The Transform node takes a JSON string produced by an earlier step (most commonly raw text returned by an AI step) and converts it into a structured object you can use in later steps of your flow.
This is the node you reach for whenever a previous step gives you "text that looks like JSON" instead of an actual usable object — for example, when an AI model returns its answer as a block of text rather than structured data.
Use a Transform node right after any step whose output is a string containing JSON, such as:
An AI/LLM step (e.g. DeepSeek, GPT) that was asked to return JSON
An HTTP Request step that returns a JSON body as raw text
Any step where you see .text in the output and need to work with its contents as fields, not as one long string
Field Required Description Step name Yes A name for this step's output, used to reference it later in your flow. Must start with a letter or underscore (no spaces or special characters). Input expression Yes A Handlebars expression pointing to the JSON string you want to parse — for example {{deepseekResult.text}}. Click + variable for a list of common variable names to insert.
Once parsed, the result is stored under the Step name you chose. You can access any field from it in later steps using standard Handlebars syntax.
Example:
Setting Value Input expression {{deepseekBom.text}} Step name bomReport
Later steps can then reference: {{bomReport.[0].style_id}}
The .[0] is used when the parsed JSON is a list of items — it grabs the first item in the list. Drop the .[0] if your JSON is a single object rather than a list.
If the AI step wrapped its JSON output in a code block (```json ... ```), Transform automatically strips those markers before parsing — you don't need to clean this up yourself.
If the input expression doesn't resolve to valid JSON, the step will fail with an error showing a preview of what it tried to parse. Double-check the source step actually returned JSON-formatted text, and that your Input expression points at the right field.
The original data from earlier steps is preserved — Transform only adds your new field, it doesn't remove anything else available in your flow.
Navigate