Nucleic
Support

Nucleic Help Center

Everything you need to run agents safely — isolating them with Nucleic Control, sandboxing their work, and handing each one a whole virtual machine of its own. Pick a topic below.

Nucleic Control

The hardened path where Nucleic clones, sandboxes, and observes a project directly — the foundation for unattended work, Autoship, and Orchestra.

What is Nucleic Control?

What Control does, how to turn on the container service, and what it unlocks — sandboxed sessions, certain git observation, Autoship, and Orchestra.

Sandboxes

Every sandboxed session runs in its own lightweight Linux VM — isolated from your host and from other sessions.

How agent sandboxing works

What the sandbox is, what it isolates, the per-project options (idle timeout, host exec, custom images), and why the Swift toolchain lives in a VM instead.

Virtual Machines

Give an agent a whole machine of its own — a per-session macOS or Linux VM for building, testing, and computer use, so parallel work never collides on your host.

Virtual machines: overview

What macOS and Linux VMs are, the hardware and disk requirements, how per-session VMs are cloned and torn down, and how many run at once.

Building the base image

The one-time golden base install: what the build does, choosing a guest macOS version, and pointing at a prebuilt bundle or a specific restore image.

Computer use: the two kinds

Let an agent see and drive a VM's screen. The two kinds — kernel-level use and semantic understanding — what each does and when to reach for it.

Setting up AX-based computer useAdvanced

The optional in-guest agent for semantic control — act on a UI element by identity, and inspect an app's accessibility tree. Fully automated on macOS 27.

Can't find what you need?

These guides grow with Nucleic. If something's missing or unclear, email hello@nucleic.blakeslee.xyz — we're happy to help.