Articles in this section

Flows overview

A flow is the unit of automation in integrator.io: it retrieves records from a source application, optionally enriches, transforms, filters, or routes them, and sends them to one or more destination applications. Every flow belongs to an integration, runs on a schedule, in real time, or on demand, and reports its runs and errors on the Dashboard. This article explains what a flow is made of, how records move through it, and where to go to build, run, and manage one.

The steps in a flow

A flow is a sequence of steps. It starts with at least one source step and ends with at least one destination step; any other steps sit between them.

Step kind What it does Where it's explained
Export Retrieves records from a source application on the flow's schedule or on demand. What are exports?
Listener Receives records that a source application pushes to a webhook URL in real time. Introduction to webhook listeners
Data loader Uploads a file you choose as the flow's one-time source. Create a Data loader flow
Transfer Moves files out of a source, or into a destination, without changing them. Create a flow
Lookup Retrieves additional data for each record from another application and merges it into the record. Add lookup, transfer, and other steps to a flow
Import Creates, updates, or deletes records in a destination application. What are imports?
Tool, AI agent, guardrail Calls a reusable tool, an AI agent, or a guardrail check on each record. Use tool import steps in a flow, Create an AI agent, Understand Celigo guardrails
Publish Publishes each record as a message to a topic that other flows subscribe to. Publish messages from a flow
Branching Routes records down different paths based on conditions. Apply branching to your flow

Each step also has its own options — field mapping, transformations, filters, and hook scripts — that you set on the step itself; see Configure flow step options.

How records move through a flow

An export retrieves records in pages of 20 by default. Pages are processed in parallel, so records from one page can finish before records from an earlier page. A page can't be larger than 5 MB, and neither can a single record.

A flow with several sources runs each source as its own job, and the records from one source pass through every later step before the next source's records start, unless you set the flow to run its sources in parallel; see Configure flow settings. Listeners don't use pages: each message a listener receives runs through the flow on its own.

How a flow runs

A flow runs only while it's turned on. Enabled on the integration's Flows tab or in Flow builder turns it on; see Monitor flows in an integration. A flow that's on runs in one of three ways:

  • On a schedule — at the cadence you set, from every 15 minutes to once a week, or on a cron expression. See Schedule a flow.
  • In real time — whenever its listener receives data or a message is published to a topic it subscribes to. A real-time flow has no schedule.
  • On demand — when you select Run. See Run a flow.

Each run appears on the Dashboard with its status, record counts, and errors; see Monitor flows on the Dashboard.

Flows and integrations

Every flow belongs to an integration, and flows in the same integration can share connections and run one after another. Flows that were created before integrations were required live in an integration named Standalone flows; you can move them into any other integration. See Integration page tabs and buttons and Move a flow to another integration.

Inside an integration, flows can be grouped into flow groups; see Organize flows and APIs into groups.

Where to work with flows

To build many copies of one flow that differ only in their connections and settings, use a multi-instance flow; see Learn more about multi-instance flows.