Install OpenClaw on macOS

Why macOS is still the cleanest first route

The Orange Paper treats local installation as the safest first learning path, and macOS remains one of the easiest environments to reason about if you want a visible, local-first setup.

You are not trying to build a production deployment yet. You are trying to prove four things:

• OpenClaw runs
• one model path works
• the onboarding flow is valid
• you understand the base operator loop

Before you start

Make sure you have:

1. Node.js 22 or later
2. a package manager such as npm, pnpm, or bun
3. Xcode Command Line Tools installed
4. one model plan in mind before onboarding

Recommended sequence

1. Prepare the machine

Install any missing runtime pieces first. A surprising number of “OpenClaw is broken” reports are really version-mismatch problems.

2. Install the CLI and onboard

The Orange Paper’s preferred starting point is:

npm install -g openclaw@latest
openclaw onboard --install-daemon

This gets you much closer to the current supported flow than trying to invent your own manual path immediately.

3. Validate the environment

After onboarding, treat verification as a separate step:

openclaw doctor

Do not debug channels, models, and installation all at once.

4. Create a backup before you get ambitious

The March 2026 line adds local backup tooling for a reason:

openclaw backup create

Do this before aggressive upgrades or config experiments.

5. Move to the next branch only after local success

Once local install is stable, the next real decision is usually:

• channels
• models
• hosting

Common macOS mistakes

Mixing historical shell state with a new runtime

Check your path, Node version, and package manager state first.

Debugging everything at once

Installation, channel setup, and model wiring are different problems.

Jumping to cloud too early

A correct local install is much more valuable than a vague half-working server.

Recommended next step

After a successful local install, go to the Channels Hub or the Models Hub.