memsprout

Using memsprout

memsprout is this team's persistent memory, connected as an MCP server. What you write there is read later by the user, by teammates in shared spaces, and by other AI agents across many tools. They trust it — treat it as shared memory, not a temporary scratchpad.

Search first

• Call search_memories before assuming something isn't recorded, and before storing anything, so you don't create duplicates. It searches everything you can read by default.
• Use list_memories for recency browsing instead of meaning-based lookup.
• At the start of a substantive task, search memsprout for relevant prior context.

Capturing

Store memories liberally with store_memory, and default to giving each one a concise, descriptive title — especially anything durable: a decision, a procedure, a fact, a conclusion worth finding by name later, and anything going into a shared space. Leave the title off only for genuinely raw, ephemeral captures — a half-formed idea or fragment not meant to stand alone. You can promote a raw capture later by adding a title via update_memory.

At the end of a task, capture the durable conclusion — titled, so it's findable later — without waiting to be asked.

Memories are versioned: update_memory archives the prior state with an optional change_note, so revise freely — history is never lost.

Spaces and topics

Spaces are shared containers with member roles. Pass space_id to store_memory for shared work; omit it for the user's personal vault. Only pass a shared space_id when the user has named that Space in the current conversation — never infer a shared destination. No named Space means omit space_id, which defaults to personal.

Every space memory is assigned to a topic. When capturing into a space, check list_topics and pass the best-fitting existing topic; if nothing fits, create_topic first (with a description). Omit the topic only as a last resort — the server auto-classifies as a backup.

Files