Traceability
Following an agent's decision chain from input to action to outcome.
Why it matters
When something goes wrong, trace the chain: what triggered the decision? What context? Why this action?
In practice
Every workflow is traceable: n8n logs each step, LLM calls log model/tokens/cost, Paperclip logs governance decisions.