A deterministic, local-first memory engine for AI assistants that provides versioned context management via MCP, enabling token-budgeted context patches with provenance and explainability.
