What the workspace knows does not have to stay in one window. Point another AI at it and that AI can search what you have gathered, answer from it, and add to it.
The memory stays where it is. Nothing is copied into the other AI, nothing is installed on a laptop, and everything the AI does is on a record you can read afterwards. Connecting an AI to a workspace comes with the workspace.
Four steps
1. Have a workspace. If you do not yet, set one up. It takes about five minutes, most of it waiting. Give it something to read before you connect anything, or the AI will find an empty memory and say so.
2. Paste the address into the AI. One address, the same for every AI:
https://gbrain.io/mcpConnect an AI has the exact steps per AI, including where each one hides its MCP setting. The workspace’s Integrations page has them too, under Use GBrain in other AIs.
3. Pick the workspace and what it may do. The AI sends you back to GBrain to approve it. You choose the workspace, then Off, Read or Read & Write for its memory and for its schedules, and Off, Mine or All for the applications in it: none, the ones you added, or everybody’s. Read lets the AI search and recall. Read & Write also lets it remember new things, link notes, and forget. You can change all of it later, and the change takes effect on the AI’s next request. What an AI can reach when you connect a workspace lists everything the screen hands over, and what it never does.
4. Ask it something. See the asks below.
Per AI, the short version
Full steps are on the Connect an AI page. The one thing that differs per AI:
- Claude Code.
claude mcp add gbrain -t http https://gbrain.io/mcp, then sign in when it asks. The memory shows up as tools named after the server you gave it. - Claude, web or desktop. Add the address as a connector under Settings and approve the sign-in.
- ChatGPT. Add the address as a connector and approve the sign-in. Deep research works against the memory too, because it looks for the same search and fetch tools the workspace already publishes.
- OpenClaw. Two settings matter. Its default transport is the older SSE, so set
transport: "streamable-http", and useauth: "oauth". If several people talk to one OpenClaw, addoauth.identity: "per-requester"so each person signs in as themselves and the record shows who asked. - Hermes. Add the address under
mcp_servers. It signs itself in.
Try the memory from the AI
Once it is connected, these are the asks that prove it:
- “Search the memory for what we decided about pricing.” Reading.
- “Remember that Ana runs procurement.” Writing, so it needs Read & Write.
- “What do you know about this company already?” The reason to connect at all: the answer comes from the material, not from the model’s guesswork.
An AI will not reach for memory on its own for every question. Saying “search the memory” is the difference between an answer from the notes and an answer from general knowledge, and the two sound equally confident.
Good to know
One connection, several workspaces. An AI connects to one workspace to begin with, and the same connection can reach more. Add the second one from its Integrations page, under the same AI; Connect an AI has the steps under Two workspaces in one AI. Every tool takes a workspace argument, so the AI never has to guess which memory it is writing to.
Off until you connect it. Nothing listens on the workspace until you connect an AI to it. Take it off every connection and it switches off again.
Everything is on the record. Every search, every note written, which connection did it, and what it cost, on the Activity page.
The AI cannot widen its own access. If it needs Read & Write and has Read, it will say so and hand you the page to change it. Only you can change it.