Extract Field

Pull a specific field or dot-notation path from upstream JSON data.

Overview

The Extract Field node reads structured JSON from an upstream node and extracts one field (or a path of nested fields) as its output. It has two input modes: a Dropdown that lists fields detected from the upstream schema, and a Custom path mode where you type a dot-notation path manually. The output can be a single text string, a list of items for fan-out, or a raw JSON value for chaining into another Extract Field or JSON Process node.

This node executes inline (no job created, no credits charged).

When to Use

• Pull title or url out of a Google Search or RSS result set before feeding a List fan-out
• Extract a nested property like authorMeta.name from TikTok posts
• Convert a JSON array from Web Scrape into a text list for Generate Text
• Chain two Extract Field nodes to drill into deeply nested JSON

Configuration

Field	Type	Default	Description
Field	dropdown / text	""	The field or dot-path to extract. Empty = whole item (returns each array element as-is). Detected options appear when an upstream node is connected.
Mode	toggle (auto)	dropdown	Auto-selected. Switches to Custom path… for arbitrary dot-notation entry.
Output Type	select	text	How the result is emitted to downstream nodes (see below).

Output Type options

Value	Behaviour
text	Values are joined by newline into a single string. Works with any text-accepting node.
list	Each match is a separate list item — supports item:N, fan-out, and List-aware nodes.
json	Raw JSON value — use when feeding another Extract Field or a JSON-consuming node.

Keeping fields paired

To use two fields of the same JSON array together — for example prompt and negative from a list of concepts — add one Extract Field per field, both with Output Type list, and wire each into its own input of the downstream node (directly with both wires set to Each, or through two columns of a List).

When the lists are fanned out together they are paired by array element: run 3 gets element 3’s prompt and element 3’s negative. An element with no negative gives that run no negative; it does not shift the ones after it. A run is never started for an element that has no value at all, so a field that only some elements carry (a videoUrl that only some posts have) still fans out over exactly the values that exist.

Addressing by position is unchanged: Item N, ranges and Bundle count only the values that exist, and the text output is still those values, one per line.

If the upstream is an LLM, it must return raw JSON — no code fence and no sentence before or after it — or the node fails with “Input is not valid JSON”.

Inputs & Outputs

Inputs: Any upstream node that emits JSON, a JSON string, or a list.

Outputs: text — the extracted value(s) in the selected format.

Pricing

Free — no credits charged.

Common Use Cases

• Web Scrape → Extract Field (title) → List → Generate Image (one per headline)
• Web Scrape (RSS) → Extract Field (url, list) → Selector → text-prompt chaining
• JSON Process → Extract Field (nested path) → Generate Text
• TikTok scrape → Extract Field (caption) → Generate Text for remixing captions

Tips

• Connect an upstream node first — the dropdown auto-populates with detected fields from the upstream schema, saving you from memorising key names.
• Leave Field blank and set Output Type to list to emit each top-level item of an array individually — useful when the JSON is already a flat list of values like ["a", "b", "c"].
• Use dot notation in custom mode for nested objects: authorMeta.name, videoMeta.duration.
• Chain two Extract Field nodes when the path has array segments in the middle (e.g., first extract items, then extract items[].title by chaining a second node with field title).