How does the deduplication logic work?

It uses a similarity engine to identify and merge records that are more than 85% identical, preventing redundant or conflicting context from bloating the AI's context window.

What is Memory Fabric?

Memory Fabric is a skill for Claude Code that acts as an orchestration layer for knowledge graph memory, improving how the AI extracts, stores, and retrieves project-specific context.

Does it require a specific MCP server?

Yes, it requires the standard Anthropic Memory MCP server to function as its underlying storage engine for graph nodes.

Is Memory Fabric user-invocable?

No, it is designed to operate in the background as a system enhancement, improving Claude's internal memory management processes without needing manual commands.

What is 'Cross-Reference Boosting'?

It is a ranking technique that automatically increases the priority of memories that are logically linked to other existing entities in your project's knowledge graph, ensuring high-authority information surfaces first.

Memory Fabric

Name: Memory Fabric
Author: yonatangross

byyonatangross

•

Data Science & ML

Orchestrates complex knowledge graph memory with entity extraction, deduplication, and intelligent cross-referencing for Claude Code.

Memory Fabric is a sophisticated orchestration layer for Claude Code that enhances the standard memory MCP server by implementing a graph-first approach to knowledge retrieval. It automates entity extraction from natural language, parses complex queries, and performs high-precision deduplication to ensure a clean, unified context. By utilizing multi-hop graph traversal and a custom ranking algorithm that balances recency, relevance, and source authority, it allows Claude to maintain deep, persistent project knowledge across long-running development sessions and complex architectural tasks.

Key Features

01Weighted ranking system based on recency, semantic relevance, and source authority

02Intelligent query parsing to identify search intent and entity hints

03Automated entity extraction and relationship mapping from natural language inputs

04High-precision deduplication engine using an 85% similarity threshold to merge redundant data

05Cross-reference boosting that elevates the relevance of interconnected graph nodes

0692 GitHub stars

Use Cases

01Tracing architectural decisions and engineer recommendations across multiple development sessions

02Enhancing RAG capabilities within the Claude Code CLI for more accurate contextual answers

03Building a persistent knowledge base for large-scale codebases with complex architectural dependencies

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add yonatangross/orchestkit memory-fabric

For use in Claude.ai and ChatGPT

Download Skill