Skip to content

White Paper

Jump to bottom
Bellok edited this page Apr 4, 2026 · 4 revisions

White Paper: Neurodivergent Memory MCP Server

Back to Home

Date: 2026-04-04 | Version: 0.3.0 | Status: Research Preview (pre-1.0)

Author: Jerimiah Meyer (Jerry)

Founder Note

This project is my brain-child. The goal is straightforward: build a memory system that supports neurodivergent, non-linear cognition while still meeting engineering standards for agent tooling, release discipline, and supply-chain trust.

Executive Summary

Neurodivergent Memory MCP Server is a research-preview memory platform for agent workflows. It combines district-based memory organization, graph relationships, canonical metadata tags, and ranked retrieval.

Release 0.3.0 is the current contextual-intelligence milestone, building on the trust-and-telemetry foundation introduced in 0.2.0. This document records design intent, measured release outcomes, operational lessons, and the production-readiness gap to 1.0.0.

Problem Statement

Most generic memory stores are not designed for:

  • non-linear recall patterns,
  • explicit context partitioning,
  • associative graph traversal,
  • and repeatable operational release delivery.

This project addresses those gaps with a district-aware graph model, canonical tag schema, BM25 retrieval, and hardened release automation.

Architecture Summary

For full architectural detail, see Architecture.

District Model

Memories are stored in five cognitive districts:

  • logical_analysis
  • emotional_processing
  • practical_execution
  • vigilant_monitoring
  • creative_synthesis

The district model provides intentional context boundaries while still allowing cross-district graph connections.

Canonical Tag Schema

Memories use structured tags in five namespaces:

  • topic:X
  • scope:X
  • kind:X
  • layer:X
  • persistence:X

This schema improves consistency, discoverability, and retrieval quality across human and agent-authored entries.

Retrieval and Graph Operations

The system provides:

  • BM25-ranked lexical retrieval,
  • relationship-aware graph traversal,
  • and state/health inspection through memory statistics.

Release 0.3.0 Outcomes

For detailed release notes, see Release-Notes.

Delivery Channels Completed

The current release line publishes across:

  • npm package registry,
  • GitHub release assets,
  • Docker Hub container registry,
  • and GitHub Container Registry (GHCR).

Supply-Chain and Integrity Controls

Release automation includes:

  • provenance-aware npm publishing,
  • package and container attestations,
  • checksum generation,
  • and multi-architecture container builds.

Trust & Telemetry Milestone Additions

Release 0.3.0 added higher-level cognitive and operator workflow features:

  • structured emotional-to-logical distillation via distill_memory,
  • active loop guardrails with no-net-new-info and distillation suggestions,
  • agent_id attribution and per-agent visibility in memory_stats,
  • goal-aware and recency-aware retrieval controls,
  • LUCA-addressed custom districts,
  • richer import/storage diagnostics and snapshot migration controls,
  • and packaged agent-kit templates with an explicit installer workflow.

Release 0.2.0 remains the underlying trust-and-telemetry foundation that made these features safe to add:

  • write-ahead journal persistence with startup replay and compaction,
  • async write serialization with bounded queue backpressure,
  • stable NM_E error taxonomy for MCP failures,
  • structured Pino logging,
  • observe-only loop telemetry,
  • and published benchmark baselines for 1k, 5k, and 10k memory datasets.

Operational Lessons Incorporated

The release process surfaced and resolved key issues:

  • npm immutability behavior on reruns,
  • 2FA/EOTP behavior in automation contexts,
  • main/release branch synchronization gaps,
  • and package visibility mismatch when GHCR was not included.

Workflow hardening now includes:

  • automatic release-branch sync on version tags,
  • rerun-safe npm publish skip when version already exists,
  • and GHCR publication with required permissions for GitHub Packages visibility.

Agentic Value Assessment

Current Value

The platform is strong for:

  • single-agent memory workflows,
  • structured reflection and recall,
  • and early-stage multi-agent experiments.

Current Constraints

The project remains pre-1.0 and is not yet suitable for high-criticality production workloads due to remaining gaps in:

  • cross-process write coordination,
  • stronger durability guarantees beyond the current WAL plus snapshot model,
  • capability and introspection clarity across heterogeneous MCP clients,
  • and orchestration-native control surfaces.

Multi-Agent Roadmap Direction

For the full roadmap, see Roadmap.

Roadmap intent includes support for council-style and Kanban-like agent orchestration, where multiple CLI agents coordinate through shared memory and explicit lifecycle semantics.

Planned capabilities:

  • lifecycle hooks for memory events,
  • conflict-aware write and merge patterns,
  • goal-aware retrieval and ranking,
  • and reference orchestration patterns for parallel agent execution.

Readiness Position

Release 0.3.0 should be interpreted as:

  • ready for research and controlled pilot use,
  • not yet ready for production-scale, multi-tenant deployment.

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

Milestones to 1.0.0

See Roadmap for the full milestone breakdown.

Milestone Focus
0.2.x Structured logs, error taxonomy, write serialization, crash recovery, loop telemetry
0.3.x–0.4.x Distillation, contextual retrieval, orchestration-aware retrieval, multi-agent reference implementations
1.0.0 Durability guarantees, stable API contract, validated operational runbooks

Publication Strategy

This wiki serves as the canonical, versioned technical record for the project. The full white paper is maintained here and linked from the repository README.

Primary supporting materials in-repo:

  • README
  • CHANGELOG
  • EXPERIMENT_REPORT
  • SMOKE_TEST_REPORT
  • SECURITY
  • .github/workflows/release.yml

Disclaimer

This document describes a research-preview release and does not constitute a production SLA, compliance certification, or formal security guarantee.

See also: Architecture · Release-Notes · Roadmap

Home

Setup

Reference

Reports

v0.3.0 · Research Preview

npm · Docker Hub · GHCR

Clone this wiki locally