	Yes	Yes
Agent harness API	Yes	No	Yes
MCP server	Yes	Yes	No
GPU	No	No	No
Windows	No	No	No
Arm64	No	Yes	No
Self-hosting	Yes	Yes	No
BYOC	Yes	Yes	No
Open source	No	Yes	No
Scoped API keys	Not published	Not published	Yes
Spend limits	Yes	Yes	Yes
Usage per sandbox (API)	Yes	Yes	Yes
Audit logs	Not published	Not published	No
SSO	Yes	Yes	No
SOC 2	Yes	No	No
HIPAA	No	No	No
EU data residency	No	No	Yes
Zero data retention	No	Not published	Yes

Performance

Claude Managed AgentsNot measured

smol machinesNot measured

bochka26.69 runs/s

Neither has a comparable CPU measurement: Managed Agents and smol cloud were not part of the hpc-sandbox-benchmarks run, and Anthropic does not publish how many vCPUs a sandbox gets. bochka did 26.69 runs per second on that Node.js build-and-test loop at 4 vCPU / 8 GB. smol machines publishes a boot time under 200ms for smolvm, which is a start-time figure, not CPU speed.

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

Scenario	Claude Managed Agents, a month	smol machines, a month	bochka, 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	$13 · Claude API (pay-as-you-go, usage tiers) · with limits	$25 · Standard	$20 · $20/mo
One agent that never stops 4 vCPU / 8 GB, 50 GB disk, running 24/7 all month	$53 · Claude API (pay-as-you-go, usage tiers) · with limits	$93 · Standard	$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	$634 · Claude API (pay-as-you-go, usage tiers) · with limits	$1,139 · Pro	$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	$360 · Claude API (pay-as-you-go, usage tiers) · with limits	$465 · Startup	$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	$796 · Claude API (pay-as-you-go, usage tiers) · with limits	$1,389 · Startup	$2,000 · $2000/mo · with limits

Claude Managed Agents: Managed Agents bills model tokens at API rates plus a flat session runtime fee, measured only while the session is running, including time spent waiting on the model. The table shows runtime only; tokens come on top and usually cost more.

smol machines: smol cloud charges a base fee for every hour a machine runs, plus the CPU and memory the guest actually uses, with a one-minute minimum and disk billed even while stopped. Pro and Startup fees are added on top of usage, not credit.

smol machines: 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

A Managed Agents sandbox has up to 8 GB of memory and 10 GB of disk, an unpublished vCPU count and Docker only in limited availability, so no scenario counts as a clean fit for it. Anthropic publishes no concurrent-session cap, only 300 creates per minute and usage-tier spend limits. smol machines caps machines by plan, 20 on Standard, 100 on Pro and 1,000 on Startup, with stopped machines holding a slot, so 50 agents needs Pro and 1,000 short jobs needs Startup. bochka queues starts above 90 a minute at 1,000 short jobs at once.

Scenario	Claude Managed Agents	smol machines	bochka
One coding agent, workdays	adjacent product, not a like-for-like sandbox machine size not published different kind of product: agent-platform machine size per session not published	None that bite	None that bite
One agent that never stops	adjacent product, not a like-for-like sandbox machine size not published different kind of product: agent-platform machine size per session not published	None that bite	None that bite
50 agents in parallel	adjacent product, not a like-for-like sandbox not confirmed: Docker inside machine size not published different kind of product: agent-platform	None that bite	None that bite
Sandboxes for your users	adjacent product, not a like-for-like sandbox machine size not published different kind of product: agent-platform not confirmed: how many machines can run at once (you need 300; the typical published limit is 50)	None that bite	None that bite
1,000 short jobs at once	adjacent product, not a like-for-like sandbox machine size not published different kind of product: agent-platform not confirmed: how many machines can run at once (you need 1,000; the typical published limit is 50)	None that bite	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 Claude Managed Agents when

you want Claude itself to do long text and code tasks with no harness to build, idle time that costs nothing, and the option to run tools on your own workers.

No agent loop to build. Teams that want Claude doing long tasks get the harness, caching, compaction, memory and webhooks ready to use.
Idle sessions sleep for free. Agents that wait a long time between messages pay runtime only while working, and resume on the next message.
Host allowlists and your own workers. Teams with strict network rules can limit sandboxes to listed hosts, or run every tool call on their own workers.

Use smol machines when

you want one open-source VM engine from your laptop to your own servers, big machines, or many branches of a running machine with its memory.

Open source, from laptop to your own servers. Teams that want the same VM locally, in a hosted cloud and on their own servers can use one open-source engine.
Live branches with memory. Useful for RL and test fan-out: branch a running machine with its memory and processes into many children.
Big machines and big bursts. Teams that need up to 32 vCPU per machine, or 1,000 machines at once on Startup, get both.

Where bochka fits

Managed Agents gives you Claude in a small container you cannot open, and smol machines gives you a bare VM and leaves the agent, desktop and keys to you. If your agents need a whole machine you can reach, with agents and a desktop ready to use, look at bochka.

A whole machine that keeps its disk

Builds that need Docker or an Android emulator stall in a Managed Agents container, and its files expire after 30 days. smol cloud runs Docker but has no /dev/kvm, so emulators mean self-hosting smolvm.

Open what the agent builds and watch it work

A Managed Agents sandbox has no inbound ports or desktop. smol cloud has up to 16 ports but no SSH or desktop, so you build and maintain the desktop and screen stream yourself.

Any agent, with keys kept outside

Managed Agents runs only Anthropic's harness with Claude, so Claude Code or Codex runs elsewhere. On smol cloud you write the harness on exec calls, and secrets sit in the VM where a prompt injection can leak them.

Use bochka when your agents need a whole machine for real work: Docker, emulators, SSH, ports you can open, a desktop and browser, a disk that lasts, and Claude Code, Codex or other agents you drive from outside. Compare it one to one: bochka vs Claude Managed Agents and bochka vs smol machines.

Questions people ask

Is Claude Managed Agents cheaper than smol machines?

For runtime alone, Managed Agents costs less in all five scenarios on this page, but that leaves out model tokens, which usually cost more than the runtime. Its sandbox size is also not published, so none of those scenarios counts as a clean fit for it.

Which is faster, Claude Managed Agents or smol machines?

Neither has a comparable CPU measurement, and Anthropic does not publish a sandbox's vCPU count. smol machines publishes a boot time under 200ms for smolvm. bochka did 26.69 runs per second on the Node.js build-and-test benchmark.

Can Claude Managed Agents or smol machines run Docker?

Managed Agents lists Docker as limited availability. smol machines runs Docker inside the VM, with Docker data on the machine's ext4 disk, but does not expose nested virtualization on hosted machines. bochka runs Docker and exposes /dev/kvm for Android emulators and VMs.

Can I run Claude Code on Claude Managed Agents or smol machines?

Managed Agents runs Anthropic's own harness with Claude models, not Claude Code. smol machines has registry images with the Codex CLI and the Harbor eval harness, run through its command API. On bochka, Claude Code, Codex and five other agents come preinstalled.

Can I self-host either one?

Managed Agents can run tools on workers in your own infrastructure while Anthropic keeps the orchestration. smol machines' engine, smolvm, is open source and runs on your laptop or servers. bochka is a hosted service only.

Do they keep API keys out of the sandbox?

Managed Agents keeps vault secrets as placeholders and swaps them in at egress. smol cloud secrets are plaintext environment variables inside the guest. bochka's credentials proxy keeps real keys outside the sandbox for model providers and any HTTPS API.

Try bochka

Install the CLI, then run bochka new.

Product preview