Master.dev – Personal Agents with OpenClaw

What You Will Master

• Understand that a personal AI agent is delegated authority, and learn how to deliberately scope both its power and its risk.
• Learn why local-first is not secure by default and how to actually lock down an assistant running on your own hardware.
• Build a clear mental model of OpenClaw’s personal-assistant trust model so you do not mistake it for multi-tenant isolation.
• Apply least privilege across three axes: who can talk to the bot, where it can act, and what it can touch.
• Read and interpret openclaw doctor, security audit, and security audit –deep output with confidence instead of treating green checkmarks as magic.
• Know when to reach for a tool, a skill, or a plugin as you expand your assistant’s capabilities.
• Leave with a reusable security checklist, a custom skill, and a working capstone automation you can keep evolving after the workshop.

Prerequisites

• Basic comfort with the terminal, including running commands, editing config files, and reading command output.
• Some exposure to AI coding tools such as Copilot, Cursor, Claude Code, ChatGPT, or similar tools.
• No prior experience with agent architecture, OpenClaw, self-hosted assistants, or secure automation is required.
• Come with Node.js LTS, npm, Git, and a code editor installed and working on your machine.
• OpenClaw should be pre-installed if possible.
• Bring an API key for at least one model provider with a low spending limit.
• Telegram and Tailscale are helpful but optional; simulated alternatives will be available.