Recall is a local-first memory layer for AI agents: Markdown for humans, SQLite for search, MCP for agents. Agents write structured durable facts through CLI/MCP/API. Humans browse, search, inspect, and audit those memories through a read/view-first UI. The source of truth stays as Markdown on your machine; SQLite is a rebuildable local index. Agents write. Humans inspect. Recall indexes. MCP retrieves.