Nullspace gives AI agents composable machines: on-demand isolated machines with first-class lifecycle primitives, all driven from Python, TypeScript, the CLI, raw HTTP, or MCP. This is what it looks like:
from nullspace import Machine

with Machine.create(template="base", timeout=300) as machine:
    result = machine.commands.run("echo 'hello from a machine'", shell=True)
    print(result.stdout)

    machine.files.write("/workspace/app.py", "print('hi')\n")
    print(machine.commands.run("python3 /workspace/app.py", shell=True).stdout)
Commands run inside an isolated VM with its own filesystem — see Commands and Filesystem.
Private beta — signup is closed for now. Public availability timing will be announced when capacity opens. If you have a beta key, you can be running these snippets in about five minutes: create your first machine. The full stack is open source and can be self-hosted on a single host.

Get started

First machine in 5 minutes

Install the CLI, save your key, and run nullspace quickstart.

Copy a runnable example

Code interpreter, agents, fork, templates, volumes, desktop, and more.

Self-host the OSS stack

Run the whole platform on one Ubuntu/KVM host.

The lifecycle primitives

Everything in Nullspace builds on six explicit operations. Pause/resume and fork are what set it apart: full VM state is preserved exactly, wakes use lazy memory loading so they feel instant, and forks branch a running machine.
PrimitiveWhat it does
CreateBoot a fresh machine from a template, image, or snapshot — or claim a pre-booted one from a warm pool
ExecRun commands and processes, stream output
HibernatePause and persist full VM state
ResumeWake exactly where it paused, near-instantly
ForkBranch a running machine into independent children
DestroyTear down and release everything

Explore the platform

Machines

The machine lifecycle end to end.
Create & connect · Persistence · Fork example

Code Interpreter

Stateful notebook-style execution.
Analyze data · Charts · Example

Agents

Coding agents in machines, or your agent as a managed service.
Functions · Local MCP · Job example

Templates & warm pools

Bake dependencies once; start instantly.
Quickstart · Warm pools · Build example

Filesystem & volumes

Machine files plus durable shared storage.
Uploads · Volumes · Shared-volume example

Networking

Signed URLs, WebSockets, custom proxies.
Preview URLs · Access control · Port example

Computer use

GUI environments for agents and humans.
Automation · Viewer · Viewer example

Observability & access

Watch and debug long-running work.
Lifecycle events · Webhooks · SSH & PTY

Reference

Client reference

Choose Python, TypeScript, CLI, MCP, or raw HTTP.

Python SDK

Python package, object references, and task guides.

TypeScript SDK

Node and browser SDK for machines, files, templates, volumes, and Functions.

CLI

nullspace commands for machines, templates, volumes, and agents.

HTTP API

The full REST contract with an interactive playground.