Zikkaron
Biologically-inspired persistent memory engine with 26 cognitive subsystems including Hopfield networks, predictive c...
What it does
Biologically-inspired persistent memory engine with 26 cognitive subsystems including Hopfield networks, predictive coding, and hippocampal replay.
Zikkaron is a persistent memory engine for Claude Code built on computational neuroscience principles. It provides 23 tools across 26 subsystems organized into five layers covering core storage, memory dynamics, knowledge representation, meta-cognition, and background consolidation.
The system uses Hopfield energy scoring, spreading activation, and cross-encoder reranking for retrieval without requiring external API calls. It includes hippocampal replay for surviving context compaction, predictive coding write gates to prevent redundant storage, and a cognitive map using successor representations for concept-space navigation. All data persists locally in a single SQLite file with WAL mode, FTS5 search, and sqlite-vec for approximate nearest neighbor lookups.
Capabilities
Server
Quality
deterministic score 0.66 from registry signals: · indexed on pulsemcp · has source repo · 55 github stars · registry-generated description present