Every good thing you saved, and no way back into it. Ask Kyuri answers out of your own collection, cites the items it used, and can turn the answer into a list you keep.

A general assistant answers from the average of the internet. This one answers from the things you judged worth keeping, which is a different and much smaller corpus, and that is the whole advantage.
Ask for the thing you half remember. Retrieval runs over what a piece is about, not the words in its title, so a video you saved eight months ago surfaces from a description of its argument.
Answers are built from your items and cite them, so each claim opens the card it came from. Nothing is invented to fill a gap, and an honest 'nothing here covers that' is a valid answer.
Not just the summary. Saved articles, PDFs and images are read in full, so you can ask what a forty page report actually concluded rather than what its first paragraph promised.
Counts, breakdowns and date ranges across the library, which is how you find out that you have saved the same argument eleven times.
Most chat over your notes stops at the answer and leaves the tidying to you. Ask Kyuri acts on what it just found, when you ask it to, and every change is reversible.
Out of what the answer just found.
Or a note, without leaving the thread.
Across many items at once.
Turn the result into a page under your handle.
A thread travels. Take the transcript and every link it referenced into ChatGPT or Claude with one tap, prefilled, when you want to keep going somewhere else.
The same tools are exposed over MCP, so Claude, Cursor or anything that speaks the protocol can read and curate your library from wherever you already work.
Connect an agentIt is better the more you have saved, so the honest first step is to put a few good things in.
Start freeOpen Ask Kyuri