Database support

Database / mechanism	Status	Mechanism and notes
PostgreSQL	Supported	Native pg adapter; information_schema/pg_catalog introspection, pooled connections, server-side query cursor behavior, and EXPLAIN (FORMAT JSON).
MySQL	Supported	Native mysql2/promise adapter; MySQL metadata queries, pooled connections, and EXPLAIN FORMAT=JSON.
MariaDB	Supported	Uses the MySQL wire-compatible mysql2/promise adapter. Server/version-specific SQL or metadata differences can limit introspection.
SQL Server	Supported	Native mssql/Tedious adapter; metadata queries and SET SHOWPLAN_XML ON in an isolated transaction for plans.
Oracle	Supported	Native oracledb thin-mode adapter; Oracle metadata and EXPLAIN PLAN with available row, byte, cost, time estimates, and predicates. No Oracle Instant Client is required for thin mode.
Generic JDBC	Experimental	JVM sidecar loads a user-provided driver JAR and java.sql.Driver class, then connects through the supplied JDBC URL.
ODBC	Experimental	Generic ODBC adapter with bounded query and stream support. Requires a separately installed 64-bit driver compatible with the target database.
MongoDB	Native + read-only SQL	Extended JSON find/aggregate, explicit writes, metadata, autocomplete, and /mongo SELECT conversion. SQL mode executes supported SELECTs natively, with DuckDB for complex queries. See MongoDB usage.

Generic JDBC

Configure the JDBC connection with:

• endpoint: JDBC URL;
• options.jarPath: filesystem path to the driver JAR;
• options.driverClassName: class implementing java.sql.Driver;
• username and password as normal connection credentials.

The JVM sidecar loads the JAR dynamically and keeps connections in memory. It uses standard DatabaseMetaData for schemas, tables, views, columns, and best-effort primary-key flags. Drivers without schemas use a default bucket. Queries are capped at 10,000 rows and use driver-dependent fetch/max-row settings. Values outside JSON primitives are stringified; byte arrays are Base64 encoded. BigDecimal, BigInteger, and integers beyond JavaScript’s safe range are transported as decimal strings. H2/JDBC streaming has integration coverage; other vendor drivers need separate validation.

Limitations are intentional because JDBC drivers differ:

• EXPLAIN is unsupported; JDBC has no portable plan syntax.
• getDefinition is unsupported; there is no portable CREATE-text query.
• Functions, indexes, foreign keys, and defaults are not exposed reliably by arbitrary drivers.
• Primary-key discovery is best effort and may be unavailable.
• Driver-specific SQL, authentication, TLS, transactions, type conversion, and metadata quality remain the responsibility of the selected driver.
• A missing JAR, invalid driver class, rejected URL, credentials error, or network error is reported by the sidecar; it is not silently converted into native-driver behavior.

Native adapters should be preferred when available because they provide dialect-specific introspection and explain support.

ODBC

Configure the connection with a DSN or ODBC connection string, plus username and password when required. Install a 64-bit ODBC driver for the target database separately; the driver manager alone does not provide a database driver.

The adapter discovers tables, views, columns, and best-effort primary keys through the driver’s metadata APIs. It supports bounded query results and streaming for local analysis. SQL syntax, authentication, TLS, metadata quality, and type conversion depend on the selected driver.

Execution plans, indexes, object definitions, functions, and result-grid row edits/inserts are not currently supported. For analytical streaming, date/time or numeric values whose precision cannot be preserved must be cast to text in the source SQL; otherwise the import fails explicitly. See the local analysis guide for import behavior.