Skip to main content
The Data explorer app lets you inspect data from existing connections without creating or running a pipeline. Use it to preview a table or stream, or to run a guarded SELECT or WITH SQL query against a supported connection.

​
Open Data explorer

In Mage Pro, select Data explorer from the sidebar, or open Apps and select Data explorer. The page is available at:

​
Supported connections

Data explorer lists existing connections from:
  • io_config.yaml SQL provider profiles.
  • Saved data integration source configs.
  • Saved data integration destination configs.
SQL-capable connections can run queries. SQLite and DuckDB configs are preview-only in Data explorer V1. Connections with neither preview nor query support are shown as unavailable so you can see why they cannot be explored instead of silently missing them. For supported io_config.yaml SQL profiles, Data explorer can browse metadata without loading the full warehouse inventory in the initial connection list. It lazily loads catalogs or databases, schemas, tables, and columns as each level is selected. When metadata loads, the browser selects the first available catalog or database, schema, and table by default, loads columns for the selected table, and keeps the selected items highlighted. The metadata browser includes its own refresh control for the selected connection. Schema browsing is currently supported for DuckDB, Microsoft Fabric, MySQL, PostgreSQL, Redshift, Snowflake, SQLite, and Microsoft SQL Server IO config profiles. Other SQL-capable IO config profiles can still preview or query data when their connection settings are complete, but you must type the table name or SQL query manually. Viewers can open Data explorer and review the connection inventory. Editor access is required to preview a table or run SQL. Saved source and destination configs follow the permissions of the pipeline that owns the config: users only see integration configs from pipelines they can view, and they must be able to edit the owning pipeline before running a preview or SQL query through that config.

​
Preview a table or stream

  1. Select a connection from All connections, IO configs, Sources, or Destinations.
  2. For io_config.yaml connections with metadata support, use the schema browser to confirm or change the selected catalog or database, schema, and table. The first available table is selected by default, and selecting another table fills the preview target and the starter SQL query.
  3. For saved source or destination configs, choose a table or stream when options are available. The selected value becomes the preview target and the starter SELECT * FROM <table_or_stream> SQL query.
  4. If the connection does not provide table options, or the schema browser does not load a matching table, type the table name.
  5. Set a row limit.
  6. Click Preview table.
Preview requests run a limited SELECT * FROM <table> query.

​
Run a SQL query

  1. Select a SQL-capable connection.
  2. Enter one SELECT or WITH query.
  3. Set a row limit.
  4. Click Run query.
Data explorer accepts one SELECT or WITH statement at a time. Mutating statements such as INSERT, UPDATE, DELETE, DROP, locking clauses such as FOR UPDATE or FOR SHARE, and multi-statement SQL are rejected. Data explorer also rejects SELECT or WITH queries that call known unsafe functions with side effects, such as functions that update sequences, terminate sessions, sleep, read server-local files, or execute provider-specific system commands. The server also clamps row limits to protect the workspace from accidentally large previews. These safeguards are not the same as database-enforced read-only execution for every provider. PostgreSQL and Redshift preview and query execution start a database read-only transaction before running the query and roll it back afterward. Metadata browsing uses server-generated SELECT queries with provider timeout settings. Metadata discovery, preview, and query execution are also bounded by server-side deadlines. For providers with native timeout controls, Data explorer passes those settings to the driver or query job. If a provider call exceeds the relevant deadline, Data explorer attempts provider-native cancellation where available, closes or stops the bounded execution operation, and returns a timeout error. Other supported providers rely on statement validation, restricted-function checks, provider timeout or lock-timeout settings where available, bounded server execution deadlines, and row-limit enforcement. For those providers, use credentials that already have the least privileges needed for exploration.

​
Review results

After a preview or query succeeds, Data explorer shows the returned rows in a table. Use the column headers to sort the returned rows, or open a column filter to narrow the rows shown in the current result. These sort and filter controls apply only to the returned sample. To filter the full source table, add the filter to the SQL query. Returned row counts describe the loaded preview window, not the total number of rows in the source table. To answer full-table cardinality questions, run an exact COUNT(*) query against the selected table or filtered query. When a tabular result is available, click Expand results to hide the connection, schema, and query browser and give the results table more space. Click ← Back to restore the explorer panels. Data explorer remembers this layout preference for future result tables.

​
API

The UI uses the data_explorations API resource. GET /api/data_explorations returns each connection with fields such as uuid, label, kind, provider, profile, pipeline_uuid, block_uuid, supports_sql, supports_preview, supports_metadata, table_options, and redacted metadata. To browse schema metadata for a supported IO config connection, send a nested data_exploration payload with action_type: "metadata":
Metadata requests accept metadata_type values of databases, schemas, tables, or columns. Table metadata requires schema; column metadata requires both schema and table. The optional database field is used by providers that need a catalog or database qualifier, such as Snowflake. Metadata limits are clamped separately from preview and query row limits. To preview or query data, send a nested data_exploration payload:
Responses include success, columns, rows, metadata, and error_message. For query responses, metadata.returned_row_count is the number of rows returned in the preview window, metadata.preview_limit_reached indicates whether the window reached the requested limit, and metadata.row_count can be null when the API did not run an exact count. Treat metadata.row_count as full-result cardinality only when metadata.row_count_mode reports that an exact or estimated count is available. Connection metadata and error messages are redacted so credentials are not exposed in the response.