Execution settings
The Execution tab holds the settings that run the step itself (models, prompts, code, documents, HTTP calls, and numeric expressions), separate from the Output tab where you shape what the next step receives.
Which steps have an Execution tab?
Only steps that perform work needing a dedicated runtime panel:
- AI: model choice, instructions (or subtype-specific bodies such as content to summarise or transform), and fields that feed the model pipeline.
- Run code: timeout, result type coercion, and the JavaScript source (tags resolved before the sandbox runs). See the Code guide.
- Random number: bounds and expressions for the sampled range, driven by tags where needed.
- Iteration: how the counter increments (expressions and inbound data).
- Document: template or DocXML execution: model instructions, template bindings, and related options for generating files.
- Webhook call: method, URL, headers, and body templates resolved against the same tag context as other steps.
Triggers use the Input tab instead (invoke schema, schedule, or webhook configuration). Logic steps (decision, switch, split) use Gate or Branch tabs. Human and termination nodes expose Output (and General) rather than Execution in the same sense.
Execution versus Output
Executionanswers "what runs right now?" Call the model with these instructions, run this script inside the timeout, POST this JSON payload.
Output answers "what object do downstream steps see as {{input.*}}?" Map keys from exe and literals into a clean JSON shape, declare globals, or define branch handles.
Tags in execution fields
Execution fields resolve with the base tag context: input, trigger_inputs, global, const, now, and metadata namespaces. See Expression variables for detail.