Services and ports

Long-running services are useful for preview servers, API sandboxes, browser targets, and test dependencies.

Start a service

mvmctl machine run --manifest . --name servicebox --port 8080:8000 -- \
  python -m http.server 8000

The exact command surface may vary by backend and profile; check CLI Commands for current flags.

Security checklist

• Bind only the ports you need.
• Prefer loopback host exposure for local development.
• Do not expose admin consoles without auth.
• Review ingress admission and teardown audit events.
• Stop or cold-pause the sandbox when the service no longer needs CPU.

Planned runtime SDK shape

Status: Planned lifecycle API.

Python

sandbox = Sandbox.create(
    image="nix:./flake#service",
    network={"mode": "none", "ports": [{
        "mapping_id": 1, "proto": "tcp", "transform": "opaque",
        "host_addr": "127.0.0.1", "host": 8080,
        "guest_addr": "127.0.0.1", "guest": 8000,
    }]},
)
proc = sandbox.exec_background(["python", "-m", "http.server", "8000"])

TypeScript

const sandbox = await Sandbox.create({
  image: "nix:./flake#service",
  network: { mode: "none", ports: [{
    mapping_id: 1, proto: "tcp", transform: "opaque",
    host_addr: "127.0.0.1", host: 8080,
    guest_addr: "127.0.0.1", guest: 8000,
  }] },
});
const proc = await sandbox.execBackground(["node", "server.js"]);