Nexo transforms stateless AI assistants like Claude Code into cognitive partners by implementing the Atkinson-Shiffrin memory model from cognitive psychology. It provides AI agents with human-like memory, featuring Ebbinghaus forgetting curves, rehearsal-based reinforcement, and memory consolidation, ensuring they remember past interactions, learn from mistakes, and adapt their behavior. This allows for semantic search of memories, metacognitive self-correction, handling cognitive dissonance, and building trust scores, all processed locally without cloud dependencies, providing a deep, evolving cognitive architecture for AI agents.
主な機能
01Cognitive dissonance detection to verbalize conflicts in instructions
02Semantic search for memories based on meaning using vector embeddings
03Atkinson-Shiffrin memory model with Ebbinghaus forgetting curves and consolidation
04Metacognitive guard system to prevent repeating past mistakes
051 GitHub stars
06Dynamic trust scoring and sentiment detection to adapt agent behavior