Sage

SAGE gives Claude institutional memory that survives restarts and validates writes through BFT consensus before committing them to disk. It exposes MCP tools for storing memories, querying by semantic similarity, and retrieving context across conversations. Every write goes through a four-validator quorum even in single-agent mode, and memories carry confidence scores that decay over time. The CEREBRUM dashboard renders your agent's memory as a force-directed neural graph with domain filtering and real-time updates. You'd reach for this when you need AI agents to build genuine long-term knowledge rather than starting from scratch every session. Ships with SQLite storage, optional AES-256-GCM encryption, and REST endpoints alongside MCP stdio.