KG Memory

About

KG Memory acts as a production-grade Model Context Protocol (MCP) Server, addressing the critical 'context amnesia' issue in AI coding assistants like Cursor, Windsurf, VS Code, and Antigravity. It maintains a dynamic, persistent Knowledge Graph of your project, allowing AI agents to remember goals, strategies, constraints, user preferences, and code relationships across sessions. This system empowers agents to learn from past outcomes, inject relevant context automatically, and perform semantic searches, evolving into a more intelligent and effective coding partner.

Key Features

• 7 GitHub stars
• Tracks goals, constraints, strategies, user preferences, and code relationships
• Active Context Injection (kg_autopilot) for task-relevant information retrieval
• Semantic & Graph Search to find connected context beyond simple string matching
• Strategic Learning that infers patterns and tracks success/failure of strategies
• Utilizes Neo4j as a graph database and Google Gemini 2.5 for LLM processing

Use Cases

• Equipping AI coding assistants with long-term memory to prevent context amnesia
• Automating the retrieval of project goals, past decisions, and coding preferences for new tasks
• Tracking code changes and linking them to active goals within a knowledge graph