How to Install OpenClaw on Windows, Mac, Linux, or the Cloud

AI Takeaway

• What is the easiest way to install OpenClaw? Use Windows Hub or the macOS app. On macOS, Linux, or WSL2, use the official installer script.
• How do you install OpenClaw on Windows? Use Windows Hub for the simplest desktop setup, PowerShell for a native CLI installation, or WSL2 for a Linux-compatible environment.
• How do you install OpenClaw on Mac? Launch the macOS app and choose This Mac, or run the installer script in Terminal. Approve only the permissions your tools need.
• How do you know the installation worked? Confirm the CLI version, run openclaw doctor, check the Gateway, open the Dashboard, and send a test message.
• Can it run without leaving a computer on? Yes. A VPS or MyClaw instance keeps it available around the clock.

Choose the Right OpenClaw Setup

Choose based on local access, uptime, and infrastructure control.

Setup	Best For	Main Trade-Off
Windows or Mac app	Easy local setup	The computer must remain available
Installer script	Terminal users on macOS, Linux, or WSL2	You maintain the runtime
Dedicated device	Local tools and a home agent	Hardware, power, security, and updates
VPS	Remote 24/7 self-hosting	Server administration and recovery
Managed hosting	A persistent agent without server maintenance	Paid hosting; model usage may be separate

Use a local app for desktop features and remote hosting for continuous work. The OpenClaw hosting guide compares these environments.

What You Need Before Installing OpenClaw

You do not need Homebrew, pnpm, or Docker before using the recommended installer.

• A supported Windows, macOS, or Linux computer
• An internet connection for downloading the runtime
• An account or API key for a supported model provider
• A browser for the Dashboard
• An optional messaging account

The installer provides Node when needed. Manual npm installs require a supported Node release; Node 26 is the recommended default. pnpm is only needed for source builds.

How to Install OpenClaw on Windows

Windows Hub is the easiest route, PowerShell suits terminal-first use, and WSL2 fits Linux-based workflows.

Option 1: Install Windows Hub

Windows Hub supports Windows 10 20H2+ and Windows 11. Choose x64 for most computers or ARM64 for Windows-on-ARM.

Launch OpenClaw Companion and choose local setup. It creates an app-owned WSL Gateway, installs OpenClaw, and pairs the app without changing existing Ubuntu distributions or requiring administrator privileges.

When the tray icon turns green, confirm connection and node health in Command Center.

Option 2: Install from PowerShell

Open PowerShell and run:

iwr -useb https://openclaw.ai/install.ps1 | iex

The installer starts guided onboarding. Then verify:

openclaw --version

openclaw doctor

openclaw gateway status --json

Managed startup uses a Scheduled Task or a per-user Startup fallback.

Option 3: Use WSL2

Choose WSL2 for Linux tools or a headless Gateway. Install Ubuntu, enable systemd, and run the Linux installer inside WSL.

Windows browsers and files may need extra configuration, so prefer Windows Hub unless you need Linux.

How to Install OpenClaw on Mac

The macOS app is simplest. Use Terminal for manual control.

Option 1: Use the Native macOS App

Install the current .dmg, launch OpenClaw, and choose This Mac for a local Gateway or select an existing remote Gateway.

Automatic setup installs Node, the matching CLI, and launchd service without Terminal, Homebrew, or administrator access.

Open System Settings > Privacy & Security to manage permissions. Accessibility and Screen Recording enable computer control; Microphone and Speech Recognition enable voice features. Basic chat does not need every permission.

Option 2: Install from Terminal

Run the official installer:

curl -fsSL https://openclaw.ai/install.sh | bash

The script installs Node if needed and launches onboarding. The daemon keeps the Gateway running after Terminal closes. If the app or Dashboard fails, check Gateway status before reinstalling.

How to Install OpenClaw on Linux

On a Linux desktop or server, use the same installer as macOS:

curl -fsSL https://openclaw.ai/install.sh | bash

Complete onboarding and enable managed startup. Run openclaw gateway status to confirm the service is active.

Docker, Podman, Nix, and source builds suit specialized deployments. Start with the installer unless you need packaging-level control.

Install OpenClaw with npm or from Source

The original global install remains useful when you already manage Node:

npm install -g openclaw@latest

openclaw onboard --install-daemon

Build from source only when you want to modify OpenClaw or contribute:

git clone https://github.com/openclaw/openclaw.git

cd openclaw

pnpm install && pnpm build && pnpm ui:build

pnpm link --global

openclaw onboard --install-daemon

Source builds require Git and pnpm. Avoid sudo with global packages; correct the npm prefix or PATH instead of running OpenClaw as root.

Skip the Setup, Run Best-in-Class AI Agents Now

Launch a managed OpenClaw or Hermes Agent workspace in minutes, with always-on hosting, updates, and support handled by MyClaw.

Get Started →

Complete Setup and Send Your First Message

The CLI is only the first checkpoint. OpenClaw also needs a working model and Gateway.

Choose a Model and Start the Gateway

OpenClaw is the runtime; its model supplies inference. Start with one. The model catalog compares cost, context, and capability.

Onboarding configures the provider, workspace, and Gateway service. Keep credentials out of prompts and shared files.

Verify the Installation

Run these checks in order:

openclaw --version

openclaw doctor

openclaw gateway status

openclaw dashboard

A healthy setup returns a version, passes Doctor and the Gateway probe, and opens an authenticated Dashboard. If the probe fails, read the last Gateway error.

Send a Test Message

From the Dashboard, try:

Tell me which model you are using and create a short system status summary.

A complete response confirms the full setup. You can also test from Terminal:

openclaw agent --message "Hello, are you working?"

Then add a channel. The OpenClaw getting-started guide covers channels and first workflows.

How to Install an OpenClaw Skill

An OpenClaw Skill is a folder built around a SKILL.md file that tells the agent when and how to use a capability.

Before installing one, inspect its publisher, source, dependencies, API keys, network calls, and security scan:

openclaw skills verify @owner/skill-name

Install the Skill into the active workspace:

openclaw skills install @owner/skill-name

Add --global only when every local agent needs access. Run openclaw skills check, complete missing configuration, and test a reversible request before connecting important accounts.

The MyClaw Skills Hub shows categories, install counts, source information, and requirements. Start with one real need; the Google Workspace Skill, for example, connects Gmail, Calendar, Drive, Docs, and Sheets.

Launch OpenClaw in the Cloud with MyClaw

Local OpenClaw suits desktop access but stops being convenient when work must continue after sleep.

MyClaw provides a private, always-on environment with deployment, supported updates, monitoring, and backups handled. You choose the model, connected accounts, and permissions; model usage is billed separately.

Step 1: Launch Your Private OpenClaw

Choose the capacity and start an isolated instance without Node setup, WSL maintenance, Gateway exposure, or backup jobs.

Get Started

Step 2: Give It a Model and a Way to Reach You

Connect one model and one channel, then stop there until a test message works.

Step 3: Install One Skill and Give It a Real Job

Choose a Skill, configure it, and run a small test:

Every weekday at 8 AM, check these three competitors and send me a sourced change brief. Ask before taking any external action.

Review early results before expanding permissions or schedules.

Fix Common OpenClaw Installation Problems

Problem	Likely Cause	First Check
Command not found	Binary missing from PATH	Check node -v and npm prefix -g
Unsupported Node	Manual install uses an old runtime	Update Node and reinstall
npm/pnpm permission error	Incorrect global prefix	Correct the prefix; avoid sudo
Dashboard unavailable	Gateway stopped or unhealthy	Run openclaw gateway status
Port already in use	Another process owns the port	Stop it or run openclaw gateway --port 18790
Windows setup fails	WSL, virtualization, network, or pairing	Open Hub diagnostics
Mac tool fails	Permission or node issue	Check permissions and node status
Skill unavailable	Missing dependency, key, config, or access	Run openclaw skills check
Linux source build fails	Build tools missing	Run sudo apt install build-essential

Start with openclaw doctor. Fix the first blocking error before considering a reinstall.

Update OpenClaw

For a global installation, switch to the current stable release:

openclaw update --channel stable

For a source checkout, pull the latest code, reinstall packages, and rebuild:

git pull origin main

pnpm install && pnpm build && pnpm ui:build

Run openclaw doctor and check the Gateway again after either update.

Conclusion: Install OpenClaw for the Way You Plan to Use It

The best way to install OpenClaw depends on what happens next. Windows Hub and the macOS app offer the simplest local start. The installer script suits terminal users, Linux, and WSL2. Choose dedicated hardware, a VPS, or managed hosting when the agent must stay online.

Whichever path you choose, finish with the same proof: a healthy Gateway, a working model, and one useful task completed end to end.