An AI "agent" isn't just the model (LLM), it is also the code that interacts with the user, the environment, and the LLM: the harness.

We use harnesses every day: GitHub Copilot, Claude Code, ChatGPT, etc.

When you build apps that _integrate_ AI, you will probably write your own harness code that calls the LLM when appropriate.

In this session you will learn how to create a robust harness that includes user interaction, cross-app interaction, long-term memory, skills, tool calling, and other behaviors a harness might implement.

Learning objectives:

1. What is a harness and what might it do?
2. How to manage user interaction and LLM "turns"
3. Leveraging the environment for memory, skills, calling external services