
Neural Memory
This is a graph-based memory system that stores knowledge as interconnected neurons and retrieves through spreading activation rather than vector similarity. It exposes 63 MCP tools, though you'll mostly use three: nmem_remember to store memories with auto-detected types and relationships, nmem_recall to traverse the graph through explicit edges like CAUSED_BY or CONTRADICTS, and nmem_health for maintenance insights. Memories decay and consolidate over time, mimicking biological retention. The free tier runs entirely offline with SQLite FTS5, while Pro adds semantic HNSW indexing for scale past 10K neurons. It includes CloudFlare-based sync you deploy yourself, document ingestion from PDF and DOCX, and migration adapters from ChromaDB and Mem0. Reach for this when you need an agent to reason through chains of causality across sessions, not just retrieve similar text.
- GitHub stars
- 202
- tools
- 3
- transport
- STDIO
- last updated
- Apr 2, 2026
Tools
3 toolsPublic tool metadata for what this MCP can expose to an agent.
nmem_rememberStore a memory — auto-detects type, tags, and connections
Store a memory — auto-detects type, tags, and connections
No parameter schema in public metadata yet.
nmem_recallRecall through spreading activation — related memories surface naturally
Recall through spreading activation — related memories surface naturally
No parameter schema in public metadata yet.
nmem_healthBrain health score (A–F) with actionable fix suggestions
Brain health score (A–F) with actionable fix suggestions
No parameter schema in public metadata yet.
Sponsors
Readme
NeuralMemory
Your AI agent forgets everything between sessions. Neural Memory gives it a brain.
Details
- Repository
- github.com/nhadaututtheky/neural-memory
- Category
- AI & LLM Tools
- Package
neural-memory-mcp- Transport
- STDIO
- Registry
- MCP registry, active