Extends the Diplomacy board game with LLM-powered agents. Each power in the game is represented by a DiplomacyAgent that maintains state, forms relationships, conducts multi-round negotiations, and generates strategic orders. The agents use a dual-layer memory system with structured diary entries and automatic consolidation to preserve context across long games. Strategic decision-making combines BFS pathfinding with context-aware order selection, and agents can detect betrayal through negotiation and order analysis. The framework supports multiple LLM providers for generating agent behaviors.