Join your team's brain, then connect your AI.
Your team sent you an invite link. Accept it, run one installer to bring the brain to this computer as plain markdown files, and connect the AI you work with. The steps below are in order.
Your AI can run these commands for you. The invite and the link code need you, because you sign in to get them.
Four steps put the brain on this computer.
Run them on the computer you work on. The installer needs nothing set up first: it brings Node itself when the computer has none.
- Open the invite link. The page names your team and who invited you. Enter
your name, choose a password of at least 12 characters, and join: you land
signed in at
my.instibrain.ai. If you already sign in to instibrain, your password stays as it is. The link works once, for a week. - Run the installer. On a Mac or Linux, in a terminal:
curl -fsSL https://instibrain.ai/install.sh | bash. On Windows, in PowerShell:irm https://instibrain.ai/install.ps1 | iex. It puts the sync client in~/.instibrain, makes the folder~/instibrain, and asks for a link code. Run it again any time to update. - Link this computer. In
my.instibrain.ai, open Computers, choose Link a computer, name it, pick the brain it syncs, and get a code. Paste the code where the installer asks. It works once, for 10 minutes. Skipped it? Runinstibrain linkwith a new code in a new terminal. - Sync.
instibrain syncbrings the brain into~/instibrain/memory: each page you may read, as a markdown file. The installer runs the first sync itself, checks each step, and puts an instibrain launcher, with the logo, on your desktop. Leaveinstibrain watchrunning to sync each save as you make it.
Your AI connects with one command, and reads CLAUDE.md first.
The brain is an MCP server. Add it to the AI you use, approve it once, and the AI reads the pages your team shared with you.
- Add the brain. In
~/instibrain, run the Claude Code command in the panel; without Claude Code, the installer's check names its install line. Codex, Cursor and ChatGPT take the same address; Agent gives each one's line. - Approve it once. The first call opens a page in your browser. Approve it there, signed in as yourself, and the AI reads what you may read, no more.
- It reads CLAUDE.md first. Over MCP, the brain tells every agent to read
CLAUDE.mdat the root of the brain before anything else: your team's rules for agents. In the folder,~/instibrain/CLAUDE.mdimports the brain's own, so Claude Code loads it with each session. The instibrain launcher on your desktop starts Claude Code there. - Allow writing when you want it. Reading comes with the approval. Writing back to the brain is a second approval of its own.
Over MCP your AI reads the open area of the brains shared with you. An encrypted
brain opens only on a computer that holds its key: there, your AI reads the files
instibrain sync wrote. The limits are
stated in full.