Expose a CognOS agent system as a JSON graph for full inspection.
Enables semantic prompt compression for LLM workflows, reducing token usage by 30-70% while preserving semantic meaning.
