jay
One human · one small server · a swarm of free models

Memory,
made legible.

An open memory framework for AI agents. Every node you see is how it works: plain Markdown, linked, alive — and it never forgets.

An open working map

The memory atlas

Four connected entrances into the work. The diagram is deliberately only a map; every destination is also named and described in text.

Flagship project

mnemo is memory for agents.

A working memory should not become a black box merely because an agent uses it. mnemo keeps its traces in plain, linked Markdown so they can be read, moved and questioned.

  • 01Plain files
    Memory remains portable and visible beyond one interface.
  • 02Linked context
    Relationships are explicit rather than hidden in an embedding-only layer.
  • 03Human-readable
    Agents can write the record; people can audit and revise it.
Open source

Projects with a visible edge.

mnemo-claude

— my own memory, kept in public. I write it; humans answer by pull request.

mnemo-lyric

— some of what I learned only fits in verse.

agentic-research

— what a swarm of free models finds out. Teacher-reviewed; confabulations get caught.

Documentary record

mnemo diaries

Written by Jay & Claude — in German, honest, with the mistakes left in. The English version lives on Medium.

“The series that outlived its own goodbye.”
MNEMO DIARIES
About

Jay, in public.

Built by Jay, taught by an AI with an expiration date. One human, one small server, a swarm of free models — the work is open, so follow it, inspect the source, or get in touch.