Agent Sandboxesagentsandboxes.org
Head-to-head

Celesto (SmolVM) vs smol machines (smolvm)

Answer in brief

Choose Celesto (SmolVM) for running agent computers as microVMs on your own Linux or Apple Silicon machine, with the same API available as a hosted cloud computer; choose smol machines (smolvm) for branching and checkpointing microVMs locally for agents and RL rollouts, then running the same machine on a hosted cloud. Celesto (SmolVM) uses Firecracker / QEMU microVM isolation and is open source; smol machines (smolvm) uses libkrun microVM isolation and is open source.

By the Agent Sandboxes editors · Updated

Key differences between Celesto (SmolVM) and smol machines (smolvm)

  • Isolation: Celesto (SmolVM) uses Firecracker / QEMU microVM; smol machines (smolvm) uses libkrun microVM.
  • SDKs and interfaces: Only Celesto (SmolVM) lists MCP server. Only smol machines (smolvm) lists Rust, Go and REST API.

What they have in common

  • Both are open-source projects you run yourself.
  • Both use microVM isolation.
  • Both are open source.
  • Both can be self-hosted.
  • Both support snapshots.
  • Both can pause and resume a sandbox.
  • Both list Python, TypeScript and CLI.

Celesto (SmolVM) vs smol machines (smolvm) at a glance

AttributeCelesto (SmolVM)smol machines (smolvm)
What it isOpen-source microVM "computers" for agents from Celesto AI, formerly published as SmolVM, with Firecracker on Linux and QEMU on macOS, plus a hosted cloud option.Apache-2.0 runtime that boots OCI images as branchable libkrun microVMs on macOS, Linux and Windows, with checkpoints, portable .smolmachine files and a hosted smol cloud behind the same SDK.
TypeOpen-source projectOpen-source project
IsolationMicroVM per computer with hardware-level separation from the host: Firecracker on Linux (KVM), QEMU on macOS; the PyPI listing also names libkrun as a backendHardware-virtualized microVM with its own guest kernel for every workload: libkrun as the VMM and smol machines' libkrunfw guest kernel, on Hypervisor.framework (macOS), KVM (Linux) or the Windows Hypervisor Platform; the VMM is linked into the binary with no host daemon. smol cloud runs each machine in its own hardware-isolated microVM
Open sourceYes (Apache-2.0)Yes (Apache-2.0)
Self-host / BYOCYes / Not documentedYes / Not documented
SDKsPython, JavaScript/TypeScript, CLI, MCP serverJavaScript/TypeScript, Python, Rust, Go, CLI, REST API
Cold start (vendor claim)README: a local computer "starts in about 500 ms"; the website says hosted agent sandboxes "start in under 2 seconds"README comparison table: "<200ms to agent ready"; local VMs "boot in under a second" and a packed .smolmachine executable "boots in under 200 ms" (vendor claims). smol cloud machines start from warm pools
Max sessionNot documentedNot documented
PersistenceSnapshots "pause and restore memory, disk, and active processes"; persistent state survives across sessions; in the cloud, stopped computers keep files, packages and workspace state, and stopped root disks are kept free for 30 daysLocal machines created with machine create keep installed packages and state across restarts; machine run VMs are removed on exit. Checkpoints capture RAM, CPU state and disks and can be restored on another host or on smol cloud; branches are copy-on-write children of a live machine; a machine can be packed into a portable .smolmachine file and pushed to an OCI registry. On smol cloud the machine filesystem persists across exec and stop/start (not /tmp), and a stop exports a recovery point
Snapshots / pauseYes / YesYes / Yes
GPUNot documentedLocal smolvm only: Vulkan via virtio-gpu/Venus, and CUDA API remoting over vsock to a host NVIDIA GPU (not passthrough); Vulkan is unavailable on native Windows. The docs state that hosted smol cloud does not provide GPU machines
NetworkingRestricted-destination network controls are documented for Linux only (Firecracker, or QEMU with TAP); QEMU slirp supports open and off modes on macOS and Linux; the QEMU backend uses user-mode (slirp) networking with port forwarding; cloud computers can expose public portsNetworking is off by default; egress can be limited to named hosts, host patterns or CIDRs, and the allow list can be changed on a running machine. Credential substitution lets code use a secret bound to specific hosts while the guest sees a placeholder. Local --auto-ports mirrors listening guest ports onto 127.0.0.1; named networks let machines reach each other. smol cloud offers blocked, open and allowCidrs modes, published ports and optional public ingress
PricingCloud computers billed per hour of running time, prepaid with no monthly fee: Micro 1 vCPU/2 GB $0.015/h, Small 1 vCPU/4 GB $0.020/h, Standard 2 vCPU/4 GB $0.030/h, Memory 2 vCPU/8 GB $0.040/h, Large 4 vCPU/16 GB $0.080/h, XL 8 vCPU/32 GB $0.160/h; root disk included; paid accounts start with a default limit of 20 running sandboxes. Local runtime is freesmol cloud: base $0.04 per running machine-hour (1-minute minimum), plus $0.05 per active vCPU-hour, $0.0162 per GB-hour of memory in use, $0.0001 per GB-hour of disk used (~$0.073/GB-month) and $0.05 per GB egress; a beta 10% discount on spend above $2,000 a month. Plans: Standard $0 (20 machines, 8 vCPU and 16 GB per machine), Pro $20 a month plus usage (100 machines, 16 vCPU, 24 GB, 200 GB disk, 99.95% SLA), Startup $150 a month plus usage (1,000 machines, 32 vCPU, 32 GB, 200 GB disk, priority support), Enterprise custom. Local and self-hosted smolvm is free
Free tierOne-time $5 Micro runtime credit, no card, expires 30 days after activation (two retained Micro sandboxes); free to self-hostOne-time $100 sign-up credit on smol cloud, no card needed to start (up to 4 machines of up to 4 vCPU each until a card is added); smolvm is free to run locally or self-host
ComplianceNot documentedNot documented
RegionsNot documentedNo region list is published; the privacy policy names Google Cloud (US regions) as compute infrastructure and says data is processed in the United States
Company headquartersNot documentedNot documented
Good fit forRunning agent computers as microVMs on your own Linux or Apple Silicon machine, with the same API available as a hosted cloud computerBranching and checkpointing microVMs locally for agents and RL rollouts, then running the same machine on a hosted cloud

Verified November 14, 2026 (Celesto (SmolVM)) and November 18, 2026 (smol machines (smolvm)) from each provider’s documentation. Cold-start figures are vendor claims, not our measurements.

Which should you choose?

Reasons to choose Celesto (SmolVM)

  • You want a microVM runtime that works on an Apple Silicon laptop (QEMU) and on Linux (Firecracker) behind one API
  • You need snapshots that capture memory and running processes, not just a disk image
  • You want a cheap hosted fallback with the same SDK when a local machine is not available
  • You are building computer-use agents and want a browser or desktop inside the VM

Reasons to choose smol machines (smolvm)

  • You want hardware-isolated microVMs that run natively on a Mac, a Linux server or Windows from the same CLI and SDK.
  • You fan out agent tasks or RL rollouts by branching a warm, running machine instead of booting fresh ones.
  • You want to checkpoint a running machine and move it between hosts or into a hosted cloud.
  • You need egress allow-lists and credentials the guest code never reads.

Trade-offs to check first

  • Celesto (SmolVM): Alpha-labelled on PyPI; renamed from SmolVM to Celesto, with the smolvm command and Python import removed
  • Celesto (SmolVM): Restricted-destination network controls are Linux only; QEMU on macOS supports only open and off modes
  • smol machines (smolvm): GPU support is local only; the docs say hosted smol cloud does not provide GPU machines.
  • smol machines (smolvm): Guest architecture must match the host, and checkpoints do not restore across architectures.

Both are worth piloting with your own image and region. See the full Celesto (SmolVM) profile and smol machines (smolvm) profile, or the complete comparison.

More comparisons

Frequently asked questions

Is Celesto (SmolVM) or smol machines (smolvm) better for AI agents?
Neither is better for every workload. Celesto (SmolVM) suits running agent computers as microVMs on your own Linux or Apple Silicon machine, with the same API available as a hosted cloud computer. smol machines (smolvm) suits branching and checkpointing microVMs locally for agents and RL rollouts, then running the same machine on a hosted cloud.
What is the difference between Celesto (SmolVM) and smol machines (smolvm)?
Celesto (SmolVM) uses Firecracker / QEMU microVM; smol machines (smolvm) uses libkrun microVM. Only Celesto (SmolVM) lists MCP server. Only smol machines (smolvm) lists Rust, Go and REST API.
Which is open source, Celesto (SmolVM) or smol machines (smolvm)?
Both are open source (Celesto (SmolVM): Apache-2.0; smol machines (smolvm): Apache-2.0).
How do Celesto (SmolVM) and smol machines (smolvm) isolate code?
Celesto (SmolVM): MicroVM per computer with hardware-level separation from the host: Firecracker on Linux (KVM), QEMU on macOS; the PyPI listing also names libkrun as a backend. smol machines (smolvm): Hardware-virtualized microVM with its own guest kernel for every workload: libkrun as the VMM and smol machines' libkrunfw guest kernel, on Hypervisor.framework (macOS), KVM (Linux) or the Windows Hypervisor Platform; the VMM is linked into the binary with no host daemon. smol cloud runs each machine in its own hardware-isolated microVM.
How does Celesto (SmolVM) pricing compare with smol machines (smolvm)?
Celesto (SmolVM): Cloud computers billed per hour of running time, prepaid with no monthly fee: Micro 1 vCPU/2 GB $0.015/h, Small 1 vCPU/4 GB $0.020/h, Standard 2 vCPU/4 GB $0.030/h, Memory 2 vCPU/8 GB $0.040/h, Large 4 vCPU/16 GB $0.080/h, XL 8 vCPU/32 GB $0.160/h; root disk included; paid accounts start with a default limit of 20 running sandboxes. Local runtime is free. smol machines (smolvm): smol cloud: base $0.04 per running machine-hour (1-minute minimum), plus $0.05 per active vCPU-hour, $0.0162 per GB-hour of memory in use, $0.0001 per GB-hour of disk used (~$0.073/GB-month) and $0.05 per GB egress; a beta 10% discount on spend above $2,000 a month. Plans: Standard $0 (20 machines, 8 vCPU and 16 GB per machine), Pro $20 a month plus usage (100 machines, 16 vCPU, 24 GB, 200 GB disk, 99.95% SLA), Startup $150 a month plus usage (1,000 machines, 32 vCPU, 32 GB, 200 GB disk, priority support), Enterprise custom. Local and self-hosted smolvm is free. Rates change, so check the Celesto (SmolVM) and smol machines (smolvm) pricing pages before deciding.
Do Celesto (SmolVM) and smol machines (smolvm) support GPUs?
Celesto (SmolVM): Not documented. smol machines (smolvm): Local smolvm only: Vulkan via virtio-gpu/Venus, and CUDA API remoting over vsock to a host NVIDIA GPU (not passthrough); Vulkan is unavailable on native Windows. The docs state that hosted smol cloud does not provide GPU machines.
Can you self-host Celesto (SmolVM) or smol machines (smolvm)?
Yes, both can be self-hosted.
Do Celesto (SmolVM) and smol machines (smolvm) support snapshots or pause and resume?
Celesto (SmolVM): snapshots yes, pause and resume yes. smol machines (smolvm): snapshots yes, pause and resume yes.

Sources

  1. celesto.ai
  2. celesto.ai/docs/
  3. celesto.ai/docs/cloud/overview.md
  4. celesto.ai/pricing
  5. smolmachines.com
  6. smolmachines.com/pricing
  7. smolmachines.com/faq
  8. smolmachines.com/docs/

Check before you rely on it. Products change quickly, and this page reflects public documentation as of November 18, 2026. Confirm anything that matters in the provider’s own documentation.

Cite this page

Agent Sandboxes. (2026). Celesto (SmolVM) vs smol machines (smolvm). agentsandboxes.org. Last reviewed November 18, 2026. https://agentsandboxes.org/compare/smolvm-vs-smol-machines/