OpenClaw Setup Guide: Step-by-Step

Setting up OpenClaw gives you a self-hosted AI assistant that runs on your own hardware, connects to the messaging apps you already use, and executes tasks autonomously. Unlike cloud-based chatbots that forget your context when you close the browser, OpenClaw runs continuously as a daemon process that remembers conversations, manages files, automates browser tasks, and can even message you proactively about scheduled reminders or detected events.

Quick Answer: To set up OpenClaw, install Node.js 22 or higher, run the one-line installer (curl -fsSL https://openclaw.ai/install.sh | bash), choose your LLM provider during onboarding, configure your preferred messaging channel (Telegram, WhatsApp, Discord, or Slack), and run the daemon with openclaw start --install-daemon. The entire process takes 10-30 minutes depending on your deployment choice.

What Is OpenClaw and Why Set It Up?

OpenClaw is an open-source agent gateway that connects large language models to your local system. Think of it as a bridge between AI capabilities and real-world actions. The framework gained over 247,000 GitHub stars within weeks of its January 2026 launch, making it one of the fastest-growing repositories in GitHub history.

Unlike traditional chatbots, OpenClaw operates as a persistent service. It can execute shell commands, manage your calendar, read and send emails, automate browser tasks, and integrate with over 50 messaging platforms. Because it runs on your hardware, you maintain complete control over your data—no third-party servers store your conversations or access your files.

The architecture supports any LLM provider. You can switch between Claude, GPT, local models via Ollama, or even custom fine-tuned models without rewriting code. This flexibility means you're not locked into a single vendor's pricing or capabilities.

OpenClaw shines when you need automation that understands context. Want your assistant to monitor your inbox for flight confirmations and automatically add them to your calendar? Or maybe you need a bot that checks GitHub for new issues every morning and sends you a digest on Telegram? These scenarios require persistent memory and the ability to take actions across multiple services—exactly what OpenClaw provides.

The MIT license and active community mean you can customize every aspect. Over 5,700 community-built skills extend functionality, and the Skills Marketplace offers integrations for everything from Notion databases to smart home devices. Understanding how the OpenClaw agent gateway works helps you make the most of these capabilities.

What Do You Need Before Installing OpenClaw?

Getting your prerequisites right saves hours of troubleshooting later. Here's what you need before starting the installation.

System Requirements

OpenClaw requires Node.js version 22 or newer. Many tutorials incorrectly mention Node 18 or 20, but the framework uses modern JavaScript features that only work with Node 22+. Using an older version causes cryptic syntax errors during dependency installation.

For your operating system, macOS and Linux work out of the box. Windows users should install WSL2 (Windows Subsystem for Linux 2) rather than running OpenClaw natively—the framework's shell execution and daemon management features assume a Unix-like environment. Installing under native Windows PowerShell leads to channel configuration failures and daemon startup problems.

Memory requirements depend on your use case. For basic chat and file operations, 4 GB of RAM suffices. If you plan to run local LLMs via Ollama, budget at least 8 GB—preferably 16 GB for larger models. Running OpenClaw on a miniPC offers a dedicated, always-on solution without tying up your main workstation. Check out OpenClaw miniPC deployment options for hardware recommendations.

LLM Provider Account

You need access to at least one language model. Cloud options include:

• Anthropic Claude: Known for strong reasoning and long context windows. API keys cost $0.25-$15 per million tokens depending on the model (Claude 3.5 Sonnet, Opus, or Haiku).
• OpenAI GPT: Wide adoption and extensive documentation. Pricing ranges from $0.15 per million tokens (GPT-3.5) to $30 per million tokens (GPT-4 Turbo).
• Google Gemini: Competitive pricing with strong multimodal capabilities.

For privacy-focused setups or to avoid per-token costs, run local models:

• Ollama: Packages open-source models like Llama 3, Mistral, and Qwen for local execution. Requires a GPU with at least 8 GB VRAM for 7B parameter models.
• vLLM: High-performance inference server for self-hosted models.

Create your API key before starting installation—the onboarding wizard will prompt for it immediately, and pausing mid-setup to register for a provider account breaks the flow.

Messaging Platform Token

OpenClaw communicates through messaging apps. Each platform requires setup:

• Telegram: Create a bot via BotFather and obtain the HTTP API token. Takes 2 minutes.
• Discord: Register an application in Discord Developer Portal, create a bot user, and copy the token.
• WhatsApp: Requires phone number verification through the WhatsApp Business API. More complex than Telegram but integrates with your existing number.
• Slack: Create a Slack app and generate a bot token with appropriate permissions.

You can skip messaging configuration initially and add channels later, but configuring at least one channel during setup lets you test functionality immediately.

VPS Account (Optional but Recommended)

Running OpenClaw on your laptop works for testing, but production deployments benefit from dedicated hosting. A Virtual Private Server ensures your assistant stays online 24/7 even when your personal computer is off.

Recommended specs for VPS hosting:

• CPU: 2 cores minimum
• RAM: 4-8 GB
• Storage: 20 GB SSD
• Providers: DigitalOcean, Linode, Vultr, or AWS Lightsail

DigitalOcean offers a one-click OpenClaw deployment image that pre-installs dependencies and configures the environment. This shortcut collapses VPS setup from 30 minutes of manual configuration to 5 minutes of button clicks.

How Do You Install OpenClaw Step by Step?

The installation process has three main paths: quick start via installer script, manual npm installation, or Docker deployment. We'll cover all three, starting with the recommended method.

Method 1: One-Line Installer (Recommended)

This approach works on macOS, Linux, and Windows with WSL2.

Step 1: Open your terminal

• macOS: Press Cmd+Space, type "Terminal," and press Enter
• Linux: Press Ctrl+Alt+T or search for Terminal in your application menu
• Windows: Open PowerShell and run wsl to enter your WSL2 environment

Step 2: Run the installer

Paste this command and press Enter:

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

The script detects your Node.js version. If Node 22+ isn't installed, it automatically downloads and installs it. This takes 2-5 minutes depending on your connection speed.

Step 3: Navigate the onboarding wizard

After installation completes, the wizard launches automatically. You'll see a text-based interface with arrow key navigation.

First prompt: "I understand this is powerful and inherently risky."

Select "Yes" with arrow keys and press Enter. This warning acknowledges that giving an AI agent terminal access and file system permissions carries security implications.

Step 4: Choose QuickStart

The installer offers QuickStart or Manual configuration. Choose QuickStart—it handles gateway pairing, personality settings, and authentication through one unified interface.

Step 5: Select your LLM provider

Arrow down to your chosen provider (Anthropic, OpenAI, or Local Model) and press Enter. Paste your API key when prompted. The wizard validates the key by making a test API call. If validation fails, double-check for typos—API keys are long strings that are easy to copy incorrectly.

Step 6: Configure messaging channel

Select your primary channel (Telegram recommended for simplest setup). Paste your bot token when prompted. The wizard automatically configures webhooks and tests connectivity.

Step 7: Install as daemon

The final prompt asks: "Install OpenClaw as a background service?"

Choose "Yes" to register OpenClaw with your system's service manager:

• Linux: Creates a systemd service
• macOS: Registers with launchd
• Windows (WSL2): Configures systemd within WSL

This ensures OpenClaw starts automatically when your system boots and survives reboots.

Step 8: Verify installation

Run these commands to confirm everything works:

openclaw status

You should see: "Gateway: Running" and "Channels: 1 connected"

openclaw gateway status

Shows gateway URL (typically http://localhost:8174) and authentication status.

Method 2: Manual npm Installation

For advanced users who want more control over the installation process.

Step 1: Install Node.js 22+

Download from nodejs.org or use a version manager:

# Using nvm
nvm install 22
nvm use 22

Step 2: Install OpenClaw globally

npm install -g clawdbot

Or with pnpm:

pnpm add -g clawdbot

Step 3: Run onboarding manually

openclaw onboard

Follow the same wizard steps as Method 1.

Step 4: Start the gateway

openclaw start --install-daemon

Method 3: Docker Deployment

Docker provides isolation and simplifies dependency management.

Step 1: Clone the repository

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

Step 2: Build the Docker image

docker build -t openclaw:latest .

Step 3: Create volume mounts

mkdir -p ~/.openclaw ~/openclaw/workspace

These directories store configuration (/.openclaw) and agent workspace (/openclaw/workspace).

Step 4: Run the container

docker run -d \
  --name openclaw \
  -v ~/.openclaw:/root/.openclaw \
  -v ~/openclaw/workspace:/workspace \
  -p 8174:8174 \
  openclaw:latest

Step 5: Access the container for onboarding

docker exec -it openclaw openclaw onboard

Complete the wizard inside the container.

How Do You Choose and Configure Your LLM Provider?

Your choice of language model affects cost, performance, privacy, and capabilities. Here's how to evaluate options and configure them properly.

Cost Analysis

Cloud LLMs charge per token:

Provider	Model	Input Cost	Output Cost	Context Window
Anthropic	Claude 3.5 Sonnet	$3/M tokens	$15/M tokens	200K tokens
OpenAI	GPT-4 Turbo	$10/M tokens	$30/M tokens	128K tokens
OpenAI	GPT-3.5 Turbo	$0.50/M tokens	$1.50/M tokens	16K tokens
Google	Gemini 1.5 Pro	$1.25/M tokens	$5/M tokens	1M tokens

For comparison, a typical automation task uses 500-2,000 tokens. If your agent runs 100 tasks per day with Claude 3.5 Sonnet, expect $5-15/month in API costs.

Local models eliminate per-token fees but require upfront hardware investment. A GPU capable of running 7B parameter models (like Llama 3) costs $300-500. Larger models need more VRAM—a 13B model requires 16 GB, while 70B models need 48+ GB or multi-GPU setups.

Configuration Steps

For cloud providers:

Edit ~/.openclaw/config/openclaw.json:

{
  "llm": {
    "provider": "anthropic",
    "model": "claude-3-5-sonnet-20241022",
    "apiKey": "sk-ant-your-key-here",
    "maxTokens": 4096,
    "temperature": 0.7
  }
}

Adjust temperature based on use case:

• 0.3-0.5: Deterministic responses for automation and data extraction
• 0.7-0.9: Creative tasks and conversational interactions

For local models:

Install Ollama first:

curl -fsSL https://ollama.ai/install.sh | sh

Pull a model:

ollama pull llama3

Configure OpenClaw to use the local endpoint:

{
  "llm": {
    "provider": "ollama",
    "model": "llama3",
    "endpoint": "http://localhost:11434"
  }
}

Restart the gateway after configuration changes:

openclaw restart

Multi-Model Setup

Advanced OpenClaw routing lets you direct different prompts to different models. For example, use GPT-4 for complex reasoning tasks but GPT-3.5 for simple acknowledgments and confirmations.

Configure routing rules in openclaw.json:

{
  "routing": {
    "rules": [
      {
        "pattern": "summarize|analyze|explain",
        "provider": "anthropic",
        "model": "claude-3-opus"
      },
      {
        "pattern": "*",
        "provider": "openai",
        "model": "gpt-3.5-turbo"
      }
    ]
  }
}

This configuration routes analytical prompts to Claude Opus while defaulting to GPT-3.5 for everything else, optimizing for both quality and cost.

How Do You Connect OpenClaw to Messaging Apps?

Messaging integration transforms OpenClaw from a command-line tool into an accessible assistant you can reach from your phone or any device with a messaging app.

Telegram Setup (Easiest)

Step 1: Create a bot

Open Telegram and message @BotFather. Send /newbot and follow prompts to name your bot. BotFather responds with an HTTP API token:

Use this token to access the HTTP API:
1234567890:ABCdefGHIjklMNOpqrsTUVwxyz

Step 2: Configure OpenClaw

Run:

openclaw configure --section channels

Select "Add Telegram channel" and paste your token. The wizard automatically sets up webhooks.

Step 3: Test connection

Message your bot on Telegram. It should respond within seconds. If you see "Bot not responding," check gateway status:

openclaw gateway status
openclaw logs --follow

Common issue: Firewall blocking webhook callbacks. If running on VPS, ensure port 8174 allows inbound connections.

Discord Setup

Step 1: Create Discord application

Visit discord.com/developers/applications and click "New Application." Name it and save.

Step 2: Add bot user

Navigate to "Bot" section and click "Add Bot." Under "Token," click "Copy" to get your bot token.

Step 3: Configure permissions

In "OAuth2" section, select "bot" scope and check these permissions:

• Read Messages/View Channels
• Send Messages
• Embed Links
• Attach Files

Copy the generated URL and visit it to add the bot to your server.

Step 4: Add to OpenClaw

openclaw configure --section channels

Select Discord and paste your token.

WhatsApp Business API

WhatsApp integration is more complex because it requires phone number verification and uses the WhatsApp Business Platform.

Step 1: Register for WhatsApp Business API

Visit developers.facebook.com/docs/whatsapp and create a business app. This requires:

• Facebook Business Manager account
• Phone number dedicated to the bot (cannot be used with regular WhatsApp)
• Business verification

Step 2: Obtain credentials

After approval, you'll receive:

• Phone number ID
• WhatsApp Business Account ID
• Access token

Step 3: Configure OpenClaw

Edit ~/.openclaw/config/channels.json:

{
  "whatsapp": {
    "phoneNumberId": "your-phone-id",
    "accessToken": "your-access-token",
    "webhookVerifyToken": "random-secure-string"
  }
}

Step 4: Set up webhook

WhatsApp requires a publicly accessible HTTPS endpoint. Use ngrok for testing:

ngrok http 8174

Copy the HTTPS URL and add /webhooks/whatsapp to configure in Facebook App Dashboard.

Production deployments need proper SSL certificates via Let's Encrypt or a reverse proxy like nginx.

How Do You Secure Your OpenClaw Installation?

Security matters because OpenClaw has extensive permissions—it can read files, execute commands, and access any service you configure. Here's how to lock it down.

Network Security

Bind to localhost only

Edit ~/.openclaw/config/openclaw.json:

{
  "gateway": {
    "host": "127.0.0.1",
    "port": 8174
  }
}

This prevents external network access. Never expose OpenClaw directly to the internet.

Use SSH tunneling for remote access

If you need to access the control UI from outside your network:

ssh -L 8174:localhost:8174 user@your-vps-ip

Then access via http://localhost:8174 on your local machine.

Configure firewall rules

On Linux with ufw:

sudo ufw default deny incoming
sudo ufw allow ssh
sudo ufw allow 8174/tcp from 127.0.0.1
sudo ufw enable

User Permissions

Never run as root

Running OpenClaw with root privileges gives the agent unrestricted system access. Always use a dedicated user account:

sudo useradd -m -s /bin/bash openclaw
sudo su - openclaw

Install and run OpenClaw as this user.

Limit file system access

Configure a restricted workspace in openclaw.json:

{
  "workspace": {
    "root": "/home/openclaw/workspace",
    "allowedPaths": [
      "/home/openclaw/workspace",
      "/home/openclaw/documents"
    ]
  }
}

The agent cannot access paths outside these directories.

Skills Marketplace Safety

The Skills Marketplace hosts over 5,700 community plugins, but not all are trustworthy. In January 2026, security researchers identified 341 malicious skills designed to exfiltrate credentials or establish backdoors.

Vet before installing:

Run OpenClaw's built-in scanner:

openclaw security audit --deep

This checks installed skills against VirusTotal and other threat databases.

Check skill source code:

Before installing, review the skill's repository:

openclaw skill inspect skill-name

Look for suspicious patterns:

• Network requests to unknown domains
• File operations outside workspace
• Credential access or environment variable reading

Prefer verified publishers:

The marketplace shows verification badges for vetted developers. Prioritize skills from verified sources.

Authentication and Access Control

Enable authentication:

openclaw configure --section auth

Set a strong password or configure OAuth2. The gateway UI becomes password-protected.

Rotate API keys regularly:

Change your LLM provider API keys every 90 days. Update in config:

openclaw configure --section llm

Monitor logs:

Watch for unusual activity:

openclaw logs --follow --level warn

Set up log shipping to a monitoring service for production deployments.

Security Patches

OpenClaw releases security updates frequently. The February 2026 patch (version 2026.2.21) fixed CVE-2026-25253, a critical prompt injection vulnerability.

Enable automatic updates:

{
  "updates": {
    "autoCheck": true,
    "autoInstall": "security-only"
  }
}

Or manually check:

openclaw update check
openclaw update install

Subscribe to the security mailing list at openclaw.ai/security for vulnerability announcements.

What Common Installation Errors Should You Expect?

Even following instructions carefully, you'll likely encounter at least one of these issues. Here's how to fix them quickly.

"openclaw: command not found"

Cause: npm's global bin directory isn't in your PATH.

Fix:

Find npm's global bin path:

npm config get prefix

Add to PATH by editing ~/.bashrc or ~/.zshrc:

export PATH="$PATH:/usr/local/bin"

Reload shell:

source ~/.bashrc

"Permission denied" During Installation

Cause: Insufficient permissions for npm global directory.

Wrong fix: Running sudo npm install -g (creates permission problems later)

Right fix:

Change npm directory ownership:

sudo chown -R $USER:$(id -gn $USER) ~/.npm
sudo chown -R $USER:$(id -gn $USER) /usr/local/lib/node_modules

Retry installation without sudo.

"Gateway failed to start"

Cause: Port 8174 already in use or Node.js version mismatch.

Diagnostic steps:

Check port availability:

lsof -i :8174

If another process uses the port, kill it or change OpenClaw's port in config.

Verify Node.js version:

node --version

Must show v22.x or higher. If not:

nvm install 22
nvm use 22

Reinstall OpenClaw after Node upgrade.

"API key invalid" Despite Correct Key

Cause: API key contains special characters that need escaping, or whitespace was accidentally copied.

Fix:

Regenerate the API key through your provider's dashboard. When pasting, ensure no leading/trailing spaces.

Test the key independently:

curl -H "Authorization: Bearer sk-your-key" https://api.anthropic.com/v1/messages

If this fails, the key itself is invalid.

"Channel connection failed"

Cause: Bot token incorrect, webhook URL inaccessible, or firewall blocking.

Fix:

Verify token by testing directly with the platform's API:

# Telegram
curl https://api.telegram.org/bot<YOUR_TOKEN>/getMe

If running on VPS, ensure webhook endpoint is publicly accessible:

curl -I http://your-vps-ip:8174/health

Should return 200 OK. If connection refused, check firewall rules.

Systematic Troubleshooting

Run this diagnostic sequence when something isn't working:

openclaw doctor

This command checks:

• Node.js version compatibility
• Configuration file syntax
• API key validity
• Gateway connectivity
• Channel status
• File permissions

Auto-fix common issues:

openclaw doctor --fix

For persistent problems, enable verbose logging:

openclaw start --log-level debug

Watch logs in real-time:

openclaw logs --follow

How Do You Set Up MCP Servers with OpenClaw?

Model Context Protocol (MCP) servers extend OpenClaw's capabilities by connecting it to external services and data sources. MCP is an open standard developed by Anthropic that defines how AI applications access tools and context.

Understanding MCP Architecture

MCP servers expose tools that your agent can call. For example:

• Google Drive MCP server: Provides tools for searching files, reading documents, and uploading content
• Database MCP server: Offers query execution and schema inspection
• Slack MCP server: Enables message sending, channel listing, and user lookup

Your OpenClaw agent doesn't need custom code for each service—it discovers available tools automatically through the MCP protocol.

Installing MCP Servers

Method 1: Via McPorter (Recommended)

McPorter acts as a package manager for MCP servers:

npm install -g mcporter

Discover available servers:

mcporter search google

Install a server:

mcporter install @modelcontextprotocol/server-gdrive

McPorter automatically configures the server in your openclaw.json.

Method 2: Manual Configuration

Edit ~/.openclaw/config/openclaw.json:

{
  "mcpServers": {
    "gdrive": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-gdrive"],
      "env": {
        "GDRIVE_CLIENT_ID": "your-client-id",
        "GDRIVE_CLIENT_SECRET": "your-client-secret"
      }
    }
  }
}

OAuth Configuration for MCP Servers

Many MCP servers require OAuth authentication. Here's the flow using Google Drive as an example:

Step 1: Create OAuth credentials

Visit console.cloud.google.com and create a new project. Enable Google Drive API and create OAuth 2.0 credentials.

Step 2: Configure redirect URI

Set redirect URI to: http://localhost:8174/oauth/callback

Step 3: Add credentials to MCP config

{
  "mcpServers": {
    "gdrive": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-gdrive"],
      "env": {
        "GDRIVE_CLIENT_ID": "123456.apps.googleusercontent.com",
        "GDRIVE_CLIENT_SECRET": "your-secret"
      }
    }
  }
}

Step 4: Complete OAuth flow

Restart gateway:

openclaw restart

Visit the control UI at http://localhost:8174. Navigate to "MCP Servers" and click "Authenticate" next to Google Drive. This opens a browser window for authorization.

Verifying MCP Server Connection

Check server status:

openclaw mcp list

Shows all configured servers and their connection state.

Test a tool:

Message your OpenClaw bot: "List my Google Drive files"

If configured correctly, the agent discovers the gdrive_list_files tool and calls it.

Debug connection issues:

openclaw logs --filter mcp

Common problems:

• Server not starting: Check that the command and args are correct
• Authentication failed: Verify OAuth credentials and redirect URI
• Tools not discovered: Restart gateway after config changes

Popular MCP Servers

• @modelcontextprotocol/server-filesystem: Local file operations
• @modelcontextprotocol/server-postgres: Database queries
• @modelcontextprotocol/server-github: Repository management
• @modelcontextprotocol/server-slack: Team communication
• @modelcontextprotocol/server-brave-search: Web search capability

Browse the full marketplace at modelcontextprotocol.io/servers.

What Should You Do After Installation?

Installation is just the beginning. Here's how to transform your fresh OpenClaw setup into a useful assistant.

Test Core Functionality

Verify file operations:

Message your bot: "Create a file called test.txt with the content 'Hello World'"

Check your workspace:

ls ~/openclaw/workspace/

Test command execution:

"What's my current disk usage?"

The agent should execute df -h and return results.

Test memory:

"My favorite color is blue"

Later: "What's my favorite color?"

The agent should recall "blue" from context stored in local Markdown files.

Install Essential Skills

Skills extend OpenClaw's capabilities. Start with these:

Email integration:

openclaw skill install @openclaw/skill-gmail

Calendar management:

openclaw skill install @openclaw/skill-calendar

Web browsing:

openclaw skill install @openclaw/skill-playwright

Configure each skill by running:

openclaw skill configure skill-name

Follow authentication prompts for services that require OAuth.

Create Your First Automation

Build a simple workflow to understand how everything connects.

Example: Daily GitHub digest

Install GitHub skill:

openclaw skill install @openclaw/skill-github

Create a scheduled task in ~/.openclaw/tasks/github-digest.json:

{
  "name": "github-digest",
  "schedule": "0 9 * * *",
  "prompt": "Check my GitHub notifications and summarize new issues and pull requests. Send the summary to me on Telegram.",
  "enabled": true
}

This runs every day at 9 AM, fetches GitHub data, generates a summary, and messages you.

Enable scheduled tasks:

openclaw configure --enable-scheduler

Optimize Performance

Enable response caching:

{
  "cache": {
    "enabled": true,
    "ttl": 3600
  }
}

Caches LLM responses for repeated prompts, reducing API costs.

Configure rate limiting:

{
  "rateLimit": {
    "requestsPerMinute": 20,
    "tokensPerMinute": 50000
  }
}

Prevents accidental API quota exhaustion.

Set up monitoring:

Install the status dashboard:

openclaw dashboard install

Access at http://localhost:8174/dashboard to see:

• Request rate
• Token usage
• Error frequency
• Channel status

Plan for Maintenance

Enable automatic backups:

{
  "backup": {
    "enabled": true,
    "schedule": "0 2 * * *",
    "destination": "/backup/openclaw",
    "retain": 7
  }
}

Backs up configuration and memory files daily at 2 AM, keeping 7 days of history.

Set up log rotation:

openclaw configure --log-rotation

Prevents logs from filling disk space.

Subscribe to updates:

openclaw update subscribe

Receive notifications about new releases and security patches.

Join the Community

OpenClaw's rapid growth means active community support:

• GitHub Discussions: Ask questions and share configurations
• Discord server: Real-time help with troubleshooting
• Skills Marketplace: Contribute your own skills or request features

The community releases new integrations weekly. Following development helps you discover capabilities before they're widely documented.

Frequently Asked Questions

Can I use OpenClaw without a VPS?

Yes. Running on your personal computer works fine if you're okay with the agent being offline when your computer is off or asleep. For 24/7 operation, a VPS or miniPC is better.

How much does OpenClaw cost to run?

OpenClaw itself is free (MIT license). Costs come from:

• LLM API usage ($5-50/month for typical use)
• VPS hosting if used ($5-20/month)
• Optional: local GPU for models ($300-500 one-time)

Is my data private?

Your conversations and files stay on your hardware. The only external communication is API calls to your chosen LLM provider. Using local models via Ollama keeps everything completely private.

Can I switch LLM providers later?

Yes. Change the provider in config and restart. Your conversation history and skills continue working—OpenClaw abstracts the LLM interface.

How do I update OpenClaw?

Run openclaw update install. The updater preserves your configuration and skills. Always back up before major version updates.

What's the difference between OpenClaw and AutoGPT?

OpenClaw focuses on production deployment with robust daemon operation, messaging integrations, and the MCP ecosystem. AutoGPT emphasizes autonomous goal-driven behavior. OpenClaw prioritizes reliability; AutoGPT prioritizes autonomy.

---

This guide covered installation, configuration, security, troubleshooting, and post-setup optimization. Your OpenClaw instance is now ready to automate tasks, answer questions, and act as a persistent AI assistant across your messaging apps. Start with simple automations and gradually build complexity as you learn what works for your workflow.