In this OpenClaw crash course, I walk you through getting an autonomous AI agent running 24/7 on your own PC (or a VPS). OpenClaw (previously moltbot/clawdbot) is written in TypeScript by Peter Steinberger and now has official OpenAI backing as an open-source project, which is pretty wild. I show the quick-start install, the initial configuration, and how to choose an AI provider like OpenAI or Anthropic (I use Anthropic and pick Claude Opus 4.6). Then I demo both ways of talking to it: the terminal UI and the Web UI with a session token, plus setting identity so it remembers how to work with you. After that, I set up real-world messaging channels so you can chat with your agent from your phone. I configure WhatsApp via QR code and Telegram via BotFather (including grabbing the bot token and plugging it into OpenClaw). Then we hook OpenClaw into an MCP toolchain using Zapier as the “man in the middle,” so I can tightly control what tools it can access (like Gmail read + draft, but not send/delete). I also cover the big warnings: skills can be risky (I mention reports of honeybots), and API usage can get expensive fast—so I show how to run local models via Ollama and manage everything in a workspace you can open in VS Code/Cursor and even sync to GitHub.