OpenClaw: The Open-Source Personal AI Assistant That Actually Does Things

What Is OpenClaw?
Why OpenClaw Is Different
Core Features
Architecture Overview
Technology Stack
Installation & Quick Start
Configuration Basics
Supported Channels
The Skills Ecosystem
Security Model
Real-World Use Cases
Platform Support
Community & Ecosystem
Conclusion

What Is OpenClaw?

OpenClaw is an open-source personal AI assistant that runs entirely on your own devices. Unlike cloud-based AI assistants such as Siri, Google Assistant, or Alexa, OpenClaw is a local-first, self-hosted AI gateway that connects to your existing messaging apps and takes real actions on your behalf.

With its tagline — “The AI that actually does things” — OpenClaw has exploded in popularity, accumulating over 232,000 GitHub stars and 44,500 forks in roughly three months. Created by Peter Steinberger (@steipete) and a vibrant open-source community of 866+ contributors, this MIT-licensed project has been described as “Everything Siri was supposed to be” and “An iPhone moment” for personal AI.

The project was originally called Clawdbot, then briefly renamed to Moltbot, before settling on its current name OpenClaw. Its playful mascot is a space lobster named Molty.

Why OpenClaw Is Different

Traditional AI assistants live in corporate clouds. Your data flows through their servers, your capabilities are limited to what they expose, and you’re locked into their ecosystem. OpenClaw flips this model entirely:

1. Local-First, Privacy-Preserving

Your data never leaves your machine. The AI gateway runs on your own hardware — from a Raspberry Pi to a cloud VM — and you control everything.

2. Channel-Agnostic

Instead of being trapped in one app, OpenClaw works across all major messaging platforms simultaneously. Send a message on WhatsApp, get a response on Telegram, continue the conversation on Slack — it’s all the same assistant.

3. Full System Access

Unlike voice assistants that can only toggle a light or set a timer, OpenClaw can read and write files, run shell commands, browse the web, execute scripts, and interact with your entire operating system (with proper sandboxing).

4. Self-Extending

The AI can write its own skills and modify its own prompts. It literally learns and adapts to your workflow, becoming more capable over time.

5. Single-User Focus

OpenClaw is designed as a personal assistant, not an enterprise SaaS product. This means it’s optimized for your workflow, your preferences, and your privacy.

Core Features

OpenClaw packs an impressive feature set that goes far beyond simple Q&A:

Multi-Channel Messaging

Connect your AI to any combination of messaging platforms. OpenClaw supports WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage (via BlueBubbles), Microsoft Teams, Matrix, Zalo, WebChat, and more.

Browser Control

OpenClaw spins up a dedicated Chrome/Chromium instance with CDP (Chrome DevTools Protocol) control. It can browse the web, fill out forms, extract data, and automate web-based workflows — all hands-free.

Persistent Memory

Unlike stateless chatbots, OpenClaw remembers context across conversations and sessions. It runs 24/7, maintaining a continuous understanding of your needs and history.

Voice Capabilities

Voice Wake + Talk Mode is available on macOS, iOS, and Android (powered by ElevenLabs). Just say a wake word and start talking.

Live Canvas

Agent-driven visual workspace (A2UI) where OpenClaw can generate and display rich interactive content directly in the companion apps.

Proactive Behavior

Set up cron jobs, webhooks, Gmail Pub/Sub triggers, and heartbeat check-ins. OpenClaw doesn’t just wait for you — it can act autonomously based on schedules and events.

Camera & Screen Access

On connected devices, OpenClaw can snap photos, record the screen, and use visual context to assist you.

Multi-Agent Routing

Route different channels or contacts to isolated agents with different configurations, models, and capabilities.

Architecture Overview

OpenClaw follows a gateway-based architecture with a WebSocket control plane:

WhatsApp / Telegram / Slack / Discord / etc.
               │
               ▼
┌───────────────────────────────┐
│            Gateway            │
│       (control plane)         │
│     ws://127.0.0.1:18789      │
└──────────────┬────────────────┘
               │
               ├── CLI (openclaw …)
               ├── WebChat UI
               ├── macOS menu bar app
               ├── iOS node
               └── Android node

The Gateway is the heart of OpenClaw. It’s a local WebSocket-based control plane that coordinates all the companion apps and messaging channels. Each device — macOS, iOS, Android — connects as a “node” that can execute device-local actions like taking photos, recording screen, or using voice.

This architecture means you can run the gateway on a headless Linux server and connect your phone as a remote node, giving the AI access to your phone’s camera to do things like scan documents or read physical text.

Technology Stack

OpenClaw is a modern, polyglot project built with production-grade tooling:

Technology	Usage
TypeScript	85.5% — Core gateway, CLI, tools
Swift	10.3% — macOS and iOS apps
Kotlin	1.8% — Android app
Node.js ≥22	Runtime requirement
pnpm	Preferred package manager
WebSocket	Gateway control plane protocol
Docker	Sandboxing and deployment
Vitest	Testing framework
Oxlint/Oxfmt	Linting and formatting
Nix	Declarative configuration option

For AI models, OpenClaw supports any LLM provider but recommends Anthropic Claude Opus 4 for its long-context capabilities and prompt-injection resistance. It also works with OpenAI, local models, MiniMax, and others.

Installation & Quick Start

Getting started with OpenClaw is remarkably simple.

One-Liner Install

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

Via npm

npm install -g openclaw@latest

Guided Setup

openclaw onboard --install-daemon

This launches an interactive setup wizard that walks you through:

Configuring your AI model provider
Connecting messaging channels
Setting up security (pairing codes)
Installing essential skills

Start the Gateway

openclaw gateway --port 18789 --verbose

Send Your First Message

openclaw message send --to +1234567890 --message "Hello from OpenClaw!"

Run an Agent Command

openclaw agent --message "Check my calendar for today" --thinking high

Configuration Basics

OpenClaw uses a JSON configuration file at ~/.openclaw/openclaw.json. A minimal setup looks like:

{
  "agent": {
    "model": "anthropic/claude-opus-4-6"
  }
}

You can expand this with channel configurations, skill settings, security rules, and more. Here’s a more realistic example:

{
  "agent": {
    "model": "anthropic/claude-opus-4-6",
    "thinking": "medium"
  },
  "channels": {
    "telegram": {
      "enabled": true,
      "botToken": "YOUR_BOT_TOKEN"
    },
    "whatsapp": {
      "enabled": true
    }
  },
  "security": {
    "pairingRequired": true,
    "dockerSandbox": true
  }
}

Supported Channels

One of OpenClaw’s strongest differentiators is its wide channel support:

Channel	Status	Notes
WhatsApp	✅ Stable	Via WhatsApp Business API
Telegram	✅ Stable	Bot API integration
Slack	✅ Stable	Workspace apps
Discord	✅ Stable	Bot integration
Signal	✅ Stable	Via Signal CLI
iMessage	✅ Stable	Via BlueBubbles (macOS)
Google Chat	✅ Stable	Workspace integration
Microsoft Teams	✅ Stable	Bot framework
Matrix	✅ Stable	Open protocol
Zalo	✅ Beta	Vietnamese messenger
WebChat	✅ Stable	Built-in web interface

All channels share the same persistent context, so you can seamlessly switch between platforms mid-conversation.

The Skills Ecosystem

Skills are OpenClaw’s extension mechanism. They’re modular capabilities that the AI can use to interact with external services and perform specialized tasks.

Skill Types

Bundled Skills — Come pre-installed (file management, web browsing, shell commands)
Managed Skills — Officially maintained, installable via CLI
Community Skills — Published on ClawHub, the skill registry

ClawHub

ClawHub (clawhub.com) is the central registry for discovering and installing skills. Think of it as npm for AI capabilities. With 50+ integrations available, you can extend OpenClaw with:

Spotify — Control music playback
Philips Hue — Smart lighting control
Obsidian — Knowledge base management
GitHub — Repository management, PR creation
Gmail — Email management and automation
Twitter/X — Social media posting
Calendar — Schedule management

Self-Extending

The most remarkable aspect is that OpenClaw can write its own skills. If you ask it to do something it can’t do yet, it can create a new skill to handle that capability going forward.

Security Model

For an AI with system access, security is paramount. OpenClaw implements several layers:

Pairing Approval

DMs from unknown senders require a short pairing code approval. No random person can just message your bot and gain access.

Docker Sandboxing

Non-main sessions can run inside per-session Docker containers, isolating potentially risky operations.

Tool Allow/Deny Lists

You can explicitly control which tools are available in which contexts, limiting what the AI can do based on the channel or user.

Tailscale Integration

Secure remote access via Tailscale Serve/Funnel — no port forwarding or exposed services needed.

Health Check

Run openclaw doctor to surface risky configurations and security recommendations.

Real-World Use Cases

The OpenClaw community has shared impressive use cases:

Personal Productivity

Managing Gmail, Calendar, and WordPress entirely from Telegram
Daily briefings with weather, schedule, and news summaries
Automatic flight check-in reminders

Developer Workflow

Autonomous code testing and PR creation triggered by Sentry webhooks
Building and deploying websites from a phone
Git repository management and code review

Smart Home

Controlling air purifiers and Hue lights via natural language
Custom automation routines triggered by time or events

Health & Wellness

WHOOP health data monitoring and analysis
Custom meditation generation with text-to-speech
Daily exercise tracking and recommendations

Business Operations

Insurance claim handling and documentation
Automated email unsubscribe workflows
Customer support triage across messaging channels

Platform Support

OpenClaw runs practically everywhere:

Platform	Support Level	Features
macOS	Full	Menu bar app, Voice Wake, Push-to-Talk, Canvas, WebChat
iOS	Full	Node app, Canvas, Voice Wake, Talk Mode, Camera, Bonjour pairing
Android	Full	Canvas, Talk Mode, Camera, Screen recording, SMS
Linux	Gateway	Runs great on Raspberry Pi, cloud VMs, and home servers
Windows	WSL2	Full gateway support via Windows Subsystem for Linux

The recommended setup is running the Gateway on a Linux server or Mac, with companion apps on your phone for mobile access.

Community & Ecosystem

OpenClaw has built a remarkable community in a very short time:

232,000+ GitHub Stars — One of the fastest-growing open-source projects ever
866+ Contributors — Active open-source development
55 Releases — Rapid iteration with regular updates
Sponsors — Backed by OpenAI, Blacksmith, and others

awesome-openclaw-skills — Community-curated skill collection (20K+ stars)
awesome-openclaw-usecases — Real-world use case collection (9K+ stars)
nanobot — An ultra-lightweight OpenClaw alternative (25K+ stars)

Chat Commands

Control OpenClaw from any connected channel:

Command	Action
`/status`	Check session status
`/new` or `/reset`	Reset the session
`/compact`	Summarize and compact context
`/think <level>`	Set thinking level (low/medium/high)
`/verbose on\|off`	Toggle verbose mode
`/restart`	Restart the gateway

Community Endorsement

The project has been endorsed by Andrej Karpathy, featured in MacStories and StarryHope, and praised widely across the developer community. Users have described it as “running my company” and a signal that “a megacorp like Anthropic or OpenAI could not build this” — a testament to the power of open-source development.

Conclusion

OpenClaw represents a paradigm shift in personal AI assistants. By combining local-first architecture, multi-channel messaging, full system access, and a self-extending skills platform, it delivers on the promise of an AI that genuinely works for you.

Whether you’re a developer looking to automate workflows, a power user wanting to control your smart home through natural language, or simply someone tired of cloud AI limitations, OpenClaw offers a compelling alternative. With its MIT license, active community, and explosive growth to 232K+ stars, it’s arguably the most exciting open-source AI project of 2026.

Getting started takes just one command:

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

Give it a try, connect your favorite messaging app, and experience what a truly personal AI assistant feels like.

Official Links:

GitHub: github.com/openclaw/openclaw
Website: openclaw.ai
Documentation: docs.openclaw.ai
ClawHub: clawhub.com

Have you tried OpenClaw or any other self-hosted AI assistant? What was your experience like? Or maybe you have a cool use case idea that wasn’t mentioned above? Drop your thoughts in the comments — let’s discuss! 💬

Table of Contents