GBrain gives a team, and every AI it uses, one shared memory. Most of what a team knows is already written down somewhere: reports, meeting notes, years of exports in a folder on somebody’s computer. That folder can now go into memory, all of it, in one sitting rather than a hundred small turns in a chat.
The reading is done by the AI you already pay for. Claude Code or Codex, running on the computer that holds the folder, reads each file and saves a page for it. It works on your own account, and a thousand average documents cost it roughly $4 to $8 on a small model, $40 to $80 on a mid-sized one. Your workspace pays only to index what arrives, a fraction of a cent a document.
Documents, not pages
You get a count you can check against the folder: “412 of 1,003 documents in, 6 skipped”. A document too long for one page is saved in parts and still counts once. What could not be read (a scan that is only a picture, a file that is locked or empty) is listed at the end with the reason, rather than saved as a blank page.
It starts again where it stopped
A run over a thousand documents will stop sometimes: a closed laptop, a limit hit, a bad file. Running the same instruction again carries on from the folder instead of starting over, and nothing is saved twice. A document that changed since it was first read is saved again, in place, so memory holds the newer version. Nothing is ever deleted from memory by an import.
What you need to do
Read Bring a folder of documents into memory. It has the one line that connects your AI to the workspace, the instruction to paste, and the check at the end that tells you every document arrived. For up to about thirty files, dropping them into the workspace’s chat is still the quicker route.