Docs / Strand / nodes/overview

Node types

Nodes are the steps a workflow is built from. Each one takes input from the nodes connected before it, does one thing, and passes its result on. This page is the index; each node has its own page with configuration and examples.

A workflow containing a transform, a Python snippet, a function call, a logic branch, an approval, a delay, an HTTP request, and a print node. One of each node type in a single flow.

Calling out#

Node Use it to
HTTP Request Call any HTTP API directly, without a saved connector
Connector nodes Use a saved connector's credentials to reach an external service
AI connectors Send prompts to a configured AI provider
Connector outputs Shape and read what a connector node returns

Control flow#

Node Use it to
Logic Branch execution on a condition
Delay Pause before continuing
Approval Pause until a human approves or rejects
Flow call Run another workflow as a step
Function call Call a reusable function

Working with data#

Node Use it to
Transform Reshape a payload between steps
Python snippet Run sandboxed Python for logic that templating can't express
Print Write to the run log for debugging

Documenting the canvas#

Node Use it to
Comment Leave a text note on the canvas. It is a no-op: the event passes through unchanged

Choosing between them#

Reach for Transform and templating first — they cover most data reshaping and are easier to read in a run record. Use a Python snippet when the logic genuinely needs code; it runs under resource limits described in Python snippet resource limits and the sandbox rules in Python security.

Use a connector node rather than a raw HTTP request whenever a connector exists for the service — the credentials stay centralized and out of the workflow definition.