mthines

Fair criticism. To be precise: I dictated the outline, specific sections and ideas, then used Claude to draft and rephrase sections and iterated on it with Claude. I didn't think it mattered, as the points are mine - but I can see people care so I should have said so in the post. :)

I will take this into account for future posts.

Hey, author here. I wrote the blog post and of course got assistance by AI to rephrase sections, gather data and do comparison, and make graphics - also because I'm no native english speaker and wanted to improve the writing.

The sections, points, ideas are by me :)

A couple of weeks ago I built an open source system for handling memory for agents, and it's called LoreKit.

It supports local and remote (so you can share with colleagues for instance) memory, it's free, open source and REALLY easy to set up.

Local memories requires no sign up, but remote does obviously - but sign up is just OAuth.

It's retrieving memories lexically - so no embedding (for now), but using scopes and labels to make sure the memories you retrieve are relevant automatically. And it scales from 6 - 60.000 memories (load tested).

I've also built a website which reads usage statistics for you to help you understand and optimize your workflowz and 3 agent skills to ease the installation and integration into your existing skills. https://lorekit.io

I've written a blog post on how to get started and some of the problems it tackles, but I would like to hear your thoughts on it, like:

- what's currently holding you back from using memory which persists between sessions? - do you prefer local, local (but committed) or fully remote?

See repo: https://github.com/mthines/lorekit