Memory Card
An agent's page carries a Memory card: a read-only view of what that agent can recall from the shared memory store, and how actively it has been writing into it. Nothing on the card creates memory — entries are written by the agent as it works, or pinned from a session — the card only reports.
The card has three sections.
Recall by Scope
Entry counts for everything this agent can currently see, bucketed by scope:
| Bucket | Counts |
|---|---|
| Private | The agent's own agent-scoped entries. |
| Session | Entries bound to its live session. |
| Workspace | The workspace's shared entries. |
| Project | Entries of the project resolved from its session's working directory. |
| Global | Entries shared everywhere. |
A Total row closes the list. Buckets with nothing in them are omitted — the card shows real counts only, never a fabricated zero. When the agent can recall nothing at all, the section shows a short explainer of what memory is instead, with a link to learn more.
Freshness
One bar per lifecycle state — Active, Stale, Archived — showing each state's share of the agent's recallable entries, with the count beside it. Unlike the scope section, genuine zeros are kept here: "0 archived" is honest data in a distribution. With no entries at all, the section says freshness will appear once the agent has memories.
Freshness is the durable, forward-only lifecycle described in Versions & Deletion — the same state that makes stale entries render dimmed on the Neural Map.
Memory Writes
A day-by-day heatmap grid of this agent's memory writes — the same grid geometry and color ramp as the Consistency card, pointed at a different substrate. Every write counts: new entries, edits (superseding versions), and deletions (tombstone versions) each weigh one, because all three put a version in the store. Other agents' writes are not counted; this is one agent's hand on the store.
Until the agent has written anything, the section says "No memory writes recorded yet."
Where to go next
- Memory — the store the card is a window onto.
- Scopes — why the buckets are exactly these five.
- Capabilities & Skills — the neighboring card describing what the agent can do, rather than what it knows.