Getting Started with OpenClaw: Install to First Agent

OpenClaw is an open-source personal AI assistant that runs locally, supports 12+ model providers, and connects to your messaging apps. Getting it running takes less than 10 minutes if you know the steps. This guide walks you through every one of them.

By the end of this tutorial, you will have a working AI assistant that you can message through WhatsApp, Telegram, or your preferred messaging platform, powered by whichever AI model you choose.

If you want background on what OpenClaw is and how it fits into the AI ecosystem before diving into installation, start with our overview of OpenClaw. For a comparison with other agent tools, see OpenClaw vs Claude Code.

Key Takeaways

• Installation requires Node.js 18+ and a single npm command to get started
• The onboarding wizard guides you through model provider selection, API key setup, and messaging channel configuration
• You choose your model provider from 12+ options including Claude, GPT, Grok, and local models
• Messaging channel setup connects OpenClaw to WhatsApp, Telegram, Discord, Slack, or other platforms
• Workspace files in ~/.openclaw/workspace/ control your assistant's personality, tools, and memory

Prerequisites

Before installing OpenClaw, make sure you have:

• Node.js 18 or higher installed on your machine. Check with node --version
• npm (comes with Node.js). Check with npm --version
• An API key from at least one AI model provider (Anthropic, OpenAI, etc.)
• A messaging account you want to connect (WhatsApp, Telegram, Discord, etc.)

If you don't have Node.js installed:

# macOS (with Homebrew)
brew install node

# Ubuntu/Debian
curl -fsSL https://deb.nodesource.com/setup_18.x | sudo -E bash -
sudo apt-get install -y nodejs

# Windows (with winget)
winget install OpenJS.NodeJS

Step 1: Install OpenClaw

Open your terminal and run:

npm i -g openclaw

This installs OpenClaw globally, making the openclaw command available system-wide. The installation pulls the core runtime and its dependencies. On most systems, this takes under a minute.

Verify the installation:

openclaw --version

You should see the version number printed. If you get a "command not found" error, ensure your npm global bin directory is in your PATH.

Step 2: Run the Onboarding Wizard

Start the interactive setup:

openclaw onboard

The onboarding wizard walks you through four decisions:

1. Choose your AI model provider
2. Enter your API key
3. Select your messaging channel
4. Configure channel-specific settings

Let's walk through each one.

Step 3: Choose Your Model Provider

The wizard presents your model provider options. Here are the most common choices:

For your first setup, Anthropic (Claude) or OpenAI (GPT) are the safest choices. Both have excellent instruction-following for skill execution, well-documented APIs, and predictable pricing.

Select your provider in the wizard:

? Select your AI model provider:
> Anthropic (Claude)
  OpenAI (GPT)
  xAI (Grok)
  Google (Gemini)
  Moonshot (Kimi)
  Local (Ollama)
  ...

Step 4: Configure Your API Key

After selecting a provider, the wizard prompts for your API key:

? Enter your Anthropic API key: sk-ant-...

Where to get API keys:

• Anthropic: console.anthropic.com under API Keys
• OpenAI: platform.openai.com under API Keys
• Google: ai.google.dev under Get API Key

The key is stored locally in your OpenClaw configuration. It never leaves your machine except when making inference requests to your chosen provider.

Step 5: Select Your Messaging Channel

This is where OpenClaw differentiates from terminal-based AI tools. You choose how you want to interact with your assistant:

? Select your messaging channel:
> WhatsApp
  Telegram
  Discord
  Slack
  Signal
  iMessage
  Google Chat
  Matrix

Each channel has its own setup requirements. Here are the three most common:

WhatsApp Setup

WhatsApp uses the Baileys library to connect without an official API. The setup involves QR code pairing:

1. Select WhatsApp in the wizard
2. A QR code appears in your terminal
3. Open WhatsApp on your phone
4. Go to Linked Devices and scan the QR code
5. OpenClaw connects as a linked device

Important: WhatsApp's two-phone architecture means OpenClaw runs as a linked device on your account. Messages to your OpenClaw number (or a dedicated number) are processed by the assistant. Configure the allowFrom list to control who can interact with it.

Telegram Setup

Telegram uses the official Bot API:

1. Message @BotFather on Telegram
2. Create a new bot with /newbot
3. Copy the bot token
4. Paste the token into the OpenClaw wizard

Your bot appears as a Telegram contact. Anyone who messages your bot gets AI responses from OpenClaw.

Discord Setup

Discord uses a bot token:

1. Go to the Discord Developer Portal
2. Create a new application
3. Under Bot, create a bot and copy the token
4. Paste the token into the OpenClaw wizard
5. Invite the bot to your server using the generated OAuth2 URL

Step 6: Send Your First Message

Once the onboarding wizard completes, OpenClaw starts running:

OpenClaw is running!
Connected to: WhatsApp
Model provider: Anthropic (Claude)
Workspace: ~/.openclaw/workspace/

Waiting for messages...

Open your configured messaging app and send a message to your OpenClaw assistant:

You: What can you help me with?

OpenClaw: I'm your personal AI assistant. I can help with:
- Answering questions and research
- Writing and editing text
- Analyzing documents and data
- Managing tasks and reminders
- Running installed skills from ClawHub
- And much more. What would you like to start with?

Congratulations. You have a working AI assistant connected to your messaging platform.

Step 7: Explore Workspace Files

OpenClaw's behavior is controlled by markdown files in ~/.openclaw/workspace/. Understanding these files lets you customize your assistant:

ls ~/.openclaw/workspace/

You should see:

AGENTS.md
SOUL.md
TOOLS.md
IDENTITY.md
USER.md
HEARTBEAT.md
MEMORY.md

SOUL.md is the most impactful file for immediate customization. It controls tone, personality, and behavioral guidelines:

# Soul

## Personality
You are a direct, efficient assistant. Avoid unnecessary pleasantries.
Get to the point quickly. Use bullet points for lists.

## Communication Style
- Concise responses unless detail is requested
- Technical accuracy over simplification
- Proactive suggestions when relevant

## Boundaries
- Never share personal information about the user
- Always clarify before taking irreversible actions

Edit this file to shape how your assistant communicates. Changes take effect immediately without restarting OpenClaw.

USER.md stores information about you that the assistant uses for context:

# User Profile

## Name
Jason

## Preferences
- Prefers morning briefings at 7 AM
- Works in software development
- Uses metric system

Step 8: Install Your First Skill

With OpenClaw running, you can expand its capabilities by installing skills from ClawHub:

You: install skill daily-digest

OpenClaw: Installing daily-digest v3.2...
Skill installed successfully.

daily-digest: Generates a morning briefing with calendar events,
weather, news highlights, and task priorities.

Try it: "Give me my daily digest"

Skills install in seconds and are immediately available across all your messaging channels. Browse the ClawHub registry for skills that match your workflow.

Step 9: Configure Security

Before using OpenClaw with others, configure access controls:

In your workspace configuration, set the allowFrom list to specify which contacts or accounts can interact with your assistant:

# Security

## Allowed Contacts
- +1234567890 (me)
- +0987654321 (partner)

## Blocked Actions
- No financial transactions
- No message deletion
- Confirm before sending messages to new contacts

This prevents unauthorized users from accessing your AI assistant through your messaging channels.

Next Steps

With OpenClaw running, here are productive next steps:

1. Customize SOUL.md to match your preferred communication style
2. Browse ClawHub for skills relevant to your work
3. Try different model providers to compare response quality and speed
4. Set up HEARTBEAT.md for periodic check-ins and automated briefings
5. Connect additional channels if you use multiple messaging platforms

Frequently Asked Questions

Do I need to keep my computer running for OpenClaw to work? Yes, if self-hosting. OpenClaw runs as a local process. When your machine sleeps or shuts down, the assistant goes offline. For always-on availability, use Agent37 managed hosting or run OpenClaw on a home server or cloud VM.

Can I switch model providers after setup? Yes. Model provider configuration can be changed at any time without losing your conversation history or installed skills. Update the configuration and OpenClaw will use the new provider for subsequent requests.

How much does it cost to run OpenClaw? OpenClaw itself is free. Your costs are API usage from your model provider. Light personal use typically costs $5-20/month with cloud providers. Local models eliminate API costs entirely but require hardware with sufficient RAM and compute.

Can multiple people use the same OpenClaw instance? Yes, through the allowFrom security configuration. Each authorized contact can message the assistant through the configured messaging channel. All users share the same skills and configuration.

What happens to my data if I stop using OpenClaw? Everything is local. Your workspace files, conversation history, and installed skills live on your machine. Uninstalling OpenClaw removes the runtime but leaves your data in ~/.openclaw/ unless you manually delete it.

From Installation to Productivity

The gap between installing a tool and getting value from it determines adoption. OpenClaw compresses this gap through a well-designed onboarding wizard that handles the four critical decisions (model, API key, channel, security) in a guided flow. Within minutes, you go from a terminal prompt to a working AI assistant in your messaging app.

The real value compounds over time as you customize workspace files, install skills from ClawHub, and let the assistant learn your patterns through MEMORY.md. The first message is just the starting point.

Explore production-ready AI skills at aiskill.market/browse or submit your own skill to the marketplace.