Every AI piles up its own context. Cobrain takes it out of the individual chats: a place of your own where Claude, ChatGPT, Claude Code and the people you work with read and update the same notes.
It is not that the models have no memory: it is that the memory is locked inside somebody else's product. You cannot see all of it, you cannot correct it, you cannot take it with you.
Here the memory is yours: Markdown notes organised in folders, readable and editable by any MCP-compatible AI as well as by you. Change model or program; the context stays.
A branch — a folder, or a single note — is handed to a colleague. It appears in their tree and their AI reads it too. You work on it with Claude, they work on it with ChatGPT, both of you on the same notes.
You and your colleague work on the same context, even with different AIs. Neither of you starts over.
Markdown notes, exportable in one go and readable anywhere. Permissions are applied by the database: your AI sees only what you see, and a private branch stays out of reach.
Each one has a page of its own.
What Claude writes, ChatGPT finds. One memory, outside all of them.
ReadThe Pro planA branch to a colleague: it appears in their tree, and their AI reads it too.
ReadPortabilityMarkdown files in real folders. Export everything, reopen it with Obsidian.
ReadSecurityYour AI cannot do more than you: the rule is applied by the database.
ReadFor whoever pays for tokensRules first, then the card, then the notes. And it says what it left out.
ReadOn Monday you decide with Claude that deliveries go by express courier, never on a Friday. On Wednesday ChatGPT writes the email to the client. On Friday Claude Code checks the price list. Three AIs, one memory. Pick an AI and watch.
If you use one AI, forever, on your own, perhaps it is. But you cannot see all of that memory, you cannot correct it line by line, you cannot hand it to a colleague and you cannot take it away. A shared memory does exactly those four things.
No: it is the place where your AIs read and write. If you already have an Obsidian vault you import it, and what comes out of Cobrain reopens in Obsidian. The difference is that Claude, ChatGPT and your colleagues get in here too — the comparison line by line.
You can, but you do not have to. You connect your AI and ask it to remember the things that matter: it writes them, in Markdown, where you tell it to. And everything stays editable by hand.
Model Context Protocol: the standard way an AI connects to an outside tool. Claude, ChatGPT, Claude Code, Codex, Antigravity and many others speak it. For you it means one thing only: it connects by pasting an address.
It is free and no card is needed. You sign in with a link sent by email, connect your AI and start building a memory that does not stay locked inside a chat.
