Data Sources & Schema Preview
Last updated: Mar 2026
Transform and For Each Loop nodes need a data source — they process structured data from an upstream step. When you configure these nodes, the Input Data or Items Sourcesection lets you pick which step's output to work with.
To make this easier, a Schema Previewpanel appears below your selection. It shows you exactly what columns (fields) are available in plain language — no technical knowledge required.
Choosing a Data Source
When you open the configuration for a Transform or For Each Loop node, the first section lists all upstream steps that produce data. Click on a step to select it as your data source.
After selecting a step, you may see additional options:
- Tool results— Structured data returned by tools like Web Search, HTTP Request, or connected apps. This is usually what you want for filtering, sorting, or looping.
- LLM response— The AI's freeform text response. Since this is unstructured text, it's less suitable for data operations.
Data Types
Each column in your data has a type that describes what kind of value it holds. The schema preview uses friendly labels so you always know what to expect.
| Type | What It Means | Example |
|---|---|---|
| Text | A piece of text — a name, URL, description, etc. | "Anthropic launches Claude 4" |
| Number | A numeric value — a count, score, price, etc. | 0.95, 42, 1024 |
| Yes / No | A true-or-false value | Yes, No |
| List of items | A collection of values or objects | [result 1, result 2, ...] |
| Grouped data | A set of named properties bundled together | headers, metadata, nested object |
Schema Preview Panel
Once you select a data source, a preview panel appears below your selection. This panel shows you:
- A plain-English descriptionof what data you're working with (e.g. "Each row of data will have these columns from Web Search")
- A list of available columns with their names and types, displayed like a mini spreadsheet
- An expand button to see full descriptions of each column when you need more detail
Compact View
By default, the preview shows the first 4 columns in a compact layout with just the column name and type. This keeps the configuration panel clean while still giving you a quick overview.
Expanded View
Click "Show all columns with details" to expand the preview. The expanded view shows every column along with a description of what each one contains. This is especially helpful when you're unsure which column to use in a filter or sort operation.
Transform Sources
Transform nodes process and reshape data. In the Input Data section, select the upstream step whose data you want to transform. Common source patterns:
- A Web Search step → Filter and sort the search results
- An HTTP Request step → Pick specific fields from the API response
- A connected app (Google Sheets, Slack, etc.) → Work with rows or messages
- Another Transform step → Chain multiple data transformations
After selecting a source, the schema preview shows the columns available for your transform operations (filter, sort, rename, etc.).
For Each Loop Sources
For Each Loop nodes iterate over a list of items, running their sub-steps once per item. In the Items Source section, select the list you want to loop over.
The schema preview shows what each individual item will look like inside the loop. For example, if you loop over search results, the preview shows columns like Title, URL, and Content — those are the fields each iteration will have access to.
Tips
- Prefer tool results over LLM responses for Transform and For Each Loop nodes. Tool results have a known structure, which means the preview can show you exactly what columns are available.
- Use the expanded viewwhen you're building a filter condition and aren't sure which column to pick. The descriptions explain what each column contains.
- Chain transforms to build complex data pipelines. The output of one transform can feed directly into another.