This is a proof-of-concept server that gives LLMs structured memory through knowledge graphs. It exposes MCP tools for storing facts, learning business rules through conversation, and running logical deductions using RDFLib and SPARQL queries. You teach it by chatting naturally: tell it "Bob drives to work" and it can later propose rules like "drivers are adults" and "adults can vote" to infer new facts. Beyond the conversational mode, it includes a dashboard for extracting rules from PDFs and visualizing the knowledge graph. Works with Claude, Gemini, or any MCP client. The author marks it experimental and not production ready, but it's a solid example of neuro-symbolic architecture if you're exploring how to make LLM conversations remember and reason over structured knowledge.