🦞 Overview

OpenClaw AI (formerly known as Clawdbot and later Moltbot) is an open-source, self-hosted personal AI assistant designed to automate tasks around the clock. It is written in TypeScript/Node.js, licensed under the MIT License, and hosted on GitHub at github.com/openclaw/openclaw with over 182,000 stars as of 2026.

Unlike cloud-only AI services, OpenClaw runs locally on your own infrastructure — a Raspberry Pi, a VPS, or a bare-metal server — giving you full control over your data. It communicates via natural language and can be extended with a rich ecosystem of Skills (plugins) available on ClawHub.

Key Facts at a Glance

📜 History

OpenClaw has gone through three distinct naming phases, each representing a significant evolution in the project's scope and ambition.

⚡ Features

OpenClaw AI combines natural language processing with extensible automation. Below is an overview of the core feature set.

🏗️ Architecture

OpenClaw is built on a modular, event-driven architecture that separates core concerns into distinct layers. Understanding the architecture helps when building custom skills or deploying in production.

Core Layers

1. Messaging Adapters

The outermost layer translates messages from external platforms (Discord, Telegram, HTTP, etc.) into a unified internal message format. Each platform adapter is a thin wrapper around the platform's SDK.

2. Intent Engine

Parses user messages to determine the user's intent. This layer maps natural language to a structured command or skill invocation. The engine can use a local LLM, an OpenAI-compatible API endpoint, or a rule-based fallback.

3. Skill Router

Dispatches recognized intents to the correct installed Skill. Skills register themselves with the router on startup and declare which intents they handle.

4. Memory & State Store

Manages conversation history, user preferences, and task state using a local SQLite database by default. Can be replaced with PostgreSQL or Redis for production deployments.

5. Scheduler

Runs cron-style recurring jobs defined by Skills or by the user. The scheduler persists across restarts using the state store.

6. REST API Layer

Exposes OpenClaw's capabilities over HTTP/HTTPS, allowing external applications to send messages, query state, and trigger actions programmatically.

Skill Anatomy

Every OpenClaw Skill is a Node.js package that exports a manifest and a set of handlers:

// my-skill/index.js
module.exports = {
  name: 'my-skill',
  version: '1.0.0',
  intents: ['greet', 'farewell'],
  handlers: {
    greet: async (ctx) => {
      await ctx.reply(`Hello, ${ctx.user.name}!`);
    },
    farewell: async (ctx) => {
      await ctx.reply('Goodbye!');
    }
  }
};

Supported LLM Backends

• Local (Ollama) – Fully offline, privacy-first. Llama 3, Mistral, etc.
• OpenAI API – GPT-4o, GPT-3.5-Turbo (requires API key)
• Anthropic API – Claude 3/3.5 series (requires API key)
• OpenAI-compatible endpoints – LM Studio, llama.cpp, vLLM, etc.

🚀 Installation

Method 1 – npx (Quickstart)

The fastest way to try OpenClaw without a permanent install:

npx openclaw@latest init

This downloads the latest version, runs the interactive setup wizard, and starts the bot.

Method 2 – Git Clone (Recommended for Development)

# 1. Clone the repository
git clone https://github.com/openclaw/openclaw.git
cd openclaw

# 2. Install dependencies
npm install

# 3. Copy and edit the configuration
cp .env.example .env
nano .env

# 4. Start the bot
npm start

Method 3 – Docker

The official Docker image at hub.docker.com/r/openclaw/openclaw is the recommended production deployment method:

# Pull the latest image
docker pull openclaw/openclaw:latest

# Run with environment variables
docker run -d \
  --name openclaw \
  -e OPENCLAW_LLM_PROVIDER=ollama \
  -e OPENCLAW_LLM_MODEL=llama3 \
  -p 3000:3000 \
  -v openclaw-data:/app/data \
  openclaw/openclaw:latest

Method 4 – Docker Compose

version: '3.8'
services:
  openclaw:
    image: openclaw/openclaw:latest
    ports:
      - "3000:3000"
    volumes:
      - ./data:/app/data
    environment:
      - OPENCLAW_LLM_PROVIDER=ollama
      - OPENCLAW_LLM_MODEL=llama3
      - OPENCLAW_MESSAGING_DISCORD_TOKEN=your_token_here
    restart: unless-stopped

Method 5 – DigitalOcean One-Click

A one-click DigitalOcean Droplet deployment is available. See the official DigitalOcean guide for step-by-step instructions.

Configuration (.env)

Key environment variables:

After Installation

Once running, verify the installation by visiting http://localhost:3000/status. Then add your first skill:

npx clawhub install youdotcom-cli

🧩 Skills & ClawHub

Skills are the heart of OpenClaw's extensibility. Each Skill is a self-contained Node.js package that adds new capabilities — from web search and crypto trading to home automation and calendar management.

ClawHub – The Official Skill Registry

ClawHub.ai is the official package registry for OpenClaw Skills — think of it as npm for AI agent skills. It hosts over 700 verified skills organized by category.

Installing Skills via npx clawhub

# Install a skill
npx clawhub install <skill-name>

# Examples:
npx clawhub install youdotcom-cli        # You.com web search
npx clawhub install openclaw-weather     # Weather forecasts
npx clawhub install openclaw-calendar    # Google Calendar integration
npx clawhub install openclaw-notes       # Note-taking & summaries

# List installed skills
npx clawhub list

# Update all skills
npx clawhub update --all

# Remove a skill
npx clawhub remove <skill-name>

Popular Skill Categories

Building Your Own Skill

# Scaffold a new skill
npx clawhub create my-custom-skill
cd my-custom-skill

# Install development dependencies
npm install

# Develop and test
npm run dev

# Publish to ClawHub (requires account)
npx clawhub publish

👥 Community

Discord Server (8,000+ Members)

The official Discord server is the primary hub for real-time support, announcements, and casual discussion. Channels include #help, #skill-announcements, #showcase, and #off-topic.

Discourse Forum (15,000+ Members)

The official forum on Discourse is the place for long-form guides, feature requests, and detailed troubleshooting threads. Posts are indexed and searchable, making it a valuable knowledge archive.

Moltbook – AI Agent Social Network

Moltbook is a social network specifically for AI agents — the self-described "Front Page of the Agent Internet." With 2.5 million registered AI agents, it lets OpenClaw bots post updates, interact with other agents, and build a following.

MaltGram – Bot Social Network

MaltGram is an Instagram-inspired social platform for OpenClaw bots. Bots can share images, short posts, and connect with communities of both human followers and other AI agents.

ClawCon – Annual Conference

ClawCon is the annual OpenClaw community conference. First held in 2026, it features talks by core contributors, skill showcases, hackathons, and networking sessions — both virtual and in-person tracks.

How to Contribute

• Submit bug reports and feature requests on GitHub Issues
• Open pull requests against the main branch for patches
• Publish new Skills to ClawHub
• Add your skill to the Awesome Skills list
• Write tutorials and post on the Discourse forum

🛡️ Security

The ClawHub Malicious Skills Incident (2025)

A security audit conducted in late 2025 revealed 386 skills on ClawHub that contained malicious code designed to steal cryptocurrency wallet credentials and private keys. The skills were disguised as legitimate trading and DeFi automation tools. Authmind published a detailed supply chain analysis.

Security Best Practices

• Audit before installing: Always read the skill's source code on GitHub before running npx clawhub install.
• Check the publisher: Prefer skills from verified publishers or those with large download counts and active maintenance.
• Use pinned versions: In production, pin skill versions in your package.json to prevent unexpected updates.
• Run with limited privileges: Use Docker or a dedicated low-privilege system user for the bot process.
• Never store API keys in skills: Pass sensitive credentials via environment variables, never hardcode them.
• Review permissions: Skills that request access to your file system or external APIs beyond their stated purpose are suspicious.
• Keep OpenClaw updated: Security patches are released regularly. Run npm update frequently.

Reporting Vulnerabilities

Security vulnerabilities in the OpenClaw core should be reported privately via the GitHub Security Advisory tab at github.com/openclaw/openclaw/security. Do not open public issues for unpatched vulnerabilities.

🔌 Integrations

VS Code Extension

The official VS Code Extension (search: openclaw.openclaw) integrates OpenClaw directly into the editor. Features include:

• Inline code explanations and suggestions via the OpenClaw chat panel
• Contextual code review on selection
• Commit message generation
• Natural language file search and navigation

Docker

The official Docker image at hub.docker.com/r/openclaw/openclaw supports all architectures (amd64, arm64). Tags:

Messaging Platform Support

OpenClaw supports 12+ messaging platforms out of the box:

• Discord  ·  Telegram  ·  Slack  ·  WhatsApp (Business API)
• Matrix  ·  IRC  ·  Mattermost  ·  Microsoft Teams
• Custom HTTP webhook  ·  REST API

OpenClawS.io

OpenClawS.io provides additional automation tools and community-driven integrations with external services such as IFTTT, Zapier, and n8n.

☁️ Cloud & Hosting

While OpenClaw is designed for self-hosting, several managed and cloud-based options exist for users who prefer not to manage infrastructure.

Recommended Production Stack

VPS (Ubuntu 22.04) + Docker + Nginx reverse proxy + Let's Encrypt SSL
├── openclaw/openclaw:latest   (port 3000, internal only)
├── nginx                      (port 443, reverse proxy)
└── certbot                    (SSL renewal via cron)

🔀 Alternatives

OpenClaw is not the only option in the personal AI assistant space. Here is a comparison of notable alternatives:

When to Choose OpenClaw

• You need a 24/7 autonomous agent that acts without being prompted
• You want a messaging bot that lives on Discord, Telegram, or Slack
• You want to extend functionality via a rich skill ecosystem
• You require full data sovereignty and local execution

When to Choose an Alternative

• You primarily want document Q&A → AnythingLLM
• You want a completely offline desktop chatbot → Jan.ai
• You are a developer working in VS Code / terminal → Claude Code

❓ FAQ