OpenClaw full install guide – with Skills tutorial

This is a full OpenClaw install and usage guide: openclaw local deployment and openclawd-style setup. Follow this openclawd guide to go from zero to running with model, channels, and Skills.

Why OpenClaw?

OpenClaw is an open-source AI Agent framework that can work 24/7 on your machine. Unlike pure chatbots, it has “eyes and hands”: it can drive the browser, write code, read files, run commands, and even run tasks while you’re away. It improves over time, can run on a schedule, and is fully customizable.

Local vs cloud

Local: simpler setup (one command), data stays on your machine, better integration with your apps and files, and full features (e.g. desktop control). Cloud often gives only a fraction of that. You can run it on an old PC, Raspberry Pi, or a spare machine; upgrade to Mac mini or better if you need more power.

Install and config: three steps

Step 1: Install

Go to openclaw.ai Quick Start, copy the install command, run it in Terminal (Mac) or Command Prompt (Windows). Options: one-line script, npm install -g openclaw, or macOS .dmg app.

Step 2: Choose model

After install, the wizard asks for an AI provider: e.g. Anthropic Claude, OpenAI, or a cheaper option like Minimax. Paste your API token as a single line with no extra line breaks.

Step 3: Message service

Talk to OpenClaw via your usual apps: iMessage, Telegram, Discord. Telegram is a popular choice for its threading and readability. Follow the prompts to connect.

Skills tutorial

Skills are like apps for your agent. Install from ClawHub or the community to add web search, summarization, browser automation, email, calendar, etc. Typical flow: search for a Skill → install → set API/permissions if needed → use it in chat.

If you prefer not to run the above yourself, use a hosted OpenClaw-based product for zero config and instant use in the browser or desktop.