Your data lives in Postgres, ClickHouse, Iceberg, Memgraph, Neo4j — maybe all of them at once. Migrating everything into a single graph database isn't always realistic. The dataset is too large, the pipeline is too complex, or the cost doesn't justify it. But you still need to query relationships across those sources, and your agents need a single interface to find the data they need.

MemGQL is Memgraph's answer: a federated GQL query engine that translates ISO-standard GQL into backend-native Cypher or SQL and executes it where your data already lives. No ETL. No migration. You define the mapping from your SQL schema to a graph model — with an agentic assistant (Structured2Graph) to help you build it — and start querying via any Bolt-compatible client: mgconsole, Memgraph Lab, or your own application drivers.

In this session, Memgraph CTO Marko Budiselic will present MemGQL for the first time, walking through the architecture and showing live demos of federated graph queries across multiple data sources.

What MemGQL is and how it works — the data flow from GQL through the Bolt protocol to Cypher or SQL backends, including multi-connection mode for querying multiple sources simultaneously

From SQL schema to graph model in minutes — how the Structured2Graph agent maps your existing relational data to a queryable graph model so you can start running GQL without moving a single row

MemGQL as a semantic layer for AI agents — how a single GQL interface gives agents structured access to your entire data landscape through one endpoint, powered by Memgraph's MCP server

Whether you're evaluating graph infrastructure but can't justify a full data migration, running a multi-database environment that needs a unified query layer, or building AI systems that need structured access to data scattered across backends — this session will show you how MemGQL fits into the picture.

Introducing MemGQL: One Query Language Across All Your Data Sources