Security & reliability

Powerful automation, clear boundaries.

SlateVM gives AI agents real authority over virtual machines — so it keeps that authority local, consented, and scoped on every host.

On every host

The MCP server listens only on your machine. Nothing is exposed to the network.

Each AI client must be approved once before it can act; approvals can be reset at any time, and quitting SlateVM ends all AI access.

Screen capture and input reach only the selected VM’s guest — never your desktop.

Each VM’s process accepts commands only from SlateVM itself.

Release manifests are signed and verified before an update installs.

Each VM is a self-contained bundle with its own disk, configuration, and state, with clean stop and delete.

Signed and notarized app with hardened runtime; privileged networking separated into its own service.

Guests run in isolated partitions of Windows’ built-in hypervisor; installed under Program Files with per-user settings.

Guests run on the kernel hypervisor with the host’s standard process isolation.

Live state everywhere: installing, running, suspended, stopped — and a stop the guest ignores is reported as still running.

Diagnostics, job status, and logs help you and your agent understand and recover from failures.

Deleting a VM stops it first and moves its bundle to the trash, along with its desktop shortcut.