Installation Guide

Get OpenClaw up and running on your system

This page covers every way to get OpenClaw onto your system. We recommend the one-liner—it installs Node.js (if needed) and all dependencies automatically. You can also use npm/pnpm, Docker, or build from source. Prefer a guided flow? The Setup Wizard gives you ready-to-run commands for your OS. Stuck in the first hour? See FAQ and Troubleshooting for common fixes.

Choose your path

Jump to the section that matches how you want to install:

Ollama (fastest for Mac/Linux)

If you use Ollama 0.17+, the simplest path is one command: Ollama installs OpenClaw (if needed), configures the model, and starts the gateway automatically.

One command
ollama launch openclaw --model kimi-k2.5:cloud

Requires Ollama and Node.js. Full step-by-step: Ollama + OpenClaw tutorial.

macOS / Linux

Recommended: The one-liner installs Node.js (if needed) and OpenClaw in one go. No need to install Node yourself.

One-liner (recommended)
curl -fsSL https://openclaw.ai/install.sh | bash
openclaw onboard

Alternative: npm

Terminal
npm i -g openclaw
openclaw onboard

Alternative: pnpm

Terminal
pnpm add -g openclaw
openclaw onboard

macOS Companion App: Download for macOS (menubar access, works alongside the CLI). Requires macOS 14+.

After the wizard, set your model and connect a channel: Configuration, Channel setup. For macOS-specific steps and launchd details: macOS platform guide. First-time walkthrough: Getting Started tutorial.

Right after install

If your agent replies but never reads files or runs commands, check tools.profile in config—it may be set to messaging (chat only). Set "tools": { "profile": "full" } (or "coding") so the agent can use file, exec, and browser tools. See Troubleshooting → No output. Before you change a lot of config, run openclaw backup create so you can restore if needed. Full backup options: Workspace Backup & Restore.

Windows

Recommended: Use the one-liner. PowerShell or CMD:

PowerShell
iwr -useb https://openclaw.ai/install.ps1 | iex
openclaw onboard
CMD
curl -fsSL https://openclaw.ai/install.cmd -o install.cmd && install.cmd && del install.cmd
openclaw onboard

Alternative: npm

Terminal
npm i -g openclaw
openclaw onboard

Alternative: pnpm

Terminal
pnpm add -g openclaw
openclaw onboard

Full Windows-specific notes (paths, services, GUI): Windows platform guide.

Docker (server, Unraid, cloud)

Running OpenClaw in Docker gives you an isolated environment on a server, Unraid box, or cloud VM. You get one container that runs the Gateway; you then run openclaw onboard inside the container (or exec in) to create the workspace and link channels. Data lives in a mounted volume so it persists across restarts.

Terminal
docker run -d --name openclaw \
  -v ~/.openclaw:/root/.openclaw \
  -p 18789:18789 \
  ghcr.io/openclaw/openclaw:latest

Then attach and run openclaw onboard. For Compose, Unraid, and persistent data layout: Docker deployment guide.

From source (hackable)

For those who read source code for fun. You get the same CLI and Gateway; useful for air-gapped or custom builds.

Option 1: Install script

Terminal
curl -fsSL https://openclaw.ai/install.sh | bash -s -- --install-method git
pnpm run openclaw onboard

Option 2: Clone and build

Terminal
git clone https://github.com/openclaw/openclaw.git
cd openclaw && pnpm install && pnpm run build
pnpm run openclaw onboard

Release (v2026.3.13): Release notes, Latest release. For live reload during development: pnpm gateway:watch. See the repo README for details.

Serverless (Moltworker)

Moltworker runs OpenClaw on Cloudflare Workers—no VM, no Docker, no long-lived server. You get pay-per-use and edge deployment. Setup and deployment: Moltworker guide.

Kimi Claw (kimi.com)

OpenClaw is also available natively on kimi.com as Kimi Claw—running in your browser, online 24/7. You get access to the ClawHub skill library, cloud storage, and pro-grade search (e.g. live data from Yahoo Finance). With Bring Your Own Claw, you can connect your self-hosted OpenClaw to kimi.com, chat with your existing setup, or bridge it to apps like Telegram groups. Beta access is open for Allegretto members and above: kimi.com/bot.

When installation doesn't work

npm install fails or command not found after install: Confirm Node.js 22.12.0+ (node -v). On some systems the global npm bin isn't on your PATH—restart the terminal or add that directory to PATH. Full checklist: Troubleshooting.

Docker: port or volume errors: Make sure port 18789 isn't in use and any mounted directory exists and is readable. Details: Docker deployment guide.

Wizard quits or channel linking fails: Run openclaw onboard again (you can redo just the channel step). If the Gateway misbehaves, openclaw doctor --fix and the Troubleshooting guide usually help.

Updating and release channels

To upgrade to the latest release, run openclaw update; then openclaw doctor to verify. Current latest: v2026.3.13. Use openclaw update --dry-run to preview what would be updated without changing config or installing. You can switch channels (stable / beta / dev) with openclaw update --channel stable|beta|dev. The CLI name is openclaw; moltbot and clawdbot remain valid aliases.

After installation

Once OpenClaw is installed, configure a model and connect a channel so you can chat. Next steps:

First time? The Getting Started tutorial walks through install, config, and first message.

Don't want to self-host? Clawi.ai offers managed OpenClaw—sign up, connect a channel, and you're live. Kimi Claw on kimi.com runs OpenClaw in the browser with ClawHub access and Bring Your Own Claw (beta for Allegretto members).