The Belief Engine (BE) introduces a new layer for LLM-based agents that enhances transparency in how these agents update their stances during deliberative interactions by tracking evidence and anchoring influences. This innovation could improve the reliability of AI in negotiation and opinion exchange scenarios by making their decision-making processes more auditable and understandable.