Node types#

A flow is built from four node types. Each one has a distinct shape and colour on the canvas so it's readable at a glance.

Start#

The entry point of a flow. A flow has exactly one start node — the editor will warn you if you try to delete it or add a second.

Use the start node's description to explain what triggers the flow ("User clicks 'Sign up' on the marketing page", "GitHub webhook receives a push to main", etc.).

Step#

A single action or activity. Most nodes in a typical flow are steps. They have one input and one output by default — connect them in a chain to describe a linear sequence.

A good step description answers "what happens here, and who/what does it".

Decision#

A branching point. A decision node has one input and two or more outputs — label each outgoing edge with the condition that takes you down that branch (yes / no, paid / trial, etc.). Connections can also carry a typed payload — see Connections.

End#

A terminal state. A flow can have multiple end nodes — one for each distinct outcome (signed up, cancelled, error). Naming them clearly makes the flow much easier to read.

A worked example#

A minimal sign-up flow might look like:

[Start: user opens /signup]
        ↓
[Step:  user submits email + password]
        ↓
[Decision: email already in use?]
   yes ↓                no ↓
[End: show login link]   [Step: send confirmation email]
                                ↓
                         [End: account created]

Five nodes, four edges, two outcomes. Most useful flows stay under 20 nodes — if yours is growing past that, consider splitting it into multiple flows that link to each other in their descriptions.