01 / Costly pipelines
ETL just to ask a question
Teams stand up warehouses and extraction jobs so BI and DBAs can join data that already lives behind Gluesync agents.
Gluesync Module · Federated SQL queries
Query Forge
A secure federated SQL gateway on Core Hub. Run cross-database JOINs across isolated agents with zero raw data movement.
Why Query Forge
Cross-database answers without warehouse tax
02 / Manual stitching
Isolated client queries
Without federation, analysts run separate queries in different tools and reconcile results in spreadsheets — slow and error-prone.
03 / The platform outcome
Zero-copy federation
Query Forge turns Core Hub into a JDBC gateway so one SQL statement can join remote schemas without copying raw datasets.
Capabilities
Federated SQL on the infrastructure you already run
Built on Query Studio infrastructure, Query Forge exposes each SQL-capable agent as a schema and plans joins centrally with intelligent pushdown.
Federated architecture
How Query Forge works
Plan centrally, push filters down, join in memory
Instead of new data movement pipelines, Core Hub receives the SQL statement, pushes projections and filters to remote agents, then evaluates joins with Apache Calcite.
- Central query planning across relevant agents
- Pushdown for projections and filters (=, <>, <, >, LIKE, IS NULL, AND/OR/NOT)
- Only filtered rows travel the wire for central join evaluation
- Bridges relational engines and NoSQL stores behind agents
- Lake datasets on S3, WebDAV and NFS join live databases in the same query
SELECT c.name, SUM(o.amount)FROM pg.sales.orders oJOIN ora.crm.customers c ON c.id = o.customer_idWHERE o.status = 'paid' AND c.region = 'EMEA'GROUP BY c.name- PostgreSQLsales.orders
2.4M rows stay at the source18,240 sent
Runs on PostgreSQL
WHERE status = 'paid' - Oraclecrm.customers
310K rows stay at the source4,115 sent
Runs on Oracle
WHERE region = 'EMEA'
Core HubJoins and sums 22,355 rowsinstead of 2.71M · nothing copied, no pipeline
- Holloway Ltd€1.42M
- Webb & Co€890K
- Brandt AG€234K
Multi-source querying without data movement
Query across relational, legacy, NoSQL and data-lake environments from a single development tool or BI connection. Gluesync resolves schemas and returns results transparently.
Standard connectivity
Dedicated JDBC driver
Connect DBeaver, DataGrip, Tableau, and custom scripts
A branded JDBC endpoint (jdbc:gluesync://host:port) makes Core Hub feel like a single SQL system to external clients.
- Personal API Token authentication (gsp_…) or local account credentials
- No shared service credentials
- Predictable schema naming for single- and multi-agent pipelines
- Works with standard BI and SQL tooling
SQL clients
- DBeaverSQL IDE
- DataGripSQL IDE
- TableauBI
JDBC gateway
One connection string. The caller's token. Read-only pushdown to each agent.Agents
- PostgreSQLanalytics
- OracleERP
- Db2IBM i
Unified JDBC gateway
One connection string. Many agents. Warehouse-style querying flexibility without storage overhead or continuous extraction latency.
Enterprise safety
Read-only security model
Per-user scope with zero write risk to production
Every federated query runs under the caller’s identity through the read-only execution path, with permissions scoped to pipelines and agents.
- Strict read-only execution path
- Users only discover schemas and tables they are allowed to see
- Fine-grained permission scoping at pipeline and agent level
- Zero-copy operations — no warehouse mirror required
Settings · Query data access
Query data access
Connection mode
Schema and table rules
public
*
erp
customers
PII unmask grants
Viewers can never unmask, even with a grant.
SQL-callable AI · Available in Gluesync 2.3
An AI conversation, returned like any other query.
Query gluesync.ai from your JDBC client. The Hub uses your configured model and governed MCP tools, then returns one structured row you can inspect, store, or continue with a conversation ID.
Available in Gluesync 2.3 · Read-only tool path
SELECT answer, proposed_sql,
citations, conversation_id
FROM gluesync.ai
WHERE question = 'who made you';answerproposed_sqlcitationsconversation_id
MOLO17 built me into Gluesync AI Hub.SELECT …[2]conv_8f12…
proposed_sql is returned for review. It is never executed automatically.
One JDBC surface
gluesync.ai appears through Query Forge alongside the schemas you already query, using the same authenticated session.
A result made for code
Receive answer, proposed_sql, citations, and conversation_id as columns—not an opaque chat response.
Your model, caller identity
The request uses the Core Hub LLM provider vault. MCP discovery and invocation remain bound to the JDBC caller’s RBAC.
Suggestions stay suggestions
The answer may include SQL for a next step, but Gluesync never runs proposed_sql automatically. You decide what executes.
Explore the platform
Continue exploring Gluesync
Query Forge extends Query Studio into a federated gateway. Explore the workbench, Knowledge querying, or the full Gluesync platform.
Lake queries
Join Parquet, CSV and JSON datasets in your lake with live databases in one federated query.
Explore Lake queries →
Query Studio
Embedded SQL workbench in the Control Plane for schema browsing and live validation.
Explore Query Studio →
Schedules and events
Chronos schedules, chained workflows, webhooks, and Core Hub event triggers.
Explore Schedules and events →
Gluesync
Real-time CDC plus federated querying under one Core Hub operational interface.
Explore Gluesync →Resources
Guides, roadmap and support
Documentation
Technical guides, API reference, deployment tutorials.
Read documentation ↗
Public Roadmap
What we're building next. Submit feature requests.
View roadmap ↗
Support
Access technical support and operational assistance.
Get support →
Service Status
Real-time platform availability and incident history.
Check status ↗Federated SQL without the warehouse tax
Request a free trial or talk with the MOLO17 team about deploying Query Forge on your Core Hub.