How to Install OpenClaw: Complete A-Z Guide for Beginners

OpenClaw is a free, open-source AI assistant built by Peter Steinberger and a global community of developers — source code publicly available at github.com/openclaw/openclaw. It turns the messaging apps you already use every day — Telegram, WhatsApp, Slack, Discord, and 20+ others — into a place where you delegate work to AI: writing content, analyzing data, automating workflows, managing communities, and much more.

This guide walks you through 4 ways to install OpenClaw — from fastest (1 command) to zero-install (cloud). Each step explains what you’re doing, what you need, and how to know it worked.

New to OpenClaw? Read What is OpenClaw? The Complete Guide to understand what OpenClaw can do before installing.

---

Choose the right installation method

Not everyone needs the same path. This table helps you choose quickly:

Method	Time	Technical knowledge needed	Best for
Method 1: One-liner (auto-install script)	5–10 min	Basic Terminal familiarity	Anyone on macOS/Linux who wants the fastest path
Method 2: npm global install	5–10 min	Basic Terminal familiarity	People already familiar with Node.js
Method 3: From source code	15–30 min	Programming experience	Developers who want to read/modify the code
Method 4: TryOpenClaw.io (cloud)	2 min	None	Everyone — especially if you’d rather not install anything

Not sure which to pick? On macOS or Linux, go with Method 1 (one-liner). If you don’t want to install anything at all, pick Method 4 (TryOpenClaw cloud).

---

Before you start: Understanding OpenClaw’s architecture (1-minute read)

OpenClaw has 3 main components working together — knowing this helps you understand what each installation step is setting up, and makes debugging faster if something goes wrong:

Gateway — OpenClaw’s coordination hub. It runs in the background on your machine (like a background service), receives messages you send from Telegram/Slack/Discord, routes them to the AI model for processing, and returns results. The Gateway runs at ws://127.0.0.1:18789 internally — you don’t need to remember this, just know it runs on your machine.

AI Model (LLM) — the brain that processes requests. OpenClaw doesn’t generate answers on its own — it sends your requests to an AI model like Claude (Anthropic), GPT (OpenAI), or Gemini (Google). You need an API key from one of these providers, or you can run a model locally on your own machine via Ollama.

Channel — where you talk to OpenClaw. Instead of opening a separate app, you give instructions through your familiar messaging apps: WhatsApp, Telegram, Slack, Discord, Signal, iMessage, and 20+ others. OpenClaw also has its own web interface (WebChat) and companion apps for macOS, iOS, and Android.

The flow: You send a message → Channel delivers it to Gateway → Gateway calls the AI Model → Result comes back to Channel.

Learn more: OpenClaw Features: Everything You Need to Know — voice, Canvas, Skills, browser control, automation, and more.

---

System requirements

Here’s what your machine needs. If you choose Method 4 (TryOpenClaw cloud), you don’t need to worry about this section — everything is already set up on the server.

Component	Requirement	Notes
Operating system	macOS, Linux, or Windows (via WSL2)	Windows requires WSL2 — see the Windows guide
Node.js	Version 24 (recommended) or 22.16+	The JavaScript runtime OpenClaw needs to run
RAM	2 GB minimum, 4 GB+ recommended	On low-RAM VPS, you can add swap — see the Linux guide
Disk space	5 GB free	SSD provides faster experience
Internet	Required	OpenClaw needs to call the AI model’s API

Quick check: Open Terminal (macOS/Linux) or PowerShell (Windows) and run:

node -v

If you see v22.x.x or v24.x.x → you’re ready. If you get command not found or a version below 22, you need to install Node.js first — see the “Install Node.js” section below.

---

Install Node.js (if you don’t have it)

Node.js is the runtime OpenClaw needs. If node -v already returned v22+ above, skip this section.

macOS / Linux:

The easiest way is nvm (Node Version Manager):

curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.3/install.sh | bash

Close Terminal, reopen it, then run:

nvm install 24

Verify: Run node -v — it should show v24.x.x.

Windows:

Download Node.js directly from nodejs.org/en/download (choose LTS) and install it like any other program. Or if you have WSL2, use the macOS/Linux method above inside your WSL terminal.

Platform-specific guides: Mac · Windows · Linux

---

Method 1: Auto-install script (One-liner) — Recommended

This is the officially recommended installation method (see the README on GitHub). A single command handles everything: downloading OpenClaw, installing it, and setting up the daemon (a background program that starts automatically when your machine boots).

Step 1: Run the install command

macOS / Linux / WSL2 — open Terminal and run:

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

Windows (PowerShell) — open PowerShell as Administrator and run:

iex (irm https://openclaw.ai/install.ps1)

The script automatically downloads and installs OpenClaw. You’ll see progress output on screen. Wait until you see a success message.

How to verify this step worked: After the command finishes, run openclaw --version — if it shows a version number (e.g. v2026.4.8), OpenClaw is installed.

Step 2: Run the Onboard Wizard

This is the most important step — OpenClaw includes an interactive wizard that sets up everything you need:

openclaw onboard --install-daemon

The wizard walks you through four things:

1. Install the Gateway daemon — the background program that keeps OpenClaw always ready, even when Terminal is closed. The daemon is installed as a LaunchAgent (macOS), systemd service (Linux), or Task Scheduler entry (Windows).

1. Choose your AI model — you’ll select Claude, GPT, Gemini, or another model. The wizard prompts you for the corresponding API key. (Don’t have one yet? See the API Key Configuration Guide — takes 5 minutes to set up.)

1. Connect a channel — choose which messaging app you want to use to talk to OpenClaw (Telegram, WhatsApp, Slack, Discord…). The wizard guides you through getting a token from your chosen app and connecting it.

1. Configure your Workspace — the wizard creates ~/.openclaw/workspace with the main config files: AGENTS.md (how the agent should behave), SOUL.md (the agent’s personality), TOOLS.md (which tools the agent can use). You can edit these files later.

How to verify this step worked: After the wizard finishes, run:

openclaw doctor

openclaw doctor checks all components (Gateway, model, channel) and reports the status of each. If everything shows “OK” or a green checkmark, you’re ready to go.

Step 3: Give your first task

Open the messaging app you connected (e.g. Telegram) and send the OpenClaw bot a simple message:

“Summarize the top 3 tech news stories today”

If the bot replies with reasonable content — congratulations, OpenClaw is installed and working.

Security note: By default, OpenClaw requires authentication when a new user messages the bot. The sender receives a pairing code — you approve it with openclaw pairing approve <channel> <code>. This prevents others from using (and spending credits on) your bot.

---

Method 2: npm/pnpm/bun (Package Manager)

If you already work with Node.js package managers, this approach will feel familiar:

Step 1: Install OpenClaw globally

Choose your package manager:

npm (most common):

npm install -g openclaw@latest

pnpm:

pnpm add -g openclaw@latest

bun:

bun add -g openclaw@latest

Verify: Run openclaw --version — a version number means it worked.

Step 2: Run the Onboard Wizard

Same as Method 1:

openclaw onboard --install-daemon

The wizard walks you through daemon setup, model selection, channel connection, and workspace configuration. See Method 1 — Step 2 above for full details on each step.

Verify: openclaw doctor shows all OK.

Getting a sharp build error? Some systems need to compile native libraries. See OpenClaw Troubleshooting Guide for a quick fix.

---

Method 3: From Source Code (for Developers)

This method is for people who want to read the source code, contribute to the project, or run the latest development build.

Step 1: Clone the repository

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

Official repository: github.com/openclaw/openclaw — this is also where you can follow new releases, report bugs, and submit contributions.

Verify: The openclaw directory exists and you’re inside it. Running ls should show files like package.json and README.md.

Step 2: Install dependencies and build

OpenClaw uses pnpm as its primary package manager when building from source:

# Install pnpm if you don't have it
npm install -g pnpm

# Install dependencies + build
pnpm install && pnpm ui:build && pnpm build

This downloads all required libraries and compiles the source code. It may take 2–5 minutes depending on your internet speed.

Verify: No ERR! or error lines in the output. Terminal returns to a normal prompt.

Step 3: Start the Gateway

pnpm gateway:watch

This starts the Gateway in development mode — it automatically reloads when you change code.

Verify: Terminal shows Gateway logs without red errors.

Step 4: Configure and connect

From here, use openclaw onboard (without --install-daemon since you’re running manually) or edit ~/.openclaw/openclaw.json directly to configure your model and channel. See the API Key Configuration Guide for the full list of options.

---

Method 4: Use TryOpenClaw.io — zero installation

If you don’t want to open Terminal, install Node.js, or manage a server — TryOpenClaw.io runs OpenClaw on cloud infrastructure for you. Everything is pre-configured: OpenClaw, Gateway, security, backups, auto-updates.

Step 1: Create an account

Go to app.tryopenclaw.io/auth/register and sign up with your email.

Step 2: Choose a plan

Plan	Price	Best for
Basic	From $4/month	Individuals, trying it out
Standard	Mid-tier	Freelancers, daily work
Premium	Higher tier	Small teams, heavy workflows

All plans include: isolated cloud instance, encrypted data, daily backups, and automatic crash recovery.

Step 3: Choose your AI model and launch

After payment, the system automatically deploys your OpenClaw instance in 1–2 minutes. Go to Dashboard → click “Configure Instance” → choose a model (Claude or GPT) → paste your API key or use TryOpenClaw credits directly (no separate API key needed) → click Deploy.

Verify: Dashboard shows status “Running”. You can start giving tasks immediately.

Step 4: Connect a channel

Go to Dashboard → Channels → choose your channel (Telegram, Discord, Slack…) → follow the prompts to paste your token. Test by sending a message to the bot.

Full guide: Run OpenClaw on Cloud — No Installation Needed ·

---

Advanced installation options

Beyond the 4 main methods above, OpenClaw supports several other deployment patterns for specific needs:

Docker / Podman — run OpenClaw in a container, completely isolated from your system. Good if you’re comfortable with Docker or deploying to production servers. See the GitHub README.

Nix — declarative configuration, ideal for NixOS users or anyone who wants reproducible builds.

Ansible — automate installation across multiple servers simultaneously.

Remote Gateway — run the Gateway on a Linux server, connect to it from another machine via Tailscale or SSH tunnel. Good if you want the Gateway running 24/7 on a VPS while you control it from your laptop.

These options are for users with technical experience. Full list in the GitHub README. If you’re just getting started, stick to the 4 methods above.

---

After installing: What to do next

Installation is just the beginning. To get real value from OpenClaw, follow this progression:

Right now — test a simple task: Send a message: “Summarize the content of [any URL]” or “Give me 3 ideas for a social media post about [topic]”. The goal is to confirm OpenClaw responds correctly.

Days 1–3 — install Skills: Skills are modules that give OpenClaw specialized abilities (SEO writing, code review, data analysis, and more). Browse and install from ClawHub with openclaw skills install <name>, or through the Dashboard on TryOpenClaw.

Week 2+ — build automated workflows: Combine tasks into recurring workflows using cron jobs, webhook triggers, or chained task sequences. Example: every weekday at 8am, OpenClaw auto-summarizes your inbox → analyzes it → sends a briefing to Telegram.

Explore Skills: OpenClaw Skills Complete Guide · Best OpenClaw Skills to Install

---

Updating and maintenance

Update OpenClaw: When a new version is available:

openclaw update

You can choose an update channel: openclaw update --channel stable (stable), beta (new features, may have bugs), or dev (latest development build). Follow changelogs at github.com/openclaw/openclaw/releases.

On TryOpenClaw, just click “Upgrade” in the Dashboard — no commands needed.

Health check: Whenever something feels off:

openclaw doctor

This checks the Gateway, model connection, and channel, and tells you exactly which component has a problem.

---

5 most common installation errors (and quick fixes)

1. openclaw: command not found after install Cause: Terminal hasn’t picked up the new PATH. Fix: close Terminal and reopen, or run source ~/.bashrc (Linux) / source ~/.zshrc (macOS).

2. node: command not found or Node.js version too old Cause: Node.js not installed or version below 22. Fix: install nvm then run nvm install 24 — see the “Install Node.js” section above.

3. API key not working (401 Unauthorized error) Cause: Wrong key, no credits, or extra whitespace in the key. Fix: check your key on the provider dashboard, verify you have credits, copy the key again without any leading/trailing spaces.

4. Port already in use (EADDRINUSE) Cause: Another app is using the default port. Fix: stop that app or change the port in OpenClaw’s config.

5. Permission error (EACCES) during global package install Cause: npm is trying to write to a system directory. Fix: use nvm (it manages permissions automatically) or change your npm prefix — don’t use sudo npm install.

Detailed fixes for 15+ errors: OpenClaw Troubleshooting Guide

---

Quick comparison: Self-hosted vs TryOpenClaw Cloud

Criteria	Self-hosted (Methods 1–3)	TryOpenClaw Cloud (Method 4)
Setup time	5–30 minutes	2 minutes
Technical knowledge needed	Basic (Terminal)	None
Infrastructure cost	VPS ~$5–20/month (for 24/7)	From $4/month
Uptime	Depends on you	99.9%, self-healing
Updates	Run openclaw update	1 click in Dashboard
Backups	You set them up	Daily, automatic
Full control	Yes	Limited to Dashboard

Self-hosted is right if you want complete control or deep customization. TryOpenClaw is right if you want to focus on using OpenClaw rather than running it.

Deep dive: OpenClaw Self-Hosted vs Cloud — Which Should You Choose?

---

Frequently Asked Questions

How long does it take to install OpenClaw? The fastest methods (one-liner or npm) take 5–10 minutes if you already have Node.js. TryOpenClaw cloud takes 2 minutes. Building from source takes 15–30 minutes.

Do I need to know programming to install OpenClaw? Methods 1 and 2 only require opening Terminal and copy-pasting commands. Method 4 (TryOpenClaw) requires no technical knowledge at all. Only Method 3 (source code) requires programming experience.

Does OpenClaw run on mobile? OpenClaw runs on your computer or a server, but you interact with it through messaging apps on your phone (Telegram, WhatsApp, etc.). OpenClaw also has companion apps for iOS and Android for notifications and voice control.

How much do API keys cost? Depends on the provider and model. Anthropic (Claude) typically gives $5 in free credits to new accounts — enough for hundreds of basic tasks. Or use TryOpenClaw with built-in credits, no separate API key required.

Does OpenClaw work offline? No, because it needs to call the AI model’s API. The exception is if you use a local model via Ollama — but that requires a machine with a decent GPU, and output quality is lower than Claude/GPT.

Is my data safe? When self-hosted, OpenClaw runs on your machine and your data doesn’t pass through any third party beyond your chosen AI provider. With TryOpenClaw, each user gets an isolated instance with encrypted data and daily backups.

---

Summary

There are 4 clear paths to installing OpenClaw — choose the one that fits your skills and needs:

1. One-liner script → fast, automatic, works for most people
2. npm/pnpm/bun → familiar for Node.js developers
3. From source → for developers who want to contribute or customize
4. TryOpenClaw.io → no installation, up and running in 2 minutes

The most important thing after installing: give your first task right away. Don’t try to optimize config or install every skill first — start simple, expand from there.

---