Databases

DataMaker supports PostgreSQL, SQL Server, Oracle, MySQL and MongoDB drivers. Db2 is optional and appears only on deployments where its driver is installed. GET /connections/types returns the available drivers and is the authority for your installation.

Connect

In Connections, add a database, choose its engine and enter the connection details. Prefer a read-only database account for source inspection and extraction, and set DataMaker's read-only option where appropriate. A destination needs separately reviewed write privileges.

Test the connection, then inspect tables or collections, columns, keys and row counts. The API uses POST /connections/test and GET /connections/tables; use the deployed reference for their inputs.

A successful test checks connectivity and credentials, not that every future query or target write will succeed. Confirm schema permissions and network access from the execution host.

Understand the dialect

Relational drivers accept their engine's SQL. MongoDB uses an MQL JSON envelope, not SQL. It takes three forms, each naming its collection:

{"collection":"orders","filter":{"status":"open"},"limit":100}
{"collection":"orders","pipeline":[{"$group":{"_id":"$status","n":{"$sum":1}}}]}
{"collection":"orders","insert":[{"id":1}]}

Write intent is read from the envelope, not from the verb you had in mind. An insert is a write, and so is an aggregation whose pipeline carries $out or $merge - both replace or upsert a collection - so a read-only connection refuses them. Check which of your pipelines write before pointing one at a source you meant to leave alone.

MongoDB field discovery uses a collection validator or sampled documents. It does not infer relational foreign-key constraints, and sampled fields may not describe every document.

Build templates or extract data

Database analysis can create direct-SQL fields that read source values, or relational synthetic templates that use the schema. Review the chosen mode and table dependencies before saving. See Schema discovery.

For larger extraction and preparation, use data jobs. For custom SQL or orchestration, use a reviewed scenario. Never assume a source query has been masked simply because it is executed inside DataMaker.