OpenClaw Setup: Your Personal AI Agent in 10 Minutes

OpenClaw is an open-source personal AI agent that lives on your hardware and connects to your entire digital life. Not a chatbot. Not an API wrapper. A persistent agent with memory, tools, and the ability to act on your behalf across messaging platforms, home automation, email, calendars, and more.

I have been running OpenClaw since early 2026, and it has fundamentally changed how I work. This is the guide I wish existed when I started.

What Makes OpenClaw Different

Most AI tools are stateless — you open a chat, get an answer, close the tab. OpenClaw is stateful and persistent:

• Always on — runs as a daemon on your hardware (Mac Mini, Raspberry Pi, VPS, or Kubernetes)
• Memory — remembers conversations, decisions, preferences, and context across sessions
• Multi-channel — one agent, reachable via Discord, Telegram, WhatsApp, Signal, email, or web UI
• Tool use — executes commands, reads files, browses the web, controls smart home devices
• Self-hosted — your data stays on your hardware, with your API keys

Think of it as a personal assistant that never sleeps, never forgets (unless you tell it to), and gets better the longer you use it.

Architecture Overview

┌─────────────────────────────────────────┐
│              OpenClaw Gateway            │
│  ┌─────────┐ ┌─────────┐ ┌──────────┐  │
│  │ Discord │ │Telegram │ │ WhatsApp │  │
│  │ Channel │ │ Channel │ │ Channel  │  │
│  └────┬────┘ └────┬────┘ └────┬─────┘  │
│       └───────────┼───────────┘         │
│              ┌────▼────┐                │
│              │  Agent  │                │
│              │  Core   │                │
│              └────┬────┘                │
│    ┌──────┬──────┼──────┬──────┐       │
│    │      │      │      │      │       │
│  Memory  Tools  Web   Files  Nodes     │
│  (.md)  (exec) (fetch) (r/w) (devices) │
└─────────────────────────────────────────┘

Quick Start: Docker Installation

The fastest way to get OpenClaw running:

# Create workspace directory
mkdir -p ~/.openclaw/workspace

# Run OpenClaw with Docker
docker run -d \
  --name openclaw \
  --restart unless-stopped \
  -v ~/.openclaw:/home/node/.openclaw \
  -p 3000:3000 \
  ghcr.io/openclaw/openclaw:latest

Open http://localhost:3000 to access the Control UI.

Configuration

OpenClaw is configured through ~/.openclaw/config.yaml. Here is a minimal production configuration:

# ~/.openclaw/config.yaml
gateway:
  bind: "0.0.0.0"     # LAN access (use 127.0.0.1 for local only)
  controlUI:
    allowedOrigins:
      - "http://localhost:3000"
      - "http://192.168.1.100:3000"  # Your LAN IP

agents:
  defaults:
    model: "anthropic/claude-sonnet-4-20250514"
    # Or use a cheaper model for routine tasks:
    # model: "openai/gpt-4o-mini"

# API keys for model providers
env:
  ANTHROPIC_API_KEY: "sk-ant-..."
  # OPENAI_API_KEY: "sk-..."

# Connect messaging channels
channels:
  discord:
    enabled: true
    token: "your-discord-bot-token"
  telegram:
    enabled: true
    token: "your-telegram-bot-token"

The allowedOrigins Gotcha

The most common setup issue is the allowedOrigins configuration. If you see “Origin not allowed” errors, you need to add your browser’s origin to the allowed list:

gateway:
  controlUI:
    allowedOrigins:
      - "http://localhost:3000"
      - "http://YOUR_LAN_IP:3000"

See the complete allowedOrigins guide for Docker, LAN, Tailscale, and reverse proxy setups.

Memory: How OpenClaw Remembers

OpenClaw uses file-backed memory — Markdown files in your workspace that persist across sessions:

~/.openclaw/workspace/
├── MEMORY.md          # Long-term curated knowledge
├── SOUL.md            # Agent personality and behavior
├── USER.md            # Information about you
├── AGENTS.md          # Workspace conventions
├── memory/
│   ├── 2026-04-01.md  # Daily notes
│   ├── 2026-04-02.md
│   └── 2026-04-03.md
└── ...your files...

The agent reads these files at the start of each session. It writes to daily memory files during conversations and periodically curates MEMORY.md with long-term insights.

This is not a vector database or embeddings-based RAG. It is plain Markdown files that you can read, edit, and version control. Simple, transparent, and debuggable.

For more on the memory system, see file-backed memory guide and memory search configuration.

Multi-Channel Messaging

One agent, every platform. OpenClaw connects to:

• Discord — full server integration with channel-specific behavior
• Telegram — personal bot or group chat
• WhatsApp — via linked device (QR code pairing)
• Signal — privacy-focused messaging
• Email — Gmail integration for inbox monitoring
• Web UI — built-in Control UI at port 3000

The agent maintains context across channels. Tell it something on Telegram, reference it later on Discord — it remembers.

See the channel setup guide for detailed configuration.

Tool Use: What OpenClaw Can Do

OpenClaw is not just a chat interface. It has access to tools:

File Operations

Read, write, and edit files in your workspace. Create scripts, update configs, manage documentation.

Shell Execution

Run commands on the host machine. Git operations, system monitoring, deployment scripts, and anything you would do in a terminal.

Web Access

Search the web, fetch and read web pages, and extract information. Research topics, check documentation, and monitor websites.

Browser Automation

Control a web browser for tasks that need JavaScript rendering — fill forms, navigate SPAs, and take screenshots.

Node Control

Manage paired devices (phones, Raspberry Pis) — camera snapshots, screen recording, location, notifications, and remote command execution.

Messaging

Send messages proactively, create polls, react to messages, manage channels — the agent can participate in conversations as a full member.

Real-World Use Cases

Personal Knowledge Management

OpenClaw reads your files, remembers your decisions, and helps you find things you discussed weeks ago. It curates your memory files over time, acting as a persistent second brain.

Home Automation

Connected to Home Assistant, OpenClaw can control lights, check security cameras via Frigate, manage NAS storage, and monitor your UPS power.

Development Assistant

It reads your codebase, runs tests, commits changes, and pushes to Git. I use it to manage this very website — it writes blog posts, generates thumbnails, fixes broken links, and deploys changes.

Communication Hub

Monitor email, respond to messages across platforms, schedule meetings, and manage your calendar. One agent handling all your communication channels.

Proactive Monitoring

Heartbeat checks run periodically — the agent checks email, calendar, weather, and project status, then notifies you only when something needs attention.

Hardware Options

OpenClaw runs on anything that runs Docker:

See the hardware comparison guide and the Raspberry Pi budget setup.

Cost: What It Actually Costs to Run

OpenClaw itself is free (open source). Your costs are:

• LLM API calls: $5-30/month depending on usage and model choice
• Hardware: One-time (Pi/Mac) or $5-10/month (VPS)
• Optional: Local LLM with Ollama = $0 API costs (but needs decent hardware)

For a detailed comparison of API vs local models, see the honest comparison guide.

Security Considerations

OpenClaw runs on your hardware with your credentials. Security best practices:

1. Bind to localhost unless you need LAN access
2. Use allowedOrigins to restrict Control UI access
3. Keep API keys in config, not in workspace files
4. Review agent actions — enable confirmation for external actions (emails, public posts)
5. Use Tailscale for secure remote access without exposing ports

The agent follows safety principles: it asks before sending external communications, prefers trash over rm, and will not exfiltrate private data.

Getting Started Today

1. Install Docker on your machine
2. Run the Docker command above
3. Open http://localhost:3000
4. Add your LLM API key in the config
5. Start talking — the agent will guide you through setup

The first conversation is special: OpenClaw will ask your name, figure out its own identity, and set up the workspace files. After that, it remembers everything.

Troubleshooting

Common issues and fixes:

• “Origin not allowed” → Fix allowedOrigins
• Docker connection refused → Docker networking guide
• API key errors → API key troubleshooting
• Discord intents error → Privileged intents fix
• Gateway startup issues → Health check guide

Related Resources

• OpenClaw CLI Reference
• Config Migration Guide
• Hot Reload Configuration
• VPS Deployment Guide
• OpenClaw on Kubernetes
• OpenClaw Project

About the Author

I am Luca Berton, AI and Cloud Advisor. I run OpenClaw daily and have written 38+ articles on setup, troubleshooting, and advanced use cases. Book a consultation to discuss AI agent deployment for your team.