This Week in Neo4j: NODES 2026, Agent Memory, GraphRAG, Knowledge Layer and more

Neo4j published the NODES 2026 agenda for its free 24-hour virtual conference on November 12, featuring 100+ speakers across AI Engineering, Data Intelligence and Modern Applications tracks, alongside a free "Road to NODES" workshop series running throughout October on GraphRAG, agentic systems and memory. Neo4j also highlighted NICD research showing lightweight graph structures can nearly double multi-hop answer rates over vector-only RAG, and a new enterprise knowledge layer series that models roles rather than people as the load-bearing nodes behind governed AI agents. Community member Luis Latorre, a Dev Community Lead and Agentic Architect at Swiss Life AG, built AgentMemory for .NET, a native .NET implementation bringing graph-native short-term, long-term and reasoning memory to Microsoft Agent Frameworks, to be showcased in a September 30 livestream.

This Week in Neo4j: NODES 2026, Agent Memory, GraphRAG, Knowledge Layer and more Staff Community Manager 3 min read Welcome to This Week in Neo4j, your fix for news from the world of graph databases Road to NODES 2026 https://neo4j.registration.goldcast.io/series/94f27929-1670-4f2b-bfdd-ab4e48eb8c22?utm source=database&utm medium=email&utm campaign=dev is here – free two-hour workshops on GraphRAG, agentic systems and memory running all October, warming you up for the 24-hour virtual conference on November 12 with 100+ speakers agenda is also published to you to explore . Also this week: a self-learning multi-agent series digging into what actually flows through the Meta Knowledge Graph, from semantic-layer grounding to gated, contradiction-checked memory; NICD research showing lightweight graph structures can nearly double multi-hop answer rates over vector-only RAG; and a new enterprise knowledge layer series, modeling roles rather than people as the load-bearing nodes behind any governed AI agent. Happy Graphing, Alexander Erdl COMING UP - Livestream : Neo4j Live: AgentMemory for .NET: Persistent AI Agent Memory Engine https://youtube.com/live/6k096 wfHCY on September 30 - Conferences : Find us at Java Forum Nord https://javaforumnord.de/2026/ on September 22, AI Summit Barcelona https://aisummitbarcelona.com/ on September 22-23, AI Engineer Paris https://www.ai.engineer/paris/2026 & Big Data London https://www.bigdataldn.com/ on September 23-24, The Developer Conference https://thedevconf.com/tdc/2026/sao-paulo/ on September 23-25, WeAreDevelopers https://www.wearedevelopers.com/world-congress-north-america/agenda/sessions/building-world-aware-robots-with-agent-memory-and-context-graphs-1220041 on September 24-25 & AI Conference https://aiconference.com/ on September 29-October 1 - Meetup : Meet us in Paris, FR https://luma.com/ai-builders-paris-sept on September 21, San Francisco, US https://luma.com/utwyaywv on September 22, Richmond, US https://www.meetup.com/richmond-java-users-group/events/315682428/ on September 23, Munich, DE https://www.meetup.com/de-de/munich-datageeks/events/316577539 on September 24, Bengaluru, IN https://globalai.community/e/bd1o37ln on September 26 & Madurai, IN https://gdg.community.dev/events/details/google-gdg-cloud-madurai-presents-unlocking-google-gemini-building-enterprise-ready-agents/ on October 3 - All Neo4j Events : Webinars and More https://neo4j.com/events/ FEATURED COMMUNITY MEMBER: Luis Latorre https://www.linkedin.com/in/joslat/ Luis Latorre https://www.linkedin.com/in/joslat/ Jose is a Dev Community Lead & Agentic Architect at Swiss Life AG and Microsoft Agent Framework contributor, bringing hands-on lessons on building measurable, production-ready AI agents. Connect with him on LinkedIn https://www.linkedin.com/in/joslat/ . He recently built AgentMemory for .NET https://neo4j.com/blog/developer/agentmemory-for-net-a-native-sibling-to-neo4j-agent-memory/ , a native .NET implementation inspired by Neo4j Agent Memory, which brings graph-native short-term, long-term and reasoning memory to Microsoft Agent Frameworks – he will also be showcasing this in the live stream end of September NODES 2026: Road to NODES 2026 and Agenda published https://neo4j.registration.goldcast.io/series/94f27929-1670-4f2b-bfdd-ab4e48eb8c22?utm source=database&utm medium=email&utm campaign=dev The NODES 2026 agenda https://neo4j.com/nodes/agenda/ has dropped: a free 24-hour virtual conference on November 12 with 100+ speakers across tracks on AI Engineering, Data Intelligence, and Modern Applications – real-world sessions on multi-agent workflows, GraphRAG, context graphs and knowledge graph construction, running across Asia-Pacific, Europe and Americas time zones. To get ready, explore our free “Road to NODES” workshop series running throughout October: hands-on, two-hour sessions on GraphRAG, agentic systems, reasoning, memory and production-ready graph+GenAI patterns. AGENT MEMORY: The payload: the semantic layer, memory, and the loop that fills it https://neo4j.com/blog/genai/the-payload-the-semantic-layer-memory-and-the-loop-that-fills-it/ Part 2 of the self-learning multi-agent series from Tomaž Bratanič that digs into what actually flows through the Meta Knowledge Graph: a semantic layer Neocarta that gives agents business meaning on top of raw schemas and a three-pass memory system that captures every session, distills tool failures into reusable fix patterns and gates durable “learnings” behind contradiction checks and human review before they can rewrite an agent’s persona. It’s a detailed, opinionated look at how to make agent memory earn its place rather than just accumulate. GRAPHRAG: From Agent Memory to Portable Skills https://neo4j.com/blog/genai/from-agent-memory-to-portable-skills/ Roberto Zicari interviews Dr. Jacek Cała from the National Innovation Centre for Data NICD to dig into NICD’s research https://neo4j.com/whitepapers/nicd-reducing-hallucinations-graphrag/ showing that even a lightweight, simple graph structure can halve hallucination rates and push answer rates on complex multi-hop questions from 29% to nearly 66%, compared to vector-only RAG. Cała explains why reliable links beat brute-force vector similarity for multi-hop reasoning, and is refreshingly candid about where the findings do and don’t generalise to messy enterprise data. KNOWLEDGE LAYER: The knowledge layer for enterprise AI: Operating structure https://neo4j.com/blog/genai/the-enterprise-knowledge-layer-operating-structure/ Chapter one of a series building a full enterprise knowledge layer, starting with the “who and what” – business units, teams, roles, and reporting lines – modelled as a live, queryable graph instead of a stale org-chart slide. Using a fictional bank as the running example, Morgan shows why roles not people should be the load-bearing nodes, and why this layer is the precondition for any governed, accountable AI agent acting inside a business. STARTUPS: Hexology https://www.linkedin.com/posts/hexology culturalheritage-globaldigitalcitizenship-activity-7504122093408776192-k01K This week, Hexology presented at the V&A. Founder and CEO Darrel Butlin spoke at the museum’s one-day conference on Ukrainian cultural heritage and UK institutions, presenting “Interpreting Ukrainian Cultural Heritage Through Geospatial Digital Media.” He set out a citizen-led model trialled with academic partners in the UK and Ukraine: location-based micro-collections that document Ukrainian heritage and push back against cultural erasure. Hexology is a UK startup whose app pins digital media to physical places. Hexology also won an Innovate UK / DCMS Create Growth Award. CONTINUOUS LEARNING - GraphAcademy : Build an ontologically sound agent memory graph system https://graphacademy.neo4j.com/courses/workshop-agent-memory-90 , give any agent the power to record memories and decision traces, use a memory-enabled agent to build a custom agent, and distil it all as a skill. - Learn on Your Schedule : Go deeper into graph intelligence on Neo4j’s On-Demand webinar library https://neo4j.com/events/on-demand/? event level=intermediate - Workshops : Join our virtual classrooms workshops from Fundamentals to GenAI https://neo4j.com/events/? event type=workshops - New Webinar : Turn complex documents into scalable knowledge graphs – Americas https://go.neo4j.com/WBR-EDU-260929-Scalable-knowledge-graphs Registration.html , Europe, Middle East & Africa https://go.neo4j.com/WBR-EDU-260929-Scalable-knowledge-graphs-EMEA Registration.html , Asia Pacific https://go.neo4j.com/WBR-EDU-260929-Scalable-knowledge-graphs-APAC Registration.html POST OF THE WEEK: Alan Smith https://bsky.app/profile/alanbsmith.bsky.social A tangible example: I built a prototype graph database for all our 800+ icons. Then I put an MCP server in front of the database so we could query it deterministically and with natural language. You can see all the tags on our arrow-up icon and all the icons tagged with “direction.” — Alan B Smith @alanbsmith.bsky.social https://bsky.app/profile/did:plc:3dpbdojqyfmc5smpvw3sln4l?ref src=embed 10. September 2026 um 05:52 https://bsky.app/profile/did:plc:3dpbdojqyfmc5smpvw3sln4l/post/3mv55myjg2c2t?ref src=embed Please share it if you like it