Node library

Every workflow is built from nodes. Each node is one step: it takes some input, does its job, and passes the result down the wires to the next node. The editor ships with a set of built-in nodes for the steps you reach for most — and when none of them fit, you add your own.

This page is the map. Skim it to see what’s in the box, then follow a link when you want the details on a specific node. The two groups below match the categories in the palette.

Flow

The nodes that move data through a workflow and decide where it goes.

• Input — defines what the workflow takes in. Each field becomes an output port the rest of the graph wires to.
• Output — collects the final result. Whatever you wire into it becomes the workflow’s output.
• Conditional — branches. Compares a value against a target and routes execution down the True or False path.
• Merge — combines several inputs into one, with modes for how to join them (including zipping lists together).
• Array Operations — transforms a list: filter, map, count, sum, and more.
• Human Review — pauses the run for a person to approve or reject, then routes to the Approved or Rejected path.

AI

The nodes that call a model. Point them at a provider — see Providers & models — and they fit alongside the Flow nodes like any other step.

• LLM — sends a prompt to a language model and returns its response. Give it tools to call, ask for structured output, or feed it images.
• Image Generation — generates an image from a text prompt.

Running a node over a list

The LLM and Image Generation nodes can run once per item over a list instead of a single time — how you’d summarize fifty tickets or generate a dozen images in one step. See Map & arrays.

Where next

• Input & Output — the entry and exit of every workflow; the place to start
• LLM — the node you’ll reach for most
• Custom node types — build a step the built-ins don’t cover