Evidence-backed memory for AI agents: local-first recall, temporal history, tested skills, and gradual migration.
Terminal score: 0–1000 raw, weighted across 4 dimensions. Public score: 0–10 normalized (shown in the 30-day stars chart above).
