Neo4j Live: AgentMemory for .NET: Persistent AI Agent Memory Engine
.NET developers building AI agents have felt this gap: Neo4j Agent Memory shipped for Python and TypeScript, but .NET was left bolting things together. José L. Latorre closes it with AgentMemory for .NET - a from-scratch implementation verified 178/178 against Neo4j's own Agent Memory Test Compatibility Kit. In this livestream, he shows how the same three memory layers - short-term, long-term (POLE+O), and reasoning - now live natively in .NET, wired straight into Microsoft Agent Framework and Semantic Kernel, so your agent remembers by design, not by hoping it calls the right tool at the right time. Guest: Jose Latorre ( https://www.linkedin.com/in/joslat/ ) Blog: https://neo4j.com/blog/developer/agentmemory-for-net-a-native-sibling-to-neo4j-agent-memory/ 0:00 Introduction 2:32 Jose's Inspiration: From Sci-Fi Robots to Agent Memory 8:47 Agent Memory Basics: Context, RAG & Graph RAG Explained 17:53 Surveying the Memory Landscape: Mem0, Graphiti & Others 20:54 Choosing Storage: Files vs. Vector DBs vs. Graphs 24:29 Building Agent Memory for .NET & the POLE Schema 27:56 Live Demo: Inside the Memory Brain 32:56 Voice Interaction, Alice in Wonderland & Forgetting 43:42 Performance Optimization & Benchmarking Results 47:46 Roadmap: Extensibility, Skill Distillation & Auto-Learning 52:03 Behavioral Memory & Alignment Research 54:32 Audience Q&A & Closing #neo4j #graphdatabase #agenticai #knowledgelayer #knowledgegraph #graphrag #AIAgents #dotnet #agentframework