OpenClaw

Featured

Open-source personal AI assistant framework for privacy-first, self-hosted deployment. Enables autonomous task execution using local LLMs and tools.

Developer
OpenClaw Community (Open Source)
Type
Open-Source Framework
Pricing
Free & Open Source (Self-Hosted)
AI Model
Model-agnostic — local models via Ollama, or OpenAI / Anthropic / Gemini APIs
Difficulty
Advanced
🌐 Official Website💰 Pricing & Access
On this page

OpenClaw is a self-hosted, privacy-first personal AI agent framework — built for people who want a powerful AI assistant running on their own hardware, connected to their own tools, with their data never leaving their control. Think of it as building your own Claude or ChatGPT, but one that runs locally, integrates with any platform you choose, and extends with community-built skills.

Overview

As commercial AI assistants increasingly centralize personal data in cloud services, a community of privacy-conscious developers built OpenClaw as an open-source alternative. The framework allows users to deploy a fully autonomous AI agent on their home server, NAS device, or private cloud instance.

OpenClaw is designed around the concept of "Skills" — modular plugins that give the agent the ability to interact with the world: browse the web, read emails, control smart home devices, execute code, manage calendars, and more. The community maintains a growing library of skills on ClawHub, the community plugin registry.

The framework is model-agnostic: it works with local models via Ollama for maximum privacy, or with cloud APIs (OpenAI, Anthropic, Gemini) for maximum performance.

Key Features

  • Local-First Architecture: Designed to run on your own devices. The Gateway is only the control plane — the product is the assistant.
  • Very Wide Channel Support: OpenClaw's supported channels include WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, IRC, Microsoft Teams, Matrix, Feishu, LINE, Mattermost, Nextcloud Talk, Nostr, Synology Chat, Tlon, Twitch, Zalo, WeChat, QQ, and a built-in WebChat.
  • Skill System (Modular Plugins): The agent's capabilities are defined by installed skills. Skills live in your workspace at ~/.openclaw/workspace/skills/<skill>/SKILL.md.
  • ClawHub Skill Registry: The community skills registry at clawhub.ai.
  • Model Agnostic: Use any AI model via Ollama locally, or connect cloud APIs for tasks that need maximum intelligence.
  • Voice and Canvas: OpenClaw can speak and listen on macOS, iOS and Android, and render a live Canvas you control.
  • Sandboxing: Set agents.defaults.sandbox.mode: "non-main" to run non-main sessions inside sandboxes. Docker is the default sandbox backend; SSH and OpenShell backends are also available.
  • Automation: Cron jobs, webhooks and Gmail Pub/Sub triggers.
  • Windows Hub: A native Windows companion app for setup, tray status, chat, node mode and local MCP mode.

How It Works

  1. Install: npm install -g openclaw@latest (npm, pnpm or bun all work). Docker and Nix installs are also documented.
  2. Onboard: Run openclaw onboard, which walks you through the gateway, workspace, channels and skills. Works on macOS, Linux and Windows.
  3. Interact: Talk to your agent on whichever channel you already use.
  4. Execute: The agent plans multi-step tasks using installed skills and available tools, executing autonomously or with confirmations (configurable per skill).

Technical Architecture:

  • Core Language: TypeScript, running on Node.js.
  • Licence: MIT.
  • Inference: Ollama (local), OpenAI API, Anthropic API, Gemini (configurable).
  • Deployment: npm/pnpm/bun global install, Docker, or Nix.
  • Skill Interface: A SKILL.md-based skill format inside the agent workspace.
  • Sandboxing: Docker (default), SSH, or OpenShell backends.

Use Cases

Privacy-First Personal Assistant

  • Manage your calendar, email, and notes through a conversational AI that never sends your data to third parties.
  • Run AI queries on sensitive personal documents and financial data with full data sovereignty.

Developer Infrastructure

  • Use OpenClaw as a personal development agent: run tests, deploy code, monitor servers, and receive alerts via Telegram.
  • Chain custom skills to build automated workflows for your specific development environment.

Smart Home & Automation

  • Install Home Assistant skills to control smart devices through natural language commands.
  • Set up automated routines: "Every morning at 7am, summarize my unread emails and today's calendar, and send it to Telegram."

Research & Knowledge Management

  • Pair with Ollama + a strong local model to analyze documents without cloud exposure.
  • Build a private knowledge base from your notes and have the agent answer questions with context.

Getting Started

Requirements

  • A machine to run it on: your laptop, a home server, a NAS, or a VPS.
  • Node.js (npm, pnpm or bun) — or use the Docker or Nix install paths.
  • (Optional) Ollama for local model inference.

Step 1: Install

npm install -g openclaw@latest
# or: pnpm add -g openclaw@latest

Step 2: Onboard

openclaw onboard

The onboarding wizard walks you step by step through setting up the gateway, workspace, channels and skills. It is the recommended setup path and works on macOS, Linux and Windows. Windows desktop users can start with the native Windows Hub companion app instead.

Step 3: Configure Your Model

Point OpenClaw at whichever backend you want — a local model served by Ollama, or a cloud API from OpenAI, Anthropic or Google. See the getting started guide for the current configuration keys.

Step 4: Add Skills

Skills live in your agent workspace at ~/.openclaw/workspace/skills/<skill>/SKILL.md. Browse the community registry at ClawHub, and see the skills documentation for the difference between bundled, managed and workspace skills.

Best Practices

  • Start with a local model to validate your setup before adding cloud API keys.
  • Install skills one at a time and test each before adding the next.
  • Turn on sandboxing for group and channel sessions: agents.defaults.sandbox.mode: "non-main". Docker is the default sandbox backend.
  • Use cron jobs to automate routine daily briefings.

Pricing & Access

  • Core Framework: Free and open-source (MIT License).
  • Self-Hosted: No subscription fees. You pay for your own compute (hardware or VPS).
  • Inference Costs: If using cloud APIs (OpenAI, Anthropic, Google), you pay those providers directly.
  • ClawHub Community Skills: Community-contributed skills, published at clawhub.ai.

Limitations

  • Technical Setup Required: Not plug-and-play — requires a terminal and some comfort with configuration.
  • Performance vs. Cloud: Small local models won't match frontier cloud models on complex reasoning tasks.
  • Community Support: No official support SLA — community-driven help only.
  • Skill Maturity: Community skills vary in quality and maintenance. Review the GitHub activity of any skill before relying on it.
  • Broad Permissions: An always-on assistant wired into your messaging accounts is a large attack surface. Use the sandbox modes and per-skill confirmations.

Community & Support

Related Tools

Related Tools

ZeroClaw

Open-source AI personal assistant infrastructure written in Rust. A single small binary that swaps LLM providers, channels, and tools, and runs anywhere.

Aider

CLI-based AI pair programmer that edits code directly in your local git repository with multi-model support, git-native commits, and autonomous debugging.

n8n

Open-source agentic workflow automation. Build and scale autonomous AI agents from a low-code visual canvas. SAP took a stake at a $5.2B valuation in May 2026.

Ollama

Run powerful open-weight AI models locally with one command. Provides OpenAI-compatible APIs and GPU acceleration for Apple Silicon and Linux hardware.

Cline

Open-source autonomous AI coding agent for VS Code. Reads files, writes code, runs terminal commands, and uses MCP to control local development environments.

Explore More AI Tools

Discover other AI applications and tools.

Browse AI Tools← Back to AI Tools