Pinned Memory
A pinned rule or data mapping is a durable line of project memory that every AI provider recalls next session, not just the one you pinned it from.
1. Pin a rule or a data mapping
Right-click a running terminal — or open its own Settings panel — and choose Pin as Project Rule… or Pin as Data Mapping….
You will see: a dialog pre-filled from any text you had selected in the terminal, with a placeholder showing the kind of line expected — for example "Always add the id field to every model" for a rule, or "user.uid → auth.users.id" for a data mapping. Its note explains where the line goes: saved into the project and recalled by every AI provider from there on. Do not paste secrets — this is a file meant to be committed and shared.
Choose Pin to save it, or Cancel.
2. If it looks like a secret
You will see: if the line matches a recognizable secret pattern, a "Possible secret detected" warning names what matched and asks you to confirm. Cancel is the default; Pin Anyway is the explicit, destructive opt-in for a false positive.
3. Review what's pinned
Choose Manage Pinned Memory… from the same right-click menu or Settings panel.
You will see: a Pinned Memory panel with a Pinned section (every rule and data mapping you've confirmed, each editable or with an Unpin action) and, only when your agents have proposed something themselves, a Proposed by agents section — each proposal offers Promote (turns it into a real pin) or Discard. Nothing an agent proposes is pinned silently; you decide.
Next
Continue to Connect a CLI over MCP.
Where to go next
- Neural Map — a pinned note becomes a node there too, as long as the session's "Knowledge fabric enabled" setting is on (it is, by default).
- Settings reference — every pinning-related control in a session's own contextual settings.