August 2025 Summaries

In the latest edition of "This Week in Neo4j," readers are introduced to a variety of events and innovations in the realm of graph databases, such as the Road to NODES workshops designed to enhance graph skills and the creation of GraphRAG pipelines using Neo4j and other tools for offline or privacy-sensitive applications. The edition highlights the Neo4j Model Context Protocol (MCP), which helps transform everyday tasks into structured graphs, and the BandtoBand project that maps the interconnectedness of rock bands through shared members, creating a vast "family tree" of music relationships. Additionally, the article features Dattaraj Rao, an AI thought leader and speaker at NODES 2025, who, alongside Sadashiv Borkar, will discuss the integration of agentic AI with enterprise knowledge graphs. The newsletter also provides information on upcoming conferences, meetups, and the GraphSummit Series, inviting readers to join the Neo4j User Research panel to influence future product development.

Entiros Integrations has developed a graph-first approach to integration using Neo4j and AI, transforming integration challenges into opportunities through collaboration between humans and AI. Their system is built on three pillars: the nonlinear mapping tool Starlify, which uses Neo4j's graph technology to provide a clear overview of integration landscapes; the Certified Integrator framework, offering a standardized methodology for integrations; and Ingo AI, an intelligent agent designed for agentic reasoning within this structured landscape. This approach facilitates faster, more accurate, and sustainable integrations by enabling AI to act as a reasoning team member, helping to navigate and shape integration landscapes. The system aims to enhance human efficiency rather than replace it, ensuring robust and adaptable digital flows.

In the realm of financial services, Know-Your-Customer (KYC) and Anti-Money Laundering (AML) are critical elements for combating illicit activities, and this blog post details the implementation of a KYC agent using OpenAI’s Agents SDK, MCP, Neo4j, and Ollama. The KYC agent is designed to navigate complex networks of customer relationships using a knowledge graph, allowing it to identify potential fraud patterns and suspicious activities through interconnected data points like customers, accounts, transactions, and devices. By leveraging graph-powered data retrieval tools, the agent can perform tasks such as detecting circular transaction patterns, retrieving customer details, and generating Cypher queries from natural language inputs. This approach offers a robust framework for dynamic querying and persistent memory storage, enabling the agent to evolve its knowledge base and assist in fraud investigations effectively. The blog also highlights the potential application of these patterns and tools beyond KYC, in domains like supply chain analysis and drug discovery, emphasizing the growing relevance of graph-aware AI agents.

Retrieval-augmented generation (RAG) is increasingly limited by traditional vector-based approaches when handling complex, interconnected information, prompting the need for structured data to enhance retrieval and reasoning capabilities. By transforming unstructured documents into structured knowledge representations, tools like LlamaCloud and Neo4j facilitate sophisticated graph traversals, relationship queries, and contextual reasoning, which are particularly valuable in the legal domain. Legal documents, with their intricate webs of references and hierarchical nature, benefit from the precision of structured knowledge graphs to improve retrieval accuracy. The process involves using LlamaParse to extract text from documents, classifying contract types with an LLM, extracting relevant attributes with LlamaExtract, and storing the information in a Neo4j knowledge graph. This approach allows for intelligent retrieval systems that understand entity relationships, enabling complex queries beyond simple text fragment searches.

The tutorial on building a retrieval-augmented generation (RAG) system using a knowledge graph provides a comprehensive guide for developers and AI engineers looking to enhance large language model (LLM) applications with structured and unstructured data retrieval. It introduces the concept of GraphRAG, which combines vector search for semantic similarity with graph search for relational queries, offering a more accurate and explainable alternative to traditional vector-only RAG systems. By integrating Neo4j for knowledge graphs and LangChain for orchestration, the tutorial walks through setting up a GraphRAG system, including environment setup, vector indexing, and Cypher query implementation, to create a scalable application capable of answering complex queries with both unstructured and structured data. The guide emphasizes overcoming common challenges in RAG systems, such as hallucinations and retrieval limitations, and highlights the benefits of using a hybrid approach to build more trustworthy and adaptable LLM applications.

The blog post by Alex Gilmore provides a detailed walkthrough on setting up a ReAct agent using LangGraph and MCP tools to interact with a Neo4j graph database, specifically for generating Cypher queries to address movie-related questions. The agent, which can be deployed using either PyPI-hosted MCP servers or locally defined tools, is built with an OpenAI LLM and is capable of executing commands via the command line. It utilizes a combination of components including an LLM, prompts, and tools, with the Neo4j Cypher MCP server providing functionalities like schema retrieval and Cypher query execution. The post outlines the necessary setup and code implementation, including the definition of a local movie recommendation tool and the configuration of MCP server parameters using the uv package manager. Additionally, the agent's architecture is designed to intelligently select and execute the appropriate tools based on user input, and the system prompt guides the agent in handling query errors. The post concludes with instructions on running the agent and modifying the setup for other Neo4j or non-Neo4j implementations, highlighting its adaptability and usefulness as a template for similar projects.

Neo4j has introduced IP filtering for its Aura Virtual Dedicated Cloud and Business Critical tiers, enhancing network security by allowing users to specify permissible IP address ranges for accessing their Neo4j AuraDB graph databases. This feature simplifies network access management without the need for complex cloud infrastructure changes, offering a streamlined alternative to private endpoints, which require more setup. Users can manage IP filters through the Aura Console UI or the Aura Admin API, supporting up to 100 IP ranges for VDC databases and 20 for Business Critical instances, with higher capacities available upon request. Filters can be applied at the project level, ensuring consistent security policies across environments and reducing manual setup. IP filtering is ideal for quickly securing public instances, limiting access to trusted networks, and standardizing access policies, complementing existing private access capabilities to provide more flexibility in data protection. This feature, aimed at making enterprise-grade security more accessible and manageable, is currently available for certain tiers, with plans to expand support in the future.

"This Week in Neo4j" highlights recent advancements and applications in graph databases, focusing on tools and techniques that enhance data management and analytics. Key topics include integrating LangChain4j with Neo4j to create graph-based question-answering systems, enriching Product Hunt data into Knowledge Graphs, and introducing native conditional logic in Cypher 25 to simplify query writing. The newsletter also explores how Neo4j's Graph Data Science library can analyze complex supply networks to understand payment behaviors. Additionally, community engagement opportunities such as the Neo4j User Research panel and various upcoming events are promoted, while featuring a community member, Thomas Orth, who discusses using Generative AI to improve software documentation at Lockheed Martin.

Neo4j GraphQL is a library that facilitates the automatic generation of GraphQL APIs from type definitions, enabling complete CRUD operations, relationship traversals, and optimized Cypher queries without the need for manual resolver writing. It provides robust built-in features such as filtering, sorting, authentication, and authorization, but also allows for the extension of auto-generated resolvers with custom logic through GraphQL middleware. This middleware, implemented using the graphql-middleware library, enables users to apply reusable logic like logging, validation, or authentication consistently across multiple resolvers. A key aspect of Neo4j GraphQL resolvers is their ability to execute a single optimized Cypher query for entire query trees, addressing the N+1 problem common in traditional GraphQL setups. By integrating middleware, developers can maintain the benefits of auto-generated schemas while introducing necessary custom behaviors, making Neo4j GraphQL a powerful tool for efficient and flexible data management.

In the Neo4j Developer Blog post, Finbar Good explores the introduction of the REPEATABLE ELEMENTS match mode in Cypher 25, which allows relationships in a graph database to be traversed multiple times within a single match query. This new feature enhances pathfinding capabilities in Neo4j by accommodating scenarios where backtracking or cycle repetition is necessary, such as optimal pathfinding with constraints, covering path problems, and peer-to-peer reachability in hierarchical structures. The blog illustrates various use cases, including finding optimal routes with constraints, modeling real-world movement like a doctor's rounds, and exploring family trees for common ancestors. However, the author cautions that enabling relationship repetition significantly increases the number of potential paths, necessitating careful constraints to manage computational complexity, such as limiting path length or dividing the problem into smaller, manageable subproblems.

Sudhir Hasbe has been promoted to President of Technology and CPO at Neo4j, where his leadership has significantly advanced the company's offerings in graph databases and analytics, establishing Neo4j as a key component of modern tech stacks. His tenure has been marked by innovations such as the development of a comprehensive cloud-native graph database, enhancements in performance and security, and a focus on making graph technology accessible to all developers. Neo4j's cloud platform, AuraDB, has expanded its reach with significant partnerships with AWS, Azure, and Google Cloud, while Sudhir's vision has driven the company to innovate in areas such as agentic AI and GenAI systems. The introduction of tools like GraphRAG has improved the reliability of GenAI applications, and Neo4j's emphasis on simplifying the developer experience has contributed to substantial growth in cloud consumption. Sudhir's strategic direction aims to continue pushing innovation in graph technology and AI, supporting organizations in moving from GenAI experimentation to the development of intelligent systems.

Matthew Wood, a Product Manager at Neo4j, explores the Neo4j Spatial Plugin, highlighting its capabilities in spatial analysis and its integration with geospatial data. In the final installment of a series, he discusses extending the plugin to incorporate advanced GIS functionalities by using open-source spatial libraries like Java Topology Suite (JTS) and GeoTools. The blog demonstrates how a custom plugin was created to precisely calculate intersection points where a vessel crosses an exclusion zone, a task previously requiring Java skills and now made accessible with the help of ChatGPT for guidance. The plugin successfully identifies entry and exit points of a vessel crossing an exclusion zone, showcasing the potential of combining Neo4j with geospatial analysis for deeper insights. The series underscores the synergy between Neo4j and spatial data, inviting further exploration and enhancement ideas from the community.

The blog post discusses the challenges of moving houses, particularly for individuals with ADHD, and introduces the Neo4j MCP Server as a potential solution for managing such complex tasks. The author shares personal experiences of trying various organizational tools, which ultimately proved ineffective, leading to the exploration of Neo4j's graph database capabilities. By leveraging the Model Context Protocol (MCP) in Neo4j, users can create a personalized AI assistant that organizes tasks into interconnected graphs, akin to the non-linear thought processes typical of an ADHD brain. This system allows for the efficient management of projects, goals, tasks, contexts, and triggers without the need for extensive coding knowledge. The post outlines the setup process using tools like Visual Studio Code and highlights the benefits of using Neo4j MCP in simplifying life organization.

Modern supply chains require more than traditional linear tracking systems; they need real-time visibility and the ability to understand complex relationships between suppliers, logistics, and regulations. Neo4j AuraDB, deployed on AWS, offers businesses a way to model these relationships through knowledge graphs, enhancing supply chain sustainability while meeting environmental, social, and governance (ESG) objectives. Capgemini utilizes Neo4j AuraDB to assist clients in improving sustainability practices by building knowledge graphs that capture emissions data and alternative sourcing paths, allowing companies to align operations with ESG goals without overhauling existing systems. As ESG expectations span industries, businesses face pressure from regulatory mandates, customer demands, and investor scrutiny to track and report ESG key performance indicators effectively. Traditional relational databases are often inadequate for modeling the interconnected nature of supply chains, leading companies to adopt graph databases like Neo4j AuraDB, which provide a dynamic view of supply networks, enabling risk analysis, real-time visibility, and better decision-making. Capgemini's expertise in using Neo4j AuraDB helps organizations visualize supply chain data, understand trade-offs, and make informed choices, while the system's integration with AWS ensures seamless connectivity with existing data infrastructure. This approach not only supports sustainability goals but also becomes a strategic asset for improving supply chain operations across various business functions.

Adam Cowley, a Developer Experience Engineer at Neo4j, discusses the shift from retrieval-augmented generation (RAG)-based chatbots to agent-based systems, highlighting the evolving landscape of generative AI. He notes that although RAG is not obsolete, advancements in large language models (LLMs) and context windows have made traditional methods like chunking documents less efficient. Cowley explains that modern agent architectures, such as ReAct, offer enhanced capabilities by allowing the agent to act on the user's behalf using a list of tools, improving problem-solving and user interaction. He illustrates how these agents can address issues like database connectivity or lesson completion more effectively than RAG-based systems by providing real-time solutions and guiding users through challenges. The article also mentions the Model Context Protocol (MCP) as a new standard for tool integration, enabling developers to enhance AI applications by incorporating third-party tools, which Neo4j leverages to improve the learning experience on its GraphAcademy platform.

This Week in Neo4j provides an overview of the latest developments in graph databases, highlighting the release of the book "Essential GraphRAG" and discussing the Model Context Protocol (MCP) for integrating AI systems with Neo4j. The newsletter explores how Neo4j enhances Retrieval-Augmented Generation (RAG) pipelines by improving reasoning and accuracy, and delves into temporal graph modeling for fraud detection. It also invites users to participate in Neo4j's User Research panel, offering opportunities to influence product development and earn compensation. Upcoming events include livestreams, conferences, and meetups, while workshops and webinars aim to bolster users' skills in graph technology. The edition features community member Joakim Nilsson, who will speak on the environmental benefits of knowledge graphs in supply chain management.

In the third installment of a series exploring the Neo4j Spatial plugin, the article delves into analyzing Automatic Identification System (AIS) data, which tracks vessel locations, using Neo4j's graph database capabilities. It demonstrates the process of creating an AIS data layer, loading and organizing the data, and establishing relationships between observations to recreate vessel journeys. By converting these paths into LINESTRING geometries, the process allows for more complex geospatial analyses, such as detecting intersections with defined areas like exclusion zones. The article concludes by noting the limitations of the current Neo4j Spatial plugin for more advanced analyses, suggesting that further capabilities could be achieved by developing a custom plugin.