HumemAI iconHumemAI

About

An open source organization for AI memory.

HumemAI grew out of research into human-like memory systems for AI. Today it is an open source organization that publishes that research, the code behind it, and the benchmarks that test it.

Illustration of connected memory structures and stored knowledge

Overview

Memory should be a real system layer, not a prompt trick.

HumemAI is built around a straightforward idea: conversations and data are not the same thing.

Agentic systems need memory structures that stay structured, inspectable, and durable over time.

How it works

Open licenses and public repositories.

Every HumemAI project is published under an open license on GitHub, and the Python packages are released through the HumemAI organization on PyPI. Anyone can run, modify, and redistribute the code.

Origin

From research into usable systems.

Portrait of Taewoon Kim, who started HumemAI

HumemAI was started by Taewoon Kim, an AI researcher and engineer working on agentic memory, and is maintained together with the contributors to each project.

The organization grows out of research on human-like memory systems and explicit memory architectures for AI, with a practical goal: turn those ideas into software that anyone can use to build reliable agents.

Next

Explore the projects behind the work.