EngramMemory and Behavioral Intelligence for AI Agents

Cross-session memory with metabolic lifecycle, real-time behavior monitoring, and predictive knowledge supply

Metabolic Knowledge

Knowledge that is used survives. Knowledge that isn't, dies. Two-state lifecycle (recent → fixed), density-based dynamic metabolism. No accumulation bloat.

Receptor & Emotion Engine

Three-layer neuron model observing agent behavior. 5-axis emotion vector, adaptive thresholds, state classification — zero LLM inference.

Shadow Index

Pre-neuron monitor detecting stale file access that the neuron system structurally cannot see. Multi-index heatmap with staleness detection.

Future Probe

Predictive knowledge supply — searches near the agent's behavioral position before it knows it needs help.

Persona & Prior Block

Session experience distilled into reusable calibration (Persona) and behavioral arc (Prior Block). No cold start. Cross-session continuity.

Data Cost Protocol

Compact positional arrays for AI-to-AI communication. Schema registry with hash-based handshake. No natural language overhead.