Memories
An AI memory
you can read.
Connect Claude, ChatGPT or Claude Code to Virlow and they remember what matters between conversations: the decisions you made, how you like things done, who works on what. You never have to repeat yourself, and you never have to wonder what was kept.
Every memory is an encrypted note in your vault. Read it, edit it, pause it or throw it away, like anything else you write.
AI memories are off until you turn them on in Settings. The people and projects in these pictures are invented.
01 / The graph
See what it knows,
and how it connects.
Each memory names the people, projects and tools it is about, and how they relate: Nadia leads Harbor, Harbor uses Postgres. Virlow draws them as a graph, coloured by kind, so the shape of your work is one look away. Open an entity to see everything true about it, and the notes that mention it.
Drag the timeline back to see what was true on a date. When a fact changes, the new one ends the old one, and the old one stays as history instead of being overwritten.
02 / What I learned
Trust it without
checking every chat.
Memories are saved as your AI works, so there is one place to see what it picked up: What I learned lists this week's memories by day, and the facts that stopped being true.
Anything it got wrong is one tap from gone. Undo moves a memory to the Trash, where you can restore it, and every memory opens in the editor like a note.
Two switches stop saving: one for everything, one for the project you are in. Recall keeps working while saving is paused.
03 / Entity pages
Everything about
one person or project.
Every person, project, tool and place gets a page: what is true about it now, what was true before and when it changed, the memories behind each fact, and the notes that mention it.
Your AI reads the same page before it works with someone or on something by name, so it starts from what you already decided.
04 / How it works
Automatic, once
it is connected.
In Claude Code, the Virlow plugin loads what the project already knows when a session starts and saves what it learns as it goes. Memories are grouped by project, named after the repository.
In Claude, ChatGPT and other apps connected through AI connections, the assistant recalls before it acts and saves what will still be true next week.
- No secrets, ever. A memory that contains a password, a key or a card number is refused before it is written.
- Nothing hidden. Memories are ordinary notes in a Memories workspace, with who saved them and when.
- Tidy on demand. Merge two spellings of one name and link older memories, from the Tidy tab or with one prompt.
- Free to use. Saving and recalling memories, What I learned and entity pages are on every plan. The graph, the timeline and renaming or merging entities are part of Pro.
Where memories are read
Memories are encrypted on your device like every note, and the graph is drawn there too. An AI app can only read them through a connection you set up, and only while your vault is unlocked. With the local connector the key never leaves your computer; the optional hosted connector decrypts on Virlow's server while you have it unlocked, so web apps can use it. Running code is the other optional feature that sends content off your device: code execution runs your code on Virlow's service. The security page has the details.
Start with one fact