ChatGPT speaks MCP. It has not run on our brain yet.
The brain is an MCP server, and ChatGPT takes MCP servers as connectors. We have not run ChatGPT against our own brain, so this page says only where the address goes and the rules ChatGPT would meet.
Memory for a team, not for one chat.
Memory inside one AI is one person's, kept where that AI keeps it. A brain is pages your team keeps, each space with its own people, read by every AI you connect, with the same pages in each.
- Read by that AI alone
- Nobody on the team reads it or corrects it
- Absent from the next AI you try
- Each space shared whole with its own people, and each AI reads as its person
- Corrected by whoever reads a page, with every version kept
- Yours as markdown files on your computers
Where the address goes.
ChatGPT adds an MCP server as a connector, under its settings.
- Add the connector. In ChatGPT's settings, with the address beside.
- Approve it. The brain's approval page asks what ChatGPT may do as you.
- Give it spaces. On Devices, pick the spaces ChatGPT reads.
ChatGPT would read only the spaces you give it, and never a sealed one.
The brain holds ChatGPT to the rules it holds every client to. A new AI reads nothing until you pick its spaces, and writing is a box you tick, so ChatGPT can stay read-only. A sealed space's key stays on its people's computers, where ChatGPT never goes.
ChatGPT and the brain, answered.
Does ChatGPT work with instibrain?
Has ChatGPT run on the brain?
Could a web AI read a sealed space?
How is this different from ChatGPT memory?
Start with Claude Code, then try ChatGPT on the same address.
The installer connects Claude Code on a linked computer; ChatGPT could take the brain's address.