OpenClaw - Personal AI Assistant

Overview

OpenClaw is a free and open-source autonomous AI assistant framework that enables users to deploy a personal AI assistant on their own infrastructure. Originally developed by Austrian software engineer Peter Steinberger in November 2025 under the name "Clawdbot," the project was later renamed to "Moltbot" and finally "OpenClaw" in early 2026. It has achieved significant popularity, amassing over 117,000 GitHub stars and 16,500 forks.

Project Links

• GitHub Repository: https://github.com/openclaw/openclaw
• Official Website: https://openclaw.ai
• Documentation: https://docs.openclaw.ai
• License: MIT
• Language: TypeScript

Core Concept

OpenClaw distinguishes itself as an AI assistant that "actually does things" rather than simply responding to prompts. It operates as an autonomous agent capable of executing tasks independently through large language models (LLMs), using messaging platforms as its primary user interface.

Key Features

1. Multi-Platform Integration

• Connects to 100+ messaging services including:
  • WhatsApp (via Baileys)
  • Telegram (via grammY)
  • Slack (via Bolt)
  • Discord (via discord.js)
  • Google Chat
  • Signal
  • iMessage
  • Microsoft Teams
  • Matrix
  • BlueBubbles
  • WebChat
  • And many more

2. Local-First Gateway Architecture

• Single WebSocket control plane for sessions, channels, tools, and events
• Runs entirely on your own hardware (Mac mini, Linux server, Windows via WSL2)
• Privacy-focused: all data remains on your infrastructure
• No cloud dependencies for core functionality

3. Multi-Agent Routing

• Route inbound channels/accounts to isolated agents
• Workspace-based organization with per-agent sessions
• Support for multiple concurrent AI models and providers

4. Voice & Speech Capabilities

• Voice Wake: always-on speech recognition for macOS/iOS/Android
• Talk Mode: continuous conversation with ElevenLabs integration
• Push-to-talk support via companion apps

5. Live Canvas

• Agent-driven visual workspace
• A2UI (Agent-to-UI) framework for dynamic interfaces
• Real-time rendering and interaction

6. Advanced Tool Integration

• Browser Control: Dedicated Chrome/Chromium automation with CDP
• Canvas Operations: Push/reset, eval, snapshot capabilities
• Device Nodes: Camera, screen recording, location services
• System Integration: Notifications, local command execution
• Automation: Cron jobs, webhooks, Gmail Pub/Sub triggers

7. Skills Platform

• Modular skill system with bundled, managed, and workspace skills
• ClawdHub skill registry for discovery and installation
• Custom skill development support

8. Model Flexibility

Supports multiple AI providers:

• Anthropic (Claude models - Opus 4.5, Sonnet 4.5, Haiku 4.5)
• OpenAI (GPT models, Codex)
• Google (Gemini)
• Other LLM providers via API

Architecture

Core Components

Gateway

• Central WebSocket server (default: ws://127.0.0.1:18789)
• Manages sessions, presence, configuration
• Handles message routing and delivery
• Provides Control UI dashboard

Pi Agent Runtime

• RPC-based execution engine
• Tool streaming and block streaming
• Context window management
• Session isolation

Session Model

• Main session for direct chats
• Group isolation with activation modes
• Queue management for concurrent requests
• Reply-back routing

Security Model

Default Behavior:

• Tools run on host for main session (full access for authenticated users)
• Configurable sandbox mode for group/channel sessions
• DM pairing system to prevent unauthorized access

Sandbox Options:

• Per-session Docker containers for non-main sessions
• Tool allowlisting/denylisting
• Network isolation capabilities

Installation & Deployment

System Requirements

• Runtime: Node.js ≥22
• OS Support: macOS, Linux, Windows (via WSL2)
• Package Managers: npm, pnpm, or bun

Quick Start

# Install globally
npm install -g openclaw@latest

# Run onboarding wizard
openclaw onboard --install-daemon

# Start gateway
openclaw gateway --port 18789 --verbose

# Send a test message
openclaw message send --to +1234567890 --message "Hello from OpenClaw"

# Interact with agent
openclaw agent --message "Ship checklist" --thinking high

From Source

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

pnpm install
pnpm ui:build
pnpm build

pnpm openclaw onboard --install-daemon
pnpm gateway:watch  # Development mode with auto-reload

Configuration

Minimal Configuration

{
  "agent": {
    "model": "anthropic/claude-opus-4-5"
  }
}

Channel Configuration Examples

WhatsApp:

pnpm openclaw channels login

Telegram:

{
  "channels": {
    "telegram": {
      "botToken": "YOUR_BOT_TOKEN"
    }
  }
}

Discord:

{
  "channels": {
    "discord": {
      "token": "YOUR_DISCORD_TOKEN"
    }
  }
}

Security & Privacy

Security Features

• DM pairing system (strangers receive pairing codes)
• Allowlist-based access control
• Configurable sandbox environments
• Optional Docker isolation for untrusted sessions
• Tailscale integration for secure remote access

Privacy Advantages

• Self-hosted: complete data ownership
• No cloud telemetry by default
• Local credential storage
• Configurable network restrictions

Security Considerations

According to security researchers:

• Requires broad permissions to function effectively
• Suitable for advanced users who understand security implications
• Recommended to run in isolated environments for production
• Supply chain risks from extensible architecture
• Should not connect to production services without proper security measures

Remote Access

OpenClaw supports remote gateway access via:

• Tailscale Serve/Funnel: Secure tailnet or public HTTPS
• SSH Tunnels: Traditional secure tunneling
• Token/Password Authentication: Configurable auth modes

This allows running the gateway on a Linux server while controlling it from any device.

Chat Commands

Users can control sessions with commands:

• /status - Session information
• /new or /reset - Reset conversation
• /compact - Summarize context
• /think <level> - Adjust thinking depth
• /verbose on|off - Toggle verbose output
• /usage off|tokens|full - Usage statistics
• /restart - Restart gateway (owner only)
• /activation mention|always - Group settings

Companion Applications

macOS App

• Menu bar control for gateway health
• Voice Wake and push-to-talk
• WebChat interface
• Debug tools
• Remote gateway control

iOS Node

• Voice trigger forwarding
• Canvas surface rendering
• Node-based command execution

Android Node

• Canvas, camera, screen capture
• Node pairing via Bridge

Agent Workspace & Skills

• Workspace Root: ~/.openclaw/workspace
• Prompt Files: AGENTS.md, SOUL.md, TOOLS.md
• Skills Directory: ~/.openclaw/workspace/skills/

Use Cases

1. Personal Productivity Assistant

   • Email management and responses
   • Calendar coordination
   • Task automation

2. Development Automation

   • GitHub notifications
   • Automated deployments
   • Code review assistance

3. Multi-Channel Communication Hub

   • Unified inbox across platforms
   • Automated responses
   • Message forwarding and routing

4. Custom Workflow Automation

   • Integration with internal tools
   • Data processing pipelines
   • Scheduled tasks and reminders

Community & Development

• Active Development: 8,368+ commits
• Open Source: MIT License
• Community: Discord server available
• Contributions: Welcomes PRs and community involvement
• Documentation: Comprehensive docs at docs.openclaw.ai

Notable Achievements

• 117k+ GitHub stars
• 16.5k+ forks
• Featured on Wikipedia
• Viral popularity through Moltbook integration
• Significant interest from Silicon Valley and international developers

Comparison to Alternatives

Unlike ChatGPT or Claude AI:

• Autonomous: Executes tasks independently vs. reactive responses
• Self-Hosted: Runs on your infrastructure
• Multi-Platform: Native integration with 100+ services
• Extensible: Custom skills and tools
• Privacy-First: Complete data control

Getting Started Resources

1. Official Documentation
2. Getting Started Guide
3. Configuration Reference
4. Security Guide
5. Troubleshooting

Future Development

OpenClaw is actively maintained with:

• Regular updates and bug fixes
• New channel integrations
• Enhanced security features
• Expanded AI model support
• Growing skills ecosystem

Credits

• Creator: Peter Steinberger
• Special Thanks: Mario Zechner (pi-mono), Adam Doppelt (lobster.bot)
• Community: 200+ contributors and growing

Conclusion

OpenClaw represents a paradigm shift in personal AI assistants, emphasizing autonomy, privacy, and self-hosting. It's ideal for developers and advanced users who want complete control over their AI assistant while maintaining the flexibility to integrate with existing workflows and communication channels.

For those comfortable with self-hosting and seeking an AI assistant that truly operates as an autonomous agent rather than a chatbot, OpenClaw provides a powerful, open-source foundation.