May 2026 Summaries

Memgraph 3.10 is a production-focused update that enhances the reliability, security, and manageability of graph applications under real workload conditions by strengthening high availability stability and enterprise multitenancy features. This release introduces improved GraphRAG workflows, allowing more flexibility for AI workloads by connecting Memgraph to various text embedding providers and enhancing vector and text index performance. It also includes a new dedicated AI Platform license tier, which aligns licensing with actual usage patterns for embedding-heavy workloads, and adds a GNN import/export module to facilitate the movement of trained models. The update improves memory management, garbage collection, and operational hardening, making Memgraph more resilient under production pressure. Additionally, Memgraph 3.10 offers better server-side descriptions for client applications, implements Kerberos SSO authentication for integration with enterprise identity systems, and enhances multitenancy robustness by refining access control, memory tracking, and tenant profiles. The release also brings schema descriptions into Memgraph Lab, aiding in schema exploration with human-readable context. These improvements make it particularly valuable for teams operating in enterprise, AI, or long-term production settings where reliability and operational clarity are crucial.

Previsant's Adaptive Intelligence Platform is designed to integrate disparate AI tools and data sources into a cohesive architecture that enhances the functionality of AI applications across various domains such as healthcare, insurance, and manufacturing. By leveraging Memgraph as the graph database layer, the platform enables the connection of unstructured documents, structured data, and AI agents, facilitating a comprehensive context layer for querying relationships, rules, and extracted entities. This approach supports knowledge graph applications and cross-system data lineage, allowing users to perform document exploration, relationship analysis, and conflict detection in complex workflows. Through live demos, Previsant demonstrated the platform's capability to assist in insurance claims research and data lineage tracing, showcasing its potential to improve efficiency and transparency in AI workflows by enabling users to inspect, query, and trace data back to its source.

Memgraph Zero is a new product line that addresses the challenges of querying distributed data without the need for ETL by enabling the data to remain in its original locations while still being accessible as a graph. The first component, MemGQL, was introduced by Memgraph CTO Marko Budiselic, offering a federated GQL layer that allows for querying across different data sources such as Postgres, ClickHouse, and MySQL, through a graph model. MemGQL is designed to support a "multi-player" mode for AI agents, allowing them to share and reuse data findings rather than duplicating efforts. Implemented in Rust and utilizing the Bolt protocol, MemGQL facilitates the translation of queries into backend-native languages like Cypher and SQL, leveraging existing storage and execution capabilities of target systems. The architecture centers on a virtual graph, avoiding unnecessary data copying and enabling the execution of queries directly on source systems. While still developing features such as authentication and advanced pushdown, MemGQL aims to integrate with various systems, providing a cohesive interface for both developers and AI agents to work with distributed and heterogeneous data sources.

Memgraph has introduced Memgraph Zero and MemGQL, aimed at simplifying graph intelligence without requiring data movement. Memgraph Zero allows users to query data in place, eliminating the need for complex ETL processes, reducing data staleness, and minimizing storage duplication. MemGQL, the first product in the Memgraph Zero line, is a federated query engine that adheres to the new international standard for graph query languages, ISO/IEC 39075. It allows users to execute queries across multiple data backends without moving their data, supporting connections to various databases such as Memgraph, Neo4j, PostgreSQL, and MySQL. MemGQL is available in community and enterprise editions, with the community version offering free access to key features, while the enterprise version provides unlimited connectors and support. The platform is designed to support agentic data access, providing a centralized semantic layer for AI agents to query and traverse data across different sources seamlessly.