Concepts

Every automation is built from the same small set of ideas. A trigger decides when a run starts, actions do the work, conditions route the flow down one path or another, and loops repeat a section once per item in a list. Variables carry data between all of them, connections hold the authentication that lets a step reach an external system, agents define how an AI step behaves, and human-in-the-loop parks a run until a person answers.

This section explains what each of those things is and how it behaves, without going node by node. Once a kind makes sense here, the catalog pages for individual nodes should need no further explanation: see Built-in nodes for the nodes that ship with the product and Integrations for the ones each connected system adds.

Page	What it covers
Triggers	How a run starts, what the trigger contributes to it, and how to stop unwanted events from starting one.
Actions	The steps that do work—their inputs, their outputs, and what happens when one fails.
Conditions and branching	Routing a flow down a true or a false path, and what happens to the path not taken.
Loops and iteration	Repeating a section of the flow once per item, in series or in parallel, and collecting what each pass produced.
Variables and secrets	Referencing data from earlier steps, defining reusable values, and keeping sensitive ones encrypted.
Connections	Saved, authenticated links to external systems, and which nodes each one powers.
AI agents and models	What an agent is, and how a step decides which provider account and model its work runs on.
Human in the loop	Pausing a run for a person, and where the waiting requests are answered.