OpenBox: Agent
AI-generatedSummary
Govern the lifecycle of an AI agent interaction by submitting user prompts to an OpenBox-governed language model with optional tool integrations and governance controls.
Inputs
- promptType (required) — Selects the source of the prompt text, either from a connected chat trigger node ('auto') or a user-defined expression or static text ('define').
- text (required) — The user message prompt text for the agent; either taken from connected input data or defined statically depending on promptType.
- needsFallback — Enable use of an additional fallback language model if the main model call fails.
- options — Collection of optional settings such as automatic passing of binary images, enabling streaming output mode, maximum agent iterations, error handling strategy on tool failure, inclusion of intermediate reasoning steps in output, and a system message to initialize the agent's context.
- governance — Advanced governance options controlling human-in-the-loop approval behavior, instrumentation of database and HTTP calls, error handling policies, and tools exempted from governance.
Output shape
a single JSON object containing the final agent response text, plus optionally intermediate steps, and additional metadata.
If 'returnIntermediateSteps' is enabled, the output includes detailed reasoning and tool usage steps taken by the agent. Streaming option outputs chunks progressively if enabled and supported by the connected trigger. Governance errors or tool failures may modify the flow according to the selected options. The response follows the shape and conventions of n8n's official Agent node for compatibility.