Article

Query Forge: cross-database federated SQL querying without data movement

2026-08-05 · /gluesync/query-forge-federated-sql-querying/

Here’s the core challenge in modern data architecture: when BI teams, analysts or DBAs need to query data across disparate databases, they usually face two painful options. Either they build and maintain complex ETL pipelines to copy everything into a central data warehouse, or they manually run isolated queries across different clients and stitch the results together in spreadsheets.

As part of the expanding ecosystem of Gluesync Modules, Query Forge solves this problem by turning Gluesync Core Hub into a secure, federated SQL gateway. It enables zero-copy cross-database querying, allowing external JDBC clients to execute single SQL queries that join tables living behind completely different agents, seamlessly bridging both relational databases (like PostgreSQL or Oracle) and NoSQL stores (like MongoDB or Couchbase).

Query Forge

How Query Forge works: federated SQL architecture

Instead of establishing new data movement pipelines, Query Forge builds upon Gluesync’s existing Query Studio infrastructure. Core Hub exposes each SQL-capable agent as an accessible SQL schema.

When a client submits a query:

  • Central query planning: Query Forge receives the SQL statement and plans the join across the relevant agents.
  • Intelligent pushdown: to eliminate network congestion, Query Forge translates column projections and filters (such as =, <>, <, >, LIKE, IS NULL and AND/OR/NOT logic) and pushes them down directly to remote source agents.
  • Central evaluation: only the essential filtered rows travel over the wire, where the in-memory query engine powered by Apache Calcite performs complex joins and handles unsupported operations centrally before returning the final results.

Key capabilities and enterprise security

Query Forge is engineered with strict read-only guarantees and fine-grained security protocols.

Dedicated JDBC driver

Ships with a branded JDBC driver (jdbc:gluesync://<host>:<port>) powered internally by Apache Calcite and Avatica, making it instantly compatible with tools like DBeaver, DataGrip, Tableau and custom scripts.

Personal API token authentication

Authentication requires a Core Hub Personal API Token (gsp_…) or local account credentials. There are no shared service credentials.

Per-user scope & security

Every query executes under the user’s specific identity. Users can only discover and query the schemas and tables assigned to their permissions.

Zero-copy read-only safety

Queries run strictly through the read-only execution path, ensuring zero write risk to underlying production databases.

Predictable schema naming

Pipelines with a single agent are exposed directly under the pipeline name (e.g., sales.orders), while multi-agent pipelines use disambiguated schema names (e.g., sales__pg.orders).

Query Forge delivers the querying flexibility of a data warehouse without the operational cost, storage overhead or latency of moving raw data. For detailed configuration steps and connection strings, consult the official Query Forge Documentation.

Back to blog