BACK_TO_FEEDAICRIER_2
NeuroStack gives AI tools persistent local memory
OPEN_SOURCE ↗
REDDIT · REDDIT// 27d agoOPENSOURCE RELEASE

NeuroStack gives AI tools persistent local memory

NeuroStack is a local-first CLI and MCP server that adds persistent memory, semantic search, and knowledge graphs to any MCP-compatible AI assistant including Claude Code, Cursor, and Windsurf. It installs via npm with zero prerequisites and claims 96% fewer tokens per query compared to naive RAG approaches.

// ANALYSIS

The MCP ecosystem is becoming the connective tissue for AI tooling, and NeuroStack is a smart bet on that stack — local memory with no SaaS dependency is exactly what privacy-conscious developers want.

  • 12 MCP tools give any compatible AI assistant direct access to a personal knowledge vault with semantic search, PageRank-weighted wiki-link graphs, and session harvesting from Claude Code transcripts
  • Neuroscience-inspired "stale detection" surfaces outdated notes using prediction error signals — a clever framing that maps to real ML/memory reconsolidation concepts
  • Three install tiers (Lite ~130MB / Full ~560MB / Community ~575MB) let developers pick between basic full-text search and full GraphRAG topic clustering via the Leiden algorithm
  • Obsidian integration and profession-specific templates (researcher, DevOps, data scientist, etc.) lower friction for developers already using markdown-based PKM workflows
  • Apache 2.0, fully local, no API keys or data leaving the machine — deliberate counter-positioning to SaaS memory tools like Supermemory
// TAGS
neurostackmcpragopen-sourcedevtoolagentself-hosted

DISCOVERED

27d ago

2026-03-16

PUBLISHED

27d ago

2026-03-16

RELEVANCE

7/ 10

AUTHOR

raphasouthall