Memory scoping
Also called: memory visibility, scope, organization-scoped memory
What is memory scoping?
Memory scoping is the rule that determines who a stored memory is visible to. A memory can be scoped to one person, one machine, one project, or a whole organization. The choice decides whether what one agent learns can ever reach a teammate agent, so it is a structural decision rather than a setting.
Where the boundary can sit.
Each level solves a different problem, and the level chosen sets the ceiling on how useful the corpus becomes.
Machine scope
Memory lives with the device. Simple, and it disappears when somebody switches laptops, which makes it the least durable option.
Individual scope
Memory follows one person across devices. Better, and still means every teammate rediscovers the same things independently.
Organization scope
Memory belongs to the team. What one developer agent works out is retrievable by another, and it survives both device changes and staff changes.
Why the default matters.
Scope is the difference between a private notebook and institutional knowledge.
Sharing has to be the default
If sharing is opt-in per memory, it does not happen. Virex Memory is organization-scoped by default, with project attribution on each entry.
People leave
Memory scoped to an individual leaves with them. Memory owned by the organization stays, which is usually the point of writing it down.
Isolation still applies
Broad scope inside your organization does not mean broad scope across organizations. That boundary is enforced in storage with row-level security.
Commonly confused with.
Three related but distinct ideas.
Memory scoping questions
What scope does Virex Memory use?
Organization scope by default, with per-project attribution and subagent attribution recorded on each memory. A teammate agent can retrieve what yours wrote.
Can another organization ever see our memories?
No. Isolation between organizations is enforced at the database layer with row-level security, so a query scoped to one organization cannot return another one data.
What happens to memories when someone leaves?
They stay with the organization. Removing a person seat ends their access, it does not remove what their agents contributed.
Can I tell which project a memory came from?
Yes. Memories carry project attribution, so a corpus spanning several projects stays navigable.
Is memory scoped per client tool?
No. Memory is reached over MCP from any compatible client, so it is the same corpus whether an agent connects from Claude Code, Cursor, or another tool.
Memory that belongs to the team.
Virex Memory is organization-scoped by default, so nobody rediscovers what a teammate agent already worked out.