Currently in active development. Follow our story

Machines for AI agents.

A Machine is a dedicated computer for an agent, with its own shell, files, ports, and state.

≡+ New chat
Checkout Flow Fix Session▾

Message Agent...

Machines

An isolated workspace where agents can run tasks, keep their state, and expose a result.

Good for:

  • Agent workspaces
  • Sandboxes
  • Preview environments

Every Machine includes

  • ShellA real Linux shell your agent can drive.
  • FilesA persistent filesystem that survives between turns.
  • PortsExpose a service and get a reachable URL.
  • LogsStream what the agent is doing as it runs.
  • LifecycleCreate, clone, stop, and clean up on demand.

Machines that start in 125ms

As little as 125ms from request to a ready workspace, with compute, files, and network already in place. No provisioning step and no queue.

Every agent gets its own microVM

Each Machine runs in a dedicated microVM with its own CPU, memory, and filesystem. Nothing is shared between agents, so one workspace cannot see or affect another.

Control Machines,
from your own code.

Skip the dashboard. Create Machines, stream logs, and manage their lifecycle through the REST API, so your own systems can drive SpaceScale directly.

Example request
curl -X POST https://api.spacescale.io/v1/machines \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "name": "checkout-base",
    "image": "ghcr.io/my-org/dev-workspace:latest"
  }'

“Every agent needs a machine of its own. An isolated workspace it can actually control, with its own shell, files, ports, and state.”

SPACESCALE TEAM

IN DEVELOPMENT

We are in active development

Read the engineering blog for the design decisions and tradeoffs behind SpaceScale as we build toward the first public release.

Read how we're building it