wiki/infra.md revised
wiki/auth.md revised
wiki/sqlite-fts.md new page
2 learnings drafted
learning #41 helpful ×6 rises
learning #12 helpful ×3 rises
learning #07 misled ×2 sinks
LongMemEval-S · 500 questions
QA accuracy 94.2%
answer in top 5 98.3%
Most memory starts blank. I start with yours.
You've already done months of work with AI. That work is yours, so you bring it with you.
Scroll
One memory, from every tool you've used.
Your past sessions in Claude Code, Cursor, Codex, Antigravity and Kilo become memories on your own computer. Your tools' files are never changed, and you can delete any memory.
Then I keep learning, from every new session.
Parsers and file watchers pick up each new session within moments.
At night, I dream.
I turn raw observations into a wiki that I keep rewriting, not a pile I keep adding to.
I remember what helped.
When a memory helps you, it rises. When it misleads you, it sinks.
I check my own work.
If a change to how I search makes me worse at recalling your history, it doesn't ship.
Open a new tool tomorrow. I'm already there.
A short briefing lands in the file each tool reads before your first message.
Only you can read me.
I'm sealed with an age key that lives in your Keychain and nowhere else.
I live in your GitHub. Not on anyone's server.
One encrypted Release asset on your own private repo. If a company shuts down, nothing happens to me.
I'm awake.
One memory for every AI coding tool you use. One command installs me.
curl -fsSL https://raw.githubusercontent.com/rohithhn/Open-Agentic-Brain-and-Memory/main/install.sh | bash