Hi there! I’d like to suggest a visual editor for defining structured outputs (similar to the Structured Output (JSON Schema) editor in OpenAI’s Agent Builder)
The idea
Create a new ‘Structed Ouptut Parser’ sub-node for AI Chains and AI Actors, that allows to define a structured output schema using a visual JSON Schema builder inside n8n.
This would let creators visually design the format of an agent’s response (strings, numbers, booleans, enums, objects, arrays, nested objects, etc.), and the system would:
-
Generate a JSON Schema automatically
-
Use it to validate the node’s output at runtime
-
Optionally expose the structured fields to downstream nodes for mapping
Essentially, this would bring n8n one step closer to typed agents and validated LLM chains, improving reliability and interoperability.
Why this matters
Right now, structured outputs can only be done manually (via JSON templates, or JSON Schema).
Having a native, visual schema editor would:
-
Make agent and AI workflows far easier to control
-
Improve validation and debugging (catch malformed outputs early)
-
Enable smarter node integrations (auto-mapping of output fields)
-
Create a consistent structure for AI/LLM outputs, similar to how inputs are handled
Inspiration
Here’s an example from OpenAI’s Agent Builder, which shows the kind of interface I’m referring to, a visual schema editor that generates JSON Schema definitions automatically:
