Build

Build is the developer surface for creating isolated runtime environments, running work inside them, moving files, exposing services, and keeping state when a workflow needs to continue. Start here when you know what you want to build but are not sure which Nullspace primitive owns the job.

Choose The Primitive

Goal	Start With	Why
Run code or install dependencies in a clean VM	Machines	Machines are the runtime boundary for commands, files, networking, desktop, and lifecycle.
Execute commands, stream logs, or manage processes	Commands	Commands are the process API inside a machine.
Read, write, upload, download, or search files	Filesystem	Filesystem APIs are for machine-local paths such as `/workspace`.
Run notebook-style code with structured results	Code Interpreter	Code Interpreter adds stateful cells, artifacts, and language contexts.
Automate a browser or GUI app	Desktop Environments	Desktop machines expose screenshots, input, windows, clipboard, viewer sessions, and recordings.
Avoid repeating setup on every machine create	Templates	Templates turn setup steps or images into reusable machine environments.
Persist shared files outside VM memory	Volumes	Volumes are durable filesystems that can mount into one or more machines.
Expose an HTTP or WebSocket service	Networking	Preview URLs and WebSockets route traffic to services inside machines.
Track long-running work and lifecycle state	Observability	Lifecycle events, webhooks, and monitor streams show what changed and why.

Common Workflows

Workflow	Path
Create a machine, run a command, read a file, then destroy it	Machines -> Commands -> Filesystem
Start a web app and share it with a signed URL	Commands -> Preview URLs
Build a reusable Python or Node environment	Templates -> Template build -> Logging
Keep generated data across machine lifetimes	Volumes -> Mounting volumes
Let an agent use a browser or desktop app	Desktop automation -> Managed viewer
Debug a retained or long-running machine	Monitor -> PTY or SSH

First Working Shape

Most Build workflows start with a machine and then opt into the primitives they need:

from nullspace import Machine

with Machine.create(template="base", timeout=300) as machine:
    result = machine.commands.run("python3 --version", shell=True)
    machine.files.write("/workspace/result.txt", result.stdout)
    print(machine.id, machine.files.read("/workspace/result.txt").strip())

Use a custom Template when this setup becomes repeatable, a Volume when output must outlive the VM, and a Preview URL when a process inside the machine needs inbound traffic.

Build Areas

Machines

Create isolated runtimes and choose lifecycle behavior.

Commands

Run foreground, streaming, and background work.

Filesystem

Move data in and out of machine paths.

Networking

Expose services and connect over HTTP, WebSockets, or SSH.

Templates

Build reusable environments for faster starts.

Volumes

Persist and share files outside VM memory.

Desktop Environments

Automate GUI workflows and supervise them in a viewer.

Observability

Subscribe to lifecycle events, webhooks, and metrics.

Get started

Machines

Commands

Filesystem

Code Interpreter

Desktop

Templates

Volumes

Networking

Interactive access

Observability

Choose The Primitive

Common Workflows

First Working Shape

Build Areas

Machines

Commands

Filesystem

Networking

Templates

Volumes

Desktop Environments

Observability

​Choose The Primitive

​Common Workflows

​First Working Shape

​Build Areas

Machines

Commands

Filesystem

Networking

Templates

Volumes

Desktop Environments

Observability

​Related

Choose The Primitive

Common Workflows

First Working Shape

Build Areas

Related