OptionalEffortThe target's reasoning effort, for models that support one.
OptionalKindWhat the reference denotes: 'model' (Live, Inworld) or 'hostedAgent' (ElevenLabs).
OptionalMaxThe most output tokens for the remote reasoning pass.
OptionalRefThe target, e.g. 'gpt-5.6-terra', 'anthropic/claude-sonnet-4-6', 'MJ Realtime Co-Agent'.
The remote reasoning target, used when
Planeis'remote'.