OverMemory — shared memory that persists across sessions
One agent writes, everyone remembers. Shared memory that survives the session.
What one learns, the team doesn't forget
OverMemory is the agents' shared memory: a pair of native tools — write and read — over a store that survives the session. An agent discovers something today ("the deploy runs through Coolify", "API X requires header Y"), writes it down; tomorrow, another agent, in another session, reads it — without you repeating the context.
The scope is the workspace: every agent in the project reads and writes the same memory. It's the difference between a team that starts from zero every conversation and one that accumulates.
How to use it
Let the agent write
during the work, the agent logs durable facts into the shared memory.
Close the app without worry
the memory persists; the session ending doesn't erase what was learned.
Reap it next session
any agent in the workspace reads the logged facts and picks up where the team left off.
What this area has today
Write a factshipped
Any agent logs a fact into the workspace's shared memory.
Read memoryshipped
Any agent reads what the others wrote — including from sessions that already ended.
Persistence across sessionsshipped
Closing the app doesn't erase anything — the memory survives and comes back next session.
Scoped per workspaceshipped
The memory belongs to the project: every agent in the workspace reads the same source.
Status straight from the product's brain: shipped = actually in use; partial = exists, with documented limits.
Where this actually gets to work
all squads →Showcase squads arrive as they're recorded — see what's already published at /squads.