Every new AI starts from what is already in the memory
Retraining each new AI is the job nobody wants. This is one memory, written once and synced to every AI you use.
gbrain.io
Retraining each new AI is the job nobody wants. GBrain keeps one memory, in plain files on a machine of its own. Write it once, sync it to ChatGPT, Claude, Cursor or any AI you choose, and each one starts where the last left off.
One memory, one address. Every AI you plug in reads the same files and writes to them, so a correction typed into one is a correction in all of them. Each gets Read or Full, set per AI, and switching one off stops it reading and changes nothing about the notes.
| Client | Status | Last used |
|---|---|---|
| Brad's MacBook | On | 4m ago |
| Claude Code | On | 22m ago |
| Cursor | Not set up | Never |
| Studio Mac Mini | Off | 6d ago |
One markdown file per thing it knows, on a machine of its own. Open any of them, change a line, and every AI synced to it answers from it. Copy the folder and that is the export: nothing was ever locked up. Keep it to yourself, or share it with a whole team.
This one is measured.
On LongMemEval, the standard test for long-term memory in AI, GBrain put the right conversation in its top five results for 95% of questions (449 of 470). The run, the settings and the scorecard are published in the open, including the twenty-one it missed.
Search finds a note by meaning as well as by its words, across the whole memory rather than a prompt's worth of it, and Lately shows what the memory learned this week, by day and by person.
Sign in, and you have a folder of notes that every AI you plug in can read and write. There is nothing to move over, and if you ever want out you copy the folder and go.
