Parse JSON · Flow · One field per schema entry
Reads fields out of JSON text and saves each one on the record, so later steps can use them.
When to use it
Use it whenever an earlier step produced JSON you need to pick apart — most often an HTTP request's response body, or an AI research answer that came back as JSON.
What it does
You give it some JSON text and describe the fields you want out of it. For each record, it parses the text and saves one record field per entry in your schema, ready to filter, route, or @-mention downstream.
Configure it
Input — the JSON text to parse. Usually an
@-mention of an earlier step's raw output, like an HTTP request's response body.Schema — one row per field you want out. Each row has:
- Name — what to call the field on the record. - Path — where to find it in the JSON, using dots and brackets: data.email, items[0].id, nested as deep as you need. Leave it empty to save the whole value. - Type — text, number, whole number, yes/no, or a list. Picking the right type matters: a whole number stays 42 rather than becoming 42.0 when it lands in a prompt or a CRM field.
Cost
No credit cost. See Credits in workflows.
Gotchas
If the input isn't valid JSON, that record fails at this step. Turn on Continue on failure if a missing or malformed reply shouldn't stop the record.
You can chain two Parse JSON steps — save a nested object or a list with one, then read deeper into it with the next.