Create a team
There is no separate team install. Set up normally, then:
Open the dashboard as the owner.
Go to the Team tab and rename the team from the default; the name appears in every member's app and in the activity log, so do it before anyone joins.
Add your first member, described next.
Team mode turns on the moment a second person joins, and the dashboard shows the team features from then on. To look around first, the Team mode switch on the same tab gives you the shared layer and the sharing controls while you are still alone.
What you will see: the Team tab shows the roster and Add member. The switch is locked on while anyone else is on the team; once you are the only person left, it unlocks again.
Invite someone
On the Team tab, choose Add member.
Enter their name, optionally an email, and pick a role.
Member reads and writes their own personal layer, reads the shared layer, and can share their own memories into it.
Admin adds managing people and the activity log.
Copy the sign-in token immediately.
Send the invite yourself, by email, chat, or however you normally talk to that person. The app gives you an invite message to copy, with Copy invite message, which fills in your Worker URL and the instructions; Second Brain does not send email.
The token is shown once because the brain stores only a hash of it, a fingerprint rather than the token; there is nothing kept that could be revealed later. If it is lost, an admin uses Reset token to issue a new one, invalidating the old.
They sign in with the Worker URL and that token, in the desktop app or the dashboard, and get their own personal layer immediately.
The wiki mentions an email option for invites, but the v3.0.0 source stores a member's email and contains no email-sending code. Second Brain does not send email; the invite message is yours to copy and deliver privately.
Personal layer and shared layer
Every brain has two layers. A member joining an existing brain sees none of anyone's personal history; their personal layer starts empty, and they see only what is already placed deliberately in the shared layer.
| Personal | Shared | |
|---|---|---|
| Who can see it | Only you | Everyone on the team |
| Where it lives by default | Everything lands here | Nothing, until you share it |
| Called in the CLI and API | personal | company |
| Reachable through any client | No, private to you | Yes, for every member |
What admins can and cannot see
Admins manage people and roles, reset tokens, suspend, restore and remove members, rename the team, set capture defaults, run integrations, and review the activity feed. None of that includes reading anyone's personal layer. Admins can see a count of another person's private memories, for offboarding, and never the content. Nobody can read anyone else's personal layer, including you.
Members see the team name and the roster's names and roles. They cannot add, suspend or remove people, and they cannot see the activity log.
Shared on purpose
Open any memory and choose to share it. This is a move, not a copy: the memory keeps one ID, one history, and one set of connections. Everyone can then find it, and you can move it back to personal at any time. Only the memory's author or an admin can edit, delete, change its status, or move it back once it is shared.
SavedThe launch checklist lives in the shared layer now, so the whole team can recall it.
Each person also has a capture default, set on the Team tab: Personal keeps everything they capture private, Company sends everything to the team, and Auto follows the organization default. The composer shows which one applies before you capture, so Auto is never a guess.
Suspend and remove someone
Both are on the Team tab, and admins only.
| Suspend | Remove | |
|---|---|---|
| Access | Blocked, reversible | Revoked permanently |
| Their personal memories | Stay put | Deleted; this cannot be undone |
| Anything they shared | Stays, still credited | Stays, still credited |
You cannot remove your own account or the last active admin, because the brain would have nobody left to manage it.
Limits
- One team per brain in v3.0.0. The API and MCP tools already carry optional team parameters for the future, but the dashboard cannot create or switch between teams.
- No enforced limit on team size. The practical limit is your Cloudflare plan's budgets. A handful of people is comfortable, and a small team can often stay within free-tier budgets.
- Members do not need Cloudflare accounts. Only whoever deploys and updates the Worker does, and a member cannot update it; they are simply told when the brain is behind.
Converting a personal brain
Your memories are safe: the upgrade is additive, and nothing is moved between layers, deleted, or made visible to anyone.
Upgrade the Worker to a version that supports teams. Nothing else changes yet.
Open the dashboard once as the owner. On the first request after the upgrade, the Worker migrates its own schema: it creates the team tables, gives you an owner identity, and files every existing memory into your personal layer. Your password keeps working, and you do not sign in again.
Check that your memories are present and search works as usual.
Name the team, then add your first member.
Going back to solo
Remove everyone else: that revokes their access and deletes their personal memories, while anything they shared stays, still credited. With only you left, the brain behaves as a solo brain again, and you can switch Team mode off on the Team tab.
Turning it off hides the layer controls; it deletes nothing. Memories already in the shared layer stay where they are and remain readable, simply unlabeled. Turn it back on and the labels return.