How to Install OpenClaw

System Requirements (RAM, Node.js, OS Versions)

OpenClaw needs Node.js, npm, a supported operating system, and access to at least one model. Hosted model providers usually keep local resource needs modest because inference runs outside your machine. Local models can require substantially more RAM, disk space, and GPU capacity depending on the model size.

Before installing, confirm Node and npm are available:

node -v
npm -v

If those commands fail, install Node.js first or use the official OpenClaw installer, which handles most setup steps. For team repositories, install OpenClaw as a normal user rather than root or administrator unless your IT policy requires managed installation.

Install on macOS Apple Silicon

Open Terminal on your Mac. The same installer works on Apple Silicon machines, including M1, M2, M3, and M4 systems.

Run the official installer:

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

After the installer finishes, start onboarding and install the local daemon:

openclaw onboard --install-daemon

If openclaw is not found after installation, close and reopen Terminal so your shell reloads its PATH. You can also check where global npm binaries are installed:

npm prefix -g
echo "$PATH"

Then verify the install:

openclaw --version
openclaw doctor

For repository work, run OpenClaw from the project directory so it can inspect the correct Git workspace, dependencies, and local configuration.

Install on macOS Intel

Use the same official installer on Intel-based Macs:

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

The installer detects the machine architecture, so you do not need a separate Intel package.

If you already manage Node with a version manager or a company-approved Node distribution, install OpenClaw with npm instead:

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

Check the install before connecting a repository:

openclaw --version
openclaw doctor
openclaw gateway status

If your Mac has multiple Node versions, make sure the shell used for OpenClaw is the same shell where node, npm, and the global npm binary path resolve correctly.

Install on Windows 11 (npm + openclaw-win-x64.exe)

The official documentation does not list a standalone openclaw-win-x64.exe installer. Use the official PowerShell installer or npm unless the OpenClaw documentation adds a Windows executable.

Open PowerShell as a normal user and run:

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

You can also install with npm if Node.js is already managed on your machine:

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

After installation, verify the CLI and Gateway:

openclaw --version
openclaw doctor
openclaw gateway status

If PowerShell blocks the installer, check your execution policy and company endpoint controls. Avoid running an elevated administrator shell unless your environment requires admin-managed installs.

WSL2 is often the better choice for repositories that depend on Linux tooling, shell scripts, or Unix-style paths. Native Windows is supported, but WSL2 can reduce path, permission, and shell differences for development teams.

Install on Linux / Ubuntu

Open a terminal and run the official installer:

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

Then install the user service:

openclaw onboard --install-daemon

Do not run the Gateway as root unless your deployment requires it. A normal user install keeps project files, credentials, shell configuration, and service ownership easier to audit.

Confirm Node, npm, OpenClaw, and the Gateway are available:

node -v
npm -v
openclaw --version
openclaw gateway status

On servers, headless workstations, or remote development boxes, make sure local firewall rules and port policies allow the dashboard and Gateway to run on the expected local address. If the machine uses systemd user services, log out and back in if the service does not appear immediately.

After confirming the Linux service is healthy, OpenClaw on Windows can help you mirror the same setup expectations on a Windows workstation.

For source-level validation, the official documentation is worth checking after you understand the How to Install OpenClaw workflow described here.

Verify Your Installation With openclaw doctor

Run verification immediately after installation and onboarding:

openclaw --version
openclaw doctor
openclaw gateway status
openclaw dashboard

The dashboard normally opens at:

http://127.0.0.1:18789/

A healthy install should confirm four things: the openclaw command resolves, the installed version prints, the Gateway is running, and the dashboard opens locally.

Treat verification as separate from installation. Installation places the CLI on the machine. Onboarding configures the daemon and model access. Verification confirms that the local service, shell path, and dashboard are ready before you connect OpenClaw to a real repository.

If openclaw doctor reports a model-provider issue, finish provider authentication before starting repository work. If the dashboard does not open, check Gateway status first, then confirm that no local process or security tool is blocking the dashboard port.

After local verification succeeds, OpenClaw Docker Setup helps you run the same environment in a containerized workflow.

When details such as limits or setup steps matter, Openclaw can help confirm the latest implementation surface.

Common Install Errors: npm ENOENT, zsh: command not found, Stack Overflow

openclaw: command not found

The shell cannot find the OpenClaw binary. Check the global npm path and your shell PATH:

npm prefix -g
echo "$PATH"

Close and reopen the terminal after installation. If the command still fails, add the npm binary directory to your shell startup file, such as .zshrc, .bashrc, or the equivalent profile used by your environment.

npm ENOENT

ENOENT usually means npm cannot find a file, directory, package path, or expected install location. Confirm Node and npm are installed:

node -v
npm -v

Then rerun the official installer or reinstall with npm:

npm install -g openclaw@latest

If the error continues, clear the npm cache only after checking that the failing path in the error message is safe to remove.

Stack overflow or repeated startup

A stack overflow or repeated startup loop can come from duplicate Gateway processes, a damaged configuration, or a model/provider setting that repeatedly fails during startup.

Check the Gateway first:

openclaw gateway status
openclaw doctor

Stop duplicate Gateway processes, review recent configuration changes, and restart onboarding if the model-provider setup is incomplete. If the problem started after an update, check release notes and confirm your Node version still matches the supported range.

After OpenClaw starts cleanly, the OpenClaw Skills Guide can help you verify that your installed setup is ready for reusable workflows.

Before you budget a real project around How to Install OpenClaw, compare the claims here with Youtube.

Next Step: Connect OpenClaw to a Model

After the CLI and Gateway pass verification, run onboarding:

openclaw onboard

Choose a provider, add the required credential, and select a model. Hosted providers usually require an API key. Local models require the model runtime, model files, and enough local resources for the selected model.

Before using OpenClaw on production code, test it in a small repository or a disposable branch. Confirm that it can read the project, understand the dependency setup, and produce changes that your normal test commands can validate.

For repository-scale development, Verdent adds coordination around OpenClaw-style agent work. It separates tasks into isolated worktrees, keeps parallel work from colliding, and gives teams a review path before changes reach the main branch.