Data payload format

The canonical reference for the data field in /trigger requests. For tutorials, examples, and common patterns, see Pass input data.

Schema


data: array<row>
row: array<cell>
cell: string

Two levels deep, both arrays. The outer array holds rows, each row holds cells. Cells are strings.

Minimal example


One row of two cells:

"data": [
  ["value1", "value2"]
]

Validation rules


RuleDetail
data is optionalOmit it entirely if your automation doesn't expect input.
data is an array of rowsA flat array such as ["a", "b"] is rejected: Data must be sent as an array of arrays.
Each row is an array of cellsEmpty rows ([]) cause unpredictable behaviour, avoid them.
No third levelA cell that is itself an array is rejected: Data must not be sent as a array with more than two dimensions. One table per call.
Cells are stringsNumbers and booleans are coerced to strings on the server. Pass "42" for clarity.
Cells must not be nullUse "" for missing values. A null cell breaks the row.
Cells must not contain objectsCells are flat scalars. To pass structured data, flatten across columns.
All rows should have the same lengthRagged rows cause column-index references to misalign.

Several kinds of input


There is one table per call. When the automation needs more than one kind of value, give each its own column:

"data": [
  ["https://example.com/a", "browser automation"],
  ["https://example.com/b", "web scraping"]
]

Inside the automation, Insert data picks column 1 or column 2 of the current row.

Header rows


The automation sees every row you send, including a header. A Loop through data step works through all of them, so include a header row only when the automation was built to expect one.

Encoding


  • The data field is JSON, sent as part of the application/json request body.
  • UTF-8 throughout. Non-ASCII characters in cell values are passed through unchanged.
  • No size limit is documented per cell or per request, but very large payloads (multiple megabytes) may hit timeout or rate limit issues. For high-volume input, batch across multiple /trigger calls instead of one giant payload.