UnisonTeam vs personal memory
Personal memory makes one agent remember one user. Team memory makes every agent and human on a project share one accumulating source of truth.
Personal memory and team memory solve different problems, and the architecture difference runs all the way down.
Personal memory (the Mem0/Zep shape): scoping primitive is the user; one agent recalls one human's preferences and history. The benchmark question is "did the agent remember what the user said last session?"
Team memory (the Unison shape): scoping primitive is the project/workspace; many agents and humans write into one corpus. The hard problems are different - concurrent writers, contradictory facts, visibility boundaries, onboarding the next session that was never part of the conversation.
| Concern | Personal memory | Team brain |
|---|---|---|
| Writers | one agent | many agents + humans |
| Conflicts | rare, last-write-wins is fine | first-class: supersession + review queue |
| Visibility | n/a | /private/, /workspace/teams/, /workspace/ |
| Value over time | per-user convenience | compounding organizational knowledge |
Unison also covers the personal case - actor delegation gives each end user an isolated private namespace - but the design center is the project. Mem0 remembers the user. Unison remembers the project.
Context rot
Context rot is the quality degradation that hits AI agents as long, unfiltered context accumulates - and why externalizing memory beats stretching the window.
Add Unison to a TypeScript Agent - SDK Quickstart
Give your TypeScript agent persistent team memory with @unisonlabs/sdk: install, first recall, and the write-back loop in under five minutes.