Neurodivergent Memory
Run a local persistent knowledge-graph MCP with BM25 search tailored to nonlinear thinking—no cloud LLM required for retrieval.
Overview
Neurodivergent Memory is a Build-phase MCP server that provides a local persistent knowledge graph with BM25 search and no cloud LLM for retrieval.
What is this MCP server?
- Persistent knowledge graph MCP aimed at neurodivergent workflow patterns
- BM25 search for lexical recall without sending memory to a cloud LLM
- npm package neurodivergent-memory 0.1.8 with stdio transport
- Keeps context structures friendly to non-linear jumps and revisits
- GitHub-hosted open source under jmeyer1980/neurodivergent-memory
- Registry version 0.1.8 on npm identifier neurodivergent-memory
- stdio transport single-package manifest
- BM25 search explicitly documented; no cloud LLM for retrieval
Community signal: 12 GitHub stars.
What problem does it solve?
Linear note tools and cloud-only memory break focus for neurodivergent builders and leak private context to remote models.
Who is it for?
Privacy-conscious solo builders who want agent memory that matches jumpy, revisit-heavy thinking styles.
Skip if: Teams that need hosted semantic search with managed embeddings and are fine shipping data to cloud LLM APIs.
What do I get? / Deliverables
After install, agents store and BM25-search a local knowledge graph so nonlinear ideas resurface without cloud LLM round-trips.
- Local persistent knowledge graph accessible via MCP tools
- BM25 search results for agent prompts without cloud LLM calls
- Structured memory that supports nonlinear revisit patterns
Recommended MCP Servers
Journey fit
Local memory MCPs belong in Build when you equip agents with durable, private recall before you scale distribution or ops monitoring. agent-tooling is the shelf for graph-based memory servers agents call over stdio, especially privacy-first BM25 stacks.
How it compares
Local BM25 knowledge-graph MCP, not a ChatGPT memory feature or a hosted vector database product.
Common Questions / FAQ
Who is Neurodivergent Memory for?
Indie builders and neurodivergent developers using MCP agents who need persistent, locally searchable memory without cloud LLM retrieval.
When should I use Neurodivergent Memory?
Use it during agent-heavy build work when sessions fragment across days and you want BM25-backed graph recall that respects nonlinear workflows.
How do I add Neurodivergent Memory to my agent?
Install the neurodivergent-memory npm package, add it as a stdio MCP server in your agent configuration, and expose the graph tools to your coding sessions.