May 2026 Summaries

Neo4j has introduced the Neo4j Virtual Graph, a new feature that allows enterprises to leverage graph intelligence without moving or duplicating data from existing data warehouses, lakehouses, and operational databases. This zero-copy architecture enables users to run Cypher queries and graph algorithms directly on data stored in platforms like Snowflake and Databricks, maintaining data governance while unlocking the power of Neo4j's AI-powered Graph Tools. Virtual Graph automatically generates a graph data model from existing tables, enabling graph reasoning necessary for complex queries that traditional flat table queries cannot handle, such as tracing customer relationships or supplier disruptions. It is designed for analytical workloads that can tolerate some latency, in contrast to native Neo4j graphs, which are optimized for low-latency, high-throughput operations. Neo4j Virtual Graph is available in private preview and aims to offer a seamless way for enterprises to test the value of graph patterns on their existing data without extensive setup or migration.

This edition of "This Week in Neo4j" explores the latest advancements in graph databases, focusing on agent memory, graph architecture, and the benefits of Neo4j's semantic layer in improving Text-to-SQL agent efficiency. It highlights the neo4j-labs agent-memory three-tier model, which addresses issues in siloed data systems under concurrent pressure, and presents benchmarks showing significant reductions in token usage and improved accuracy when using Neo4j's semantic layer. The newsletter also details a full production stack for GCP agents utilizing GraphRAG and graph memory, and invites contributions to the upcoming NODES 2026 event. Additionally, it promotes the Aura Agent Hackathon and various learning opportunities through GraphAcademy, emphasizing the value of connected data models in solving complex problems in sectors like military, healthcare, and emergency systems.

A recent IDC study highlights the substantial benefits of the Neo4j Graph Intelligence Platform, reporting a 230% return on investment over three years, with an average annual benefit of $4 million and a payback period of just under eight months. The platform's ability to model and analyze relationships enables enterprises across various industries to tackle complex data challenges, including reducing project timelines, decreasing hallucination rates in large language models, and improving real-time data access for manufacturing processes. The study emphasizes that disconnected data is a common issue, and graph intelligence provides the necessary context for effective analytics and autonomous AI systems. Neo4j's flexible graph model supports a range of applications, from supply chain management to drug discovery, by structuring data around entity relationships, thereby enhancing AI outcomes and reducing hallucination rates. As data becomes increasingly interconnected, the platform is positioned as a critical component in modern data architectures, supporting the development of accurate, explainable, and governed AI systems.

The Neo4j Kubernetes Operator provides a declarative approach for managing Neo4j Enterprise deployments on Kubernetes, leveraging Custom Resource Definitions (CRDs) to handle clusters, databases, users, roles, plugins, and backups with full drift reconciliation against the live Neo4j state. This operator, currently in alpha and maintained personally, facilitates complex Neo4j setups such as TLS-terminated clusters, multi-database configurations with distinct topologies, and plugin management without custom Docker images. It also supports role-based access control with privilege enforcement and user management through Kubernetes Secrets for password handling. Observability is enhanced by surfacing diagnostics through Kubernetes status, allowing operational insights without direct shell access to Neo4j pods. Despite its capabilities, it is not officially supported by Neo4j, Inc., and APIs may change, but the core functionality remains stable for prospective users to explore in local or test environments.

In the blog post "SumoDB in Neo4j: Chaining Multiple Graph Algorithms in Snowflake — Part 3," Benjamin Squire explores how combining Neo4j Graph Analytics with Snowflake SQL can reveal deeper insights into sumo wrestling matches than either tool can achieve alone. By using a dataset of Makuuchi bouts from 2021 to 2025, the analysis employs various graph algorithms like PageRank, Betweenness Centrality, and a novel "Chaos Score" to evaluate wrestlers' performance beyond mere win counts. PageRank is used to assess the prestige of victories based on opponents' strength, revealing wrestlers who consistently beat high-quality competitors. Betweenness Centrality identifies key wrestlers who link different levels of the competitive hierarchy, highlighting their structural importance. Additionally, the study examines non-transitive rivalries, akin to rock-paper-scissors cycles, to illustrate the complexity of dominance in sumo. The combination of these methods provides a comprehensive view of the competitive landscape, emphasizing that true dominance in sumo involves both quality and structural influence, which traditional metrics cannot capture.

Neo4j Graph Intelligence for Microsoft Fabric, since its general availability in October 2025, has introduced several enhancements to streamline graph analytics workflows within Microsoft Fabric, enabling the integration of graph results into OneLake tables to be utilized by Power BI and Microsoft Copilot. This integration allows users to run algorithms like PageRank or Louvain, with results available for natural language queries via Copilot and visualization in Power BI, thereby enhancing business decision-making capabilities. The platform now supports creating graphs on the Business Critical tier and connecting existing Aura databases to Fabric, with new features such as email notifications, a Preview mode for graph creation, and a default Query view that facilitates the exploration and transformation of data. Additionally, the Export feature offers modes for updating original tables or creating new ones, catering to different data management preferences, while the improved sample dataset flow allows trial users to easily create and load data into Lakehouses for graph creation and analysis.

Neo4j is spotlighting its role in empowering AI startups through various initiatives, including a dedicated team of field engineers and a startup program designed to foster the next generation of AI unicorns. As a part of the Google ecosystem, Neo4j offers unique advantages for startups, particularly in the realm of knowledge graphs and GenAI technology. The Neo4j Startup Program aims to provide essential resources and support to startups, helping them leverage Neo4j's capabilities to build robust, scalable solutions. These efforts underscore Neo4j's commitment to supporting innovation and growth within the AI startup community.

APRA recently issued a stark warning to the financial sector about the rapid adoption of AI and inadequate governance practices, a cautionary tale that government agencies should heed due to the higher stakes involved in public sector mishaps. The letter emphasizes the need for treating AI governance as an engineering challenge rather than a bureaucratic task, advocating for "Governance as Code" to ensure automated checks during deployment and avoid the proliferation of "Shadow AI." Furthermore, it highlights the importance of using graph databases for managing AI supply chains, promoting continuous observability over static audits to detect model drift, and enhancing leadership literacy with clear dashboards that translate technical metrics into operational risks. The article suggests that automated guardrails can actually enhance agility by allowing developers to innovate within a secure framework, thus giving leaders the confidence to approve AI projects without fear of legal or ethical breaches.

Neo4j has introduced a dedicated team of field engineers to provide more technical engagement and support to AI startups, responding to feedback from over 700 startups participating in their Startup Program. While the program has already distributed over $1.5 million in credits, startups have expressed a desire for more hands-on guidance, particularly in complex areas such as GraphRAG architecture, schema design, and scaling production systems. Neo4j's new initiative aims to address these needs by offering direct support from engineers experienced in building and optimizing graph systems. The company encourages startups to utilize resources like the Neo4j Aura free tier, GraphAcademy, and the Developer Center to quickly prototype and scale their systems, while also inviting feedback to refine their support offerings further.

Integrating Neo4j knowledge graphs with Salesforce Agentforce enhances the capabilities of Salesforce agents by providing them with a comprehensive, connected enterprise context, making them more useful, explainable, and action-oriented. This integration allows agents to access detailed relationship information, such as subsidiaries, suppliers, competitors, and recent news, which would not be available through traditional CRM summaries alone. By combining Salesforce's workflow layer with Neo4j's graph data, agents can offer strategic briefings that are grounded in explicit relationships rather than generic responses. The integration uses Neo4j to perform graph lookups, which are then enriched by Salesforce's CRM records, and presented through prompt templates that shape the final response. This approach improves operational efficiency, explainability, and the ability to manage and govern the integration process, offering multiple methods for implementation, such as Apex Actions, External Service Actions, and future support for Model Context Protocol (MCP). The result is a more robust and insightful agent architecture that seamlessly fits within the existing Salesforce environment, ensuring that users receive precise, context-rich answers integrated directly into their workflows.

This week's edition of "This Week in Neo4j" highlights the ongoing developments in the world of graph databases, including the upcoming NODES 2026 conference with an open Call for Papers until June 15, and the Aura Agents Hackathon with $50,000 in credits up for grabs. The newsletter also introduces Neo4j Agent Skills, a new resource on GitHub providing updated skills for coding agents in Cypher 25, and offers insights into authentication using Okta-issued OAuth 2.0 tokens for secure application connections. Additionally, it features a retrospective from Cypher inventor Andres Taylor on the language's evolution from a side project to an ISO standard, and encourages community engagement through the Neo4j User Research panel and various events, including webinars and meetups in several cities. The edition also spotlights Ashita Prasad's session on building AI-enhanced mobile apps with Neo4j, with all NODES AI recordings now available for viewing.

Create Context Graph is a Neo4j Labs project that introduces a CLI tool designed to create full-stack context graph applications for AI agents, enhancing their memory by structuring it as a connected graph. This tool allows developers to quickly scaffold applications with integrated knowledge graphs, decision traces, and interactive graph visualizations, using the POLE+O model to categorize entities such as people, organizations, locations, events, and objects. It bridges the gap between flat chat logs and the complex, multi-hop questions AI agents need to answer by capturing the relationships and reasoning behind decisions. The tool supports a variety of domains, such as healthcare and software engineering, and offers connectors for data sources like Linear and Claude Code. By storing memory as a graph, it enables deeper understanding and reasoning capabilities in AI agents, providing real-time visualizations of decision paths. This open-source project is built on top of neo4j-agent-memory and is available for customization and community contributions.

Agent tools are essential functions and services that enable AI agents to interact with external systems and perform tasks beyond mere text generation. These tools transform an AI model from just generating language to executing actions such as querying databases, sending emails, and running workflows. The utility of AI agents largely depends on the diversity and specificity of the tools they use, which are categorized into web search, retrieval, computation, file manipulation, computer-use, and business productivity tools. The distinction between agent tools and skills is crucial; tools perform discrete actions while skills guide the reasoning process for problem-solving. The Model Context Protocol (MCP) standardizes tool integration, allowing AI agents to dynamically discover and use tools across different services without custom integration. This standardization, coupled with the ReAct pattern—a cycle of reasoning and action—enhances the reliability and functionality of AI agents. Effective tool selection and execution are paramount, necessitating precise tool definitions and robust guardrails to ensure safe and accurate agent operations. Retrieval tools are particularly significant as they provide the contextual data necessary for informed decision-making, highlighting the importance of connected knowledge systems like knowledge graphs for enhanced agent performance.

Laurent Tande discusses how using a Neo4j semantic layer can enhance the efficiency and accuracy of Text-to-SQL agents by replacing static YAML schema files with a dynamic knowledge graph approach. This shift reduces token usage by 20-30% on average and significantly increases accuracy on complex multi-table queries by approximately 10 percentage points. The Neo4j semantic layer allows agents to intelligently navigate data architecture by fetching only relevant portions of the graph, improving performance and reducing contextual noise. This method leverages database structure, constraints, and business glossaries, and incorporates user behavior from transaction logs to provide a precise context for generating SQL queries. The result is lower token usage and higher query accuracy, especially for complex questions, as the agent receives a real-time subgraph tailored to the specific query rather than a static, comprehensive schema.

The blog post explores how integrating ServiceNow with Neo4j, a graph database, enhances the value of enterprise data by creating a connected knowledge graph that supports generative AI (GenAI) in deriving meaningful insights. By connecting ServiceNow data, such as incidents, services, and operational history, to a broader enterprise context through Neo4j, organizations can enable AI to understand and reason across various business and technical elements, providing richer, more actionable intelligence. This integration addresses the challenge of fragmented data by offering a comprehensive view that aligns operational workflow data with broader enterprise systems like CRM and security tools, thereby improving the precision of AI-driven responses. Kafka is highlighted for its role in facilitating scalable, resilient data integration, while Neo4j Change Data Capture (CDC) allows for real-time updates to be shared across systems. Ultimately, this approach enhances decision-making for stakeholders by transforming operational data into connected intelligence, supporting a shift from mere workflow automation to a more integrated and intelligent enterprise system.

NODES 2026, the eighth edition of the prominent developer conference focused on graph-powered applications, knowledge graphs, and AI, is scheduled for November 12, 2026, and promises to be the most ambitious yet, following the success of NODES 2025, which attracted over 13,000 participants. This 24-hour global event offers developers, AI engineers, and data professionals opportunities to engage in technical sessions on topics such as autonomous AI agents, GraphRAG, and graph memory, with live talks and real-time Q&A sessions across all time zones. The conference also features "Road to NODES" workshops to prepare attendees with in-depth training on emerging trends. The Call for Papers, open until June 15, 2026, invites submissions of educational presentations related to Neo4j technologies, focusing on AI engineering, modern applications, and data intelligence. Participants can choose between 30-minute technical talks or two-hour hands-on workshops, with a focus on building solutions through code, AI agents, data modeling, and seamless integrations.

Part 2 of "Building Graph-Based Agentic Systems" delves into the real-world implementation of a graph-based agentic AI system called LoanGuard AI, focusing on its compliance and investigation capabilities. The system is designed to enhance explainability by using a structured approach where an orchestrator routes questions to specialist agents, ensuring that the decision-making process is transparent and traceable. The article highlights the importance of maintaining clear boundaries between planning, execution, and data access to avoid failures common in AI systems, such as unclear system boundaries and prompt injection risks. Key challenges addressed include ensuring deterministic evaluations, implementing a strict workflow sequence, and managing context window sizes to prevent degraded reasoning. The system's architecture allows for comprehensive audit trails by storing reasoning as traversable subgraphs rather than mere logs, significantly aiding in compliance and regulatory investigations. Additionally, the article outlines future directions, such as integrating temporal regulation awareness and multi-jurisdictional capabilities, emphasizing that system design and traceability are crucial for moving AI systems from experimental phases to production-ready solutions in regulated environments.

This blog post by Benjamin Squire explores the integration of Neo4j's Graph Analytics with Snowflake, specifically detailing the use of graph databases to analyze data from the world of Grand Sumo. It builds on a prior model that visualized sumo bouts using Neo4j and now describes the installation and utilization of Neo4j Graph Analytics for Snowflake. The focus is on expanding the dataset to include the top 42 rikishi from the Makuuchi division in the Haru Basho 2026, resulting in a graph of approximately 20,000 nodes and 86,000 relationships. The post outlines the advantages of using Neo4j Graph Analytics in Snowflake, such as cloud scalability and on-demand pricing, and elaborates on the application of the Louvain Community Detection algorithm to identify community structures within the sumo data. It demonstrates how the algorithm can infer the ranks of rikishi by analyzing their bout pairings, and concludes with a preview of future analyses combining multiple graph algorithms to gain deeper insights into rikishi performance and techniques.

Neo4j AuraDB is a fully managed, cloud-native graph database designed to simplify infrastructure management while enhancing performance and scalability for developers and architects. It offers two main service tiers: AuraDB Professional, suitable for production-ready apps and startups, and AuraDB Business Critical, aimed at mission-critical applications with high security demands. The platform eliminates the need for complex joins through its use of the Cypher query language and graph algorithms, allowing for faster data insights. Neo4j AuraDB reduces operational burdens by handling provisioning, patching, and upgrades automatically, and features a simple, consumption-based pricing model. Users can begin with a free trial or the AuraDB Free tier, which provides a fully functional graph database with no time limits. For startups, the Neo4j Aura Startup Program offers up to $10,000 in credits and additional support. The service is deployable via various cloud marketplaces, enabling seamless integration and scalability as applications grow.

Neo4j Agent Skills introduce a novel approach to managing updates and enhancements in Cypher and Neo4j patterns by providing specialized skills that agents can use without requiring retraining of models. These skills are organized in a community-driven, MIT-licensed repository on GitHub, offering a range of tools for various programming languages and applications, including data importation, AI, search, and graph data science. Each skill comprises a folder with a SKILL.md file that is progressively disclosed to the agent based on task relevance, allowing for opinionated guidance without overwhelming context. The repository is designed to be extensible, encouraging contributions and updates through automated processes that align with the latest Neo4j developments. This initiative enables developers to enhance their agents' capabilities by simply installing the relevant skills, ensuring they can effectively handle tasks such as writing and optimizing Cypher queries or integrating with Python drivers, thereby streamlining the development process in Neo4j environments.

Agentic RAG (retrieval-augmented generation) is an advanced AI framework where an agent dynamically controls the retrieval process, unlike standard RAG's fixed retrieve-then-generate approach. This method is particularly effective for complex queries needing multi-step reasoning, cross-source synthesis, or evidence validation before response generation. Agentic RAG iteratively retrieves and evaluates data until it achieves sufficient context or reaches a stopping point, resulting in more reliable and auditable answers. The system is characterized by its ability to select tools and adjust retrieval strategies on-the-fly, which can effectively address industries such as enterprise knowledge, finance, legal, healthcare, and customer support. However, this adaptability comes with tradeoffs, including increased latency, token costs, and complexity in system evaluation and maintenance. Implementing agentic RAG involves starting with a standard RAG baseline, identifying specific failure modes, introducing agentic patterns to address these issues, and optimizing retrieval contexts using tools like GraphRAG for improved relevance and traceability.