Own Your Agents

Build AI agents that improve and move between ChatGPT, Claude and Gemini.

The agents you build are yours. They hold how you work, what you know about your clients and the hundred small corrections that make their output sound like you instead of like everyone else. That is intellectual property, and it’s quickly becoming the thing that sets one consultant or operator apart from the next.

Most people build that IP inside one vendor’s product: a Custom GPT, a Claude Project, a Gem. It works until the vendor changes the price, retires the model, or a competitor ships something better. Then you find out your agent was never really yours. It was a settings page.

This course teaches the other way. You keep each agent as a handful of plain files you own: instructions, skills, knowledge and tools. You improve those files every time the agent gets something wrong, and you run them in ChatGPT, Claude or Gemini depending on which one is best for the job that month.

It is the approach behind my own agent kit, which runs a small team of agents for each of my businesses. I’ll show you real files from it, how I use Buzz as the place where the agents work and get corrected, and how the same agent moves from one provider to another.

What you’ll learn

  • Why your agents are an asset to develop, not a subscription to rent
  • The five parts of a portable agent, and which part locks you in
  • How to write instructions that work on any model
  • Skills and tools that travel: SKILL.md, MCP and plain scripts
  • The improvement loop, and how I run it in Buzz
  • Setting the same agent up in Claude, in ChatGPT and Codex, and in Gemini
  • How to switch providers in an afternoon, and test that nothing got worse

Who it’s for

Independent consultants and operators who already use AI for client work and want what they build to compound. You don’t need to be a developer. Lessons 6 to 8 each have a chat-app track and a command-line track, so pick the one that fits.

The course has nine short lessons. Each ends with one thing to do, and by the end you’ll have one agent of your own, in a folder you control, running on at least two providers.

What's inside

The first lesson is open to everyone. The rest open when you join the course.

  1. The anatomy of a portable agent 3 min
    Locked
  2. Writing instructions that work on any model 3 min
    Locked
  3. Skills and tools that travel 6 min
    Locked
  4. The improvement loop, and how I run it in Buzz 4 min
    Locked
  5. Running your agent in Claude 3 min
    Locked
  6. Running your agent in ChatGPT and Codex 3 min
    Locked
  7. Running your agent in Gemini 2 min
    Locked
  8. Switching providers without starting over 3 min
    Locked