Personal memory stores facts about one user inside one assistant. Team memory stores what a group decided, with the source and date, and makes it available to every member's agents. Archilas gives AI a memory of the team, and a workspace shares one memory across its members.
Personal memory in a chat app is useful: it remembers your name, your stack, how you like answers. It belongs to one person in one product. Open a different assistant and it is empty. Ask it what your teammate decided and it has no way to know.
What team memory has to get right
- Sources. Decisions are spread across Slack, GitHub, Gmail, Notion, Linear, Calendar, Drive, and AI chats. One source is never the whole picture.
- Time. A later decision changes an earlier one. Both should stay, dated, so the newer one is visible and the older one does not silently win.
- Citations. An answer should say where it came from so a person can check it.
- Abstaining. If nobody wrote it down, the answer should say that.
- Deletion. Members must be able to delete notes, disconnect a source, and delete the account.
How Archilas does it
A workspace connects its tools on the Connect page. Each member's agents reach the same memory over MCP at https://www.archilas.com/api/mcp. Answers come with sources. Disconnecting a tool deletes the content synced from it. The details of what is stored and for how long are on the privacy page.
Related posts
Questions
- Can several people share one AI memory?
- Yes, with a shared store. In Archilas, a workspace's members share one memory, and each member's MCP clients can query it.