How it works
Where your memory actually lives.
Nothing is hidden in a database you cannot reach. There is a folder, there is an index that decides which note answers which question, and there are the assistants you already use.
A folder on your Drive
Plain documents, in your own account. Not our server. We do not run one.
You own this
Query me
Decides what is worth keeping, keeps exactly one version of it, and finds the note that answers the question in front of it.
- ChatGPT
- Claude
- Gemini
↻ and whatever they learn goes back into the same folder
What it keeps
A shape you can actually picture.
This is roughly what a setup folder looks like after a few weeks of ordinary use. Every line is a document you can open.
Underneath
The parts that are easy to miss.
Storing text is the easy half. These are the behaviors that keep a memory useful after the first month, and they are closer to how a person remembers than to how a database stores.
It knows what is still true
A memory that only ever adds becomes a memory that contradicts itself. Query me separates what is true now from what used to be true, and keeps the old version only for as long as it still explains a decision you made.
It forgets, on purpose
Human memory prunes, and so does this one. Detail that has stopped earning its place is compressed or retired on a schedule, with a short record of what changed and why. The process is designed to make important changes visible.
It reads only what the question needs
Ask about one client and it opens that client’s note, not your entire working life. Less loaded means less of the conversation spent re-reading, and far fewer chances to drag something irrelevant into an answer.
Your assistants leave each other notes
When one of them changes something the others need to know about, it writes a short message into a shared mailbox. That is how three tools stay in agreement without you carrying messages between them by hand.
Ways of working live beside what you know
Your standing procedures (how you review, how you draft, how you decide) are kept as their own documents and picked up only when a task calls for them. Knowing something and knowing how you do something are stored separately, because they are used differently.
Important claims can be checked
When certainty matters, a note can record where a claim came from: something you said, something confirmed against a source, or something an assistant inferred. That gives you a trail to inspect and correct when the memory is wrong.
See what this looks like applied.
The mechanism above is the same regardless of what you use it for. The use-cases page shows what a folder like this actually does across a few different kinds of work.