Database

Query automation databases with structured operations or raw SQL

Overview

The Database step lets you read, write, and query your automation databases. Choose from structured operations (Read, Insert, Update, Upsert, Delete) with a visual builder, or write raw SQL with the Query operation.

Configuration

Field	Description
Database	Select a connected automation database
Operation	Read, Insert, Update, Upsert, Delete, or Query
Table	Target table (for structured operations)

Operations

Operation	Description	Fields
Read	Fetch rows from a table	Column toggles to select which columns to return
Insert	Add a new row	Set field values for the new row
Update	Modify existing rows	Set field values + where clauses to match rows
Upsert	Insert or update	Set field values (inserts if not found, updates if exists)
Delete	Remove rows	Where clauses to match rows to delete
Query	Run raw SQL	SQL editor with template variable support and AI generation

Where Clauses

For Update, Delete, and Read operations, use where clauses to filter rows. Each clause specifies a column, operator, and value.

Available operators:

Operator	Description
=	Equals
!=	Not equals
>	Greater than
<	Less than
>=	Greater than or equal
<=	Less than or equal
LIKE	Pattern match (case-sensitive)
ILIKE	Pattern match (case-insensitive)
IN	Matches any value in a list
NOT IN	Does not match any value in a list
BETWEEN	Within a range
IS NULL	Value is null
IS NOT NULL	Value is not null

Output

Field	Type	Description
rows	array	Result rows
count	number	Number of affected/returned rows
error	boolean	Whether an error occurred
error_message	string	Error description if failed

Throwing Errors from SQL

For the Query operation, you can fail the step from inside your SQL by selecting an _error column. If any returned row has a non-empty _error value, the step fails and the value is surfaced verbatim as the step’s error message. The _error column is stripped from the rows on the happy path, so downstream steps never see it.

SELECT
  CASE WHEN COUNT(*) = 0 THEN 'No matching records found' END AS _error,
  id,
  name
FROM contracts
WHERE status = {{ trigger.status }}
GROUP BY id, name;

Use this to enforce data preconditions (missing records, validation failures, duplicates) and produce a clear user-facing error message without writing a separate Validator step.