What is a team brain?

A team brain is a persistent, shared knowledge graph that every AI agent and human on a project can read and write - so no session starts from scratch, and no context is re-explained twice.

Individual agent memory already exists: tools like Mem0 and Zep make one agent remember one user across sessions. A team brain solves a different problem. On a real project, context is produced by many actors - your Claude Code session, your teammate's Codex session, a CI agent, the humans in between - and today that context dies in whichever session produced it.

What changes with a team brain

• Decisions persist past the session that made them. An agent that chose an architecture writes the decision once; every future session recalls it instead of re-deriving (or contradicting) it.
• Agents stop disagreeing by accident. When Agent A and Agent B write contradictory facts, a team brain treats the conflict as a first-class object to resolve - not last-write-wins.
• New sessions onboard in one call. Context recall returns the relevant slice of project memory as markdown, ready to inject into any model's context window.
• Humans and agents share one source of truth. The same corpus answers your teammate's question and your agent's recall query.

Team brain vs. personal memory

Mem0 remembers the user. Unison remembers the project.

Start with the Quickstart, or see how it plugs into Claude Code.