Agent Memory

Chapter 1. The Work of Remembering

A note for Early Release readers

With Early Release ebooks, you get books in their earliest form—the author’s raw and unedited content as they write—so you can take advantage of these technologies long before the official release of these titles.

This will be the 1st chapter of the final book. The GitHub repo will be made active later on.

If you’d like to be actively involved in reviewing and commenting on this draft, please reach out to the editor at [e-mail].

Do you remember the first time you used ChatGPT? I can recall talking with my colleagues about how Promethean the technology was. I can remember how scary yet oddly normal interacting with it felt. But I can’t remember the specific moment I started using it, what I asked it, or even if the answer I was given was particularly novel. Like the millions of input and output tokens that I have passed through LLMs since, my first interactions with ChatGPT were almost certainly unremarkable.

What we want from these AI tools has expanded as fast as their capabilities have. We now expect them to use tools themselves, to carry context from one interaction to the next, to build on what they know about us and our work. But how does that actually happen?

The answer is memory. Not of the biological sort that holds our dearest, most ephemeral experiences, but something more mechanical and pragmatic. Agent memory is a system for storing, retrieving, updating, and forgetting the knowledge ...