What is OpenClaw and Why Should You Care?
OpenClaw is an open-source AI agent framework that runs entirely on your hardware, acting as a proactive personal assistant to automate tasks across files, applications, and messaging platforms. Unlike cloud-based AI assistants, OpenClaw keeps your data local while connecting to powerful models like Anthropic Claude, OpenAI GPT-4o, Google Gemini, or even local models through Ollama.
The platform has exploded in popularity with over 199,000 GitHub stars and a thriving community building skills for everything from Gmail automation to smart home control. Whether you want to automate code scaffolding, manage your calendar, control Philips Hue lights, or clear your inbox through WhatsApp, OpenClaw delivers.
Prerequisites Before You Start
Before installing OpenClaw, you'll need three things on your machine:
- Node.js v22+ LTS — Download from nodejs.org and verify with
node -vin your terminal - Git — Install from git-scm.com (Windows) or via your package manager (Linux/macOS)
- Admin/sudo access — Required for installation and running the daemon
Installation Steps
Windows (PowerShell as Administrator)
1. Open PowerShell as Administrator
2. Visit openclaw.ai and copy the official installer script
3. Paste and run the one-liner; approve any security prompts (select "Yes")
4. During onboarding, choose "Quick Start" and add your API keys (Anthropic, OpenAI, or Ollama)
5. Skip optional channels and skills initially — you can add them later
6. Access the Web UI at http://localhost:18789
Linux and macOS
Run this command in your terminal:
curl -fsSL https://openclaw.ai/install.sh | bash
If you get a permission error, prefix with sudo. If npm isn't in your PATH, add this to your ~/.zshrc or ~/.bashrc:
export PATH="$HOME/.npm-global/bin:$PATH"
Run onboarding with: openclaw onboard
Configuring Your AI Provider
OpenClaw is model-agnostic — you choose which AI to use. The recommended setup uses Anthropic Claude Opus 4.6 for the best balance of capability and cost, but you can also use:
- OpenAI GPT-4o or GPT-5.2
- Google Gemini 2.0
- DeepSeek V3
- Local models via Ollama or LM Studio
- OpenRouter for unified API access
The model-agnostic design means OpenClaw auto-switches to backup models if your primary choice fails — critical for production automation.
Adding Skills and Channels
OpenClaw's power comes from its skills system. Skills are modular extensions stored as Markdown files in ~/.openclaw/workspace/skills/. You get 100+ prebuilt skills out of the box, plus access to 700+ more through ClawHub.
Popular skills include:
- Gmail — Send emails, search inbox, manage labels
- GitHub — Create issues, PRs, manage repositories
- Spotify — Play music, control playback
- Philips Hue — Control smart home lights
- Crypto trading — Execute trades on exchanges
Add channels to control OpenClaw from messaging apps like WhatsApp, Telegram, or Slack. This lets you text your AI agent from anywhere.
Using OpenClaw in Practice
Once running, you can interact via the Web UI, terminal, or connected messaging apps. Give it natural language commands like:
- "Scaffold a React app with Tailwind CSS" — it creates directories, edits files, runs npm
- "Find all emails from [sender] last week and summarize them"
- "Turn on the living room lights at 7 PM"
OpenClaw maintains persistent memory using Markdown files for conversations and SQLite with vector embeddings for semantic search across your chat history.
OpenClaw vs Claude Code: Which Should You Use?
If you're wondering whether to use OpenClaw or Anthropic's Claude Code, they serve different purposes. Claude Code is a specialized coding agent optimized for software engineering tasks — refactoring, tests, PRs — with deep IDE integration. OpenClaw is a general-purpose life assistant for broader automation.
Use OpenClaw for self-hosted flexibility in non-coding automation. Use Claude Code for enterprise coding needs where SOC2 compliance matters. Many developers use both: OpenClaw handles life tasks while Claude Code focuses on software builds.