Home

Neurodivergent Memory MCP Server

Status: Research Preview (pre-1.0) · Latest Release: 0.3.0 · Current Development Focus: v0.4.0 planning and post-release hardening

A TypeScript-based MCP server implementing a memory system inspired by neurodivergent cognitive styles. It organizes thoughts into five districts (knowledge domains), ranks search results using BM25 semantic ranking, and stores memories as a persistent knowledge graph with bidirectional connections.

Quick Links

Page	Description
Architecture	District model, tag schema, archetypes, tools, retrieval, and graph operations
Getting-Started	Installation, Claude Desktop config, Docker, and MCP Inspector
Release-Notes	Published release history including the 0.3.0 contextual-intelligence release
Roadmap	Milestones from 0.3.x through 1.0.0
Experiment-Report	Historical baseline experiment plus current validation context
White-Paper	Current research-preview white paper for v0.3.0

Model Flow

graph TD
  A[MCP Client] -->|tool call| B[MCP Server stdio]
  B --> C{Tool Router}
  C --> D[store_memory]
  C --> E[retrieve_memory]
  C --> F[search_memories]
  C --> G[other tools...]
  D --> H[Graph State + BM25 Index]
  E --> H
  F --> H
  H --> I[Persistence: memories.json + WAL]
  B -->|tool response| A

Loading

• Memory operations update both graph state and BM25 index.
• Persistence writes to the local snapshot file for restart continuity (WAL journal added in v0.2.0).
• All MCP responses return through stdio transport.

Features at a Glance

Five Memory Districts

District	Purpose
logical_analysis	Structured thinking, problem solving, analytical processes
emotional_processing	Feelings, emotional responses, affective states
practical_execution	Action-oriented thoughts, tasks, implementation
vigilant_monitoring	Awareness, safety concerns, protective thinking
creative_synthesis	Novel connections, creative insights, innovative thinking

17 Memory Tools

store_memory · retrieve_memory · update_memory · delete_memory · connect_memories · search_memories · traverse_from · related_to · list_memories · memory_stats · storage_diagnostics · import_memories · distill_memory · prepare_memory_city_context · prepare_synthesis_context · prepare_packetized_synthesis_context · register_district

Prompts

• explore_memory_city — Guided exploration of districts and memory organization
• synthesize_memories — Create new insights by connecting existing memories
• synthesize_memory_packets — Packetized synthesis prompt for attachment-constrained clients

Semantic Search

Uses Okapi BM25 ranking (k1=1.5, b=0.75) with optional goal context, recency bias, and epistemic-status filtering — no embeddings or cloud calls required.

Installation

npm

npm install neurodivergent-memory

Docker Hub

docker run --rm -i twgbellok/neurodivergent-memory:0.3.0

Use explicit image tags in documentation and client configs. The published release images intentionally do not maintain a floating latest tag.

GitHub Container Registry

docker pull ghcr.io/jmeyer1980/neurodivergent-memory:0.3.0

For full setup instructions including Claude Desktop configuration, see Getting-Started.

Readiness Position

Release 0.3.0 is:

• Ready for research use, controlled pilots, and single-agent memory workflows.
• Not yet ready for production-scale, multi-tenant, or high-criticality deployments.

This is intentional and consistent with semantic versioning expectations for 0.x software.

Author

Jerimiah Meyer (Jerry) — GitHub

See also: Architecture · Getting-Started · Release-Notes · Roadmap