Starting a Workflow

View as Markdown

Every run of a workflow begins at its Start node, and most workflows need some information to work with: a customer’s name, a policy number, an uploaded document. This page explains how to describe what a run needs, so the right information is there before any step runs.

Describing the information a run needs

On the Start node you can define an input schema: a short list of the fields a run expects. For each field you give a name and a type (text, number, and so on), and you mark which ones are required. For example, a quote-intake workflow might require a customer_name and an email, and accept an optional priority number.

Defining these fields has two benefits:

  • Runs are checked before they begin. If someone starts the workflow without a required field, or sends a number where text was expected, the run is turned away with a clear message instead of failing partway through.
  • The shape is documented. Anyone starting the workflow can see exactly what to provide.

Defining an input schema is optional. A workflow with no schema accepts any information, or none at all.

Whatever a run is given becomes available to your steps as values on the Start node, so later steps can read customer_name, email, and the rest. See Variables & Data Flow.

Attaching files

A run can also start with files attached, such as a PDF application or a CSV of contacts. Uploaded files are stored securely and made available to steps that work with them, like the Agent, which can read a document and answer questions about it.

A few limits apply to file uploads:

LimitValue
Files per runUp to 10
Size per fileUp to 25 MB
File typesPDF, Word, Excel, PowerPoint, text, CSV, RTF, and common images (JPG, PNG, GIF, BMP, WEBP)

Ways to start a run

A workflow can begin in more than one way:

  • Manually, by starting it yourself to try it out or run a one-off.
  • On a request from another system, using a webhook. See Triggers & Webhooks.
  • On a schedule, so it runs automatically at set times.

However a run starts, the same input schema and file rules apply, so the information reaching your steps is consistent.