Inputs & outputs
Define what a workflow takes in with the Input node and assemble what it returns with the Output node.
Every workflow has exactly one Input node and one Output node. Together they define the workflow's contract: what you give it, and what you get back.
Input
The Input node sets what the workflow accepts. It has two modes:
- Chat — the workflow takes a single piece of free-form text. Good for conversational use and quick triggers.
- Object — the workflow takes a structured object whose fields you define. Good when the workflow is called with specific values by a schedule, the API, or another workflow.
In object mode, open the side panel to define the fields — each with a name, a type, and an optional description. That definition becomes the workflow's input contract: callers provide those fields, and downstream nodes read them.
The Input node has no settings to run; it simply shapes the data entering the workflow. Everything else on the canvas reads its values through bindings.
Output
The Output node is where the workflow's result is assembled. It's not a schema editor — it's an assembly surface: you wire the steps whose values you want into it, and it builds the final result from them. It has two modes:
- Single — the workflow returns one value (a string, an object, whatever the wired step produced).
- Structured — the workflow returns an object with several named fields, each filled from a step you wire in.
Assembling the result
When you wire a node into the Output, Compass automatically adds a field named after that source — so the common case needs no extra work. To refine it, open the Edit output panel, where you can:
- rename, add, remove, and reorder fields,
- set what each field is bound to,
- and preview the shape of the result.
If nothing is wired in yet, the Output node prompts you to connect a step or open the panel to assemble a result by hand.