TL;DR: A practical guide for non-technical people who want to work with AI agents. Part A covers the basics — what agents are, how they work, and how to set one up. Part B shows the recommended process and how to build your own agent system. Part C covers building and deploying software. No coding background required.

This guide is for people who aren't developers but want to understand — and use — AI agents. Not just for coding. Agents can write content, organize data, manage projects, and automate repetitive knowledge work. No technical background required.

A note before you start: this space is moving fast. Tools change, interfaces get redesigned, new capabilities show up every few weeks. This guide is a snapshot — it captures how things work at the time of writing, but some details will drift. The good news is that the concepts (how to think about agents, how to direct them, how to structure your work) are durable even when the buttons move around. And when something specific is out of date, you can always ask Claude or any other AI agent — they're surprisingly good at explaining their own tools. That's actually why I wrote this guide: to structure the answers to questions I faced and wished someone had organized for me.

The guide is organized in three parts:

Part A: The Basics — What agents are, how they work, and how to set one up. Read this first regardless of what you plan to use agents for.

  1. What AI Agents Are (and Aren't) — What AI agents are, how to direct them, and what engines (models) power them.
  2. Memory and Context — Context windows, what gets lost between sessions, and external memory layers.
  3. Tools of the Trade — Chat interfaces, all-in-one platforms, and agent-assisted tools. The landscape and how to pick.
  4. Setting Up — Installing Claude Code, configuring VS Code, and running your first command.
  5. Git and GitHub — Version control: saving your work, going back to previous versions, and backing up online.
  6. Configuring Your Agent — Instruction files, rules, memory, skills, commands, and MCP servers.

Part B: Putting It to Work — The recommended process for agent work, and how to build your own agent system from scratch.

  1. The Meta-Process — The recommended structure for agent work: orchestrator, explorer, actor, reviewer, iterate, post-mortem, change process.
  2. Content Writing Example — The meta-process in action: a flag blog with two agents, two skills, and one command. Full build walkthrough.
  3. Writing Effective Instructions — The craft of writing instruction files: conditional logic, sub-agent dispatching, orchestration, gates, skill sizing, and external memory.
  4. Personal AI Agents — OpenClaw and the world of agents that run 24/7 on your own hardware, communicate through messaging apps, and act autonomously.

Part C: Building Software — For those who want to build and deploy applications with agents.

  1. Building Software with Agents — The same process at scale: multiple agents, spec files, sprint files, and gates for human approval.
  2. Claude Built-in Capabilities — The default slash commands that ship with Claude Code and how to use them day to day.
  3. Roy's Claude Config — A real Claude Code setup: agents, commands, rules, and links to other public configs. Take what's useful, build your own.
  4. Servers, Hosting, and Deployment — Where your project lives and how people access it.

Each chapter ends with a Practical tips section — short, concrete advice you can apply immediately. You don't need to read those tips upfront. They're there when you need them.

Glossary — Every technical term introduced in this guide, collected in one place. Not a chapter — an addendum you can flip to anytime.