A Survey on AI-Agent Architectures for Smart Agriculture: Crop Recommendation and Market Forecasting

The integration of Artificial Intelligence (AI) agents into smart agriculture promises transformative improvements in productivity, risk management, and market access. This survey systematically reviews AI-agent architectures applied to agriculture, focusing on crop recommendation, weather forecasting, and market price prediction. Following the PRISMA 2020 framework, we examine 65 studies published between 2018 and 2025, emphasizing reproducibility, critical evaluation, and practical implementation considerations. We discuss regional applicability to Karnataka, India, highlighting datasets, policy linkages, and linguistic diversity. The survey identifies key research gaps, including explainability, multilingual support, heterogeneous data integration, and agent orchestration. Furthermore, we propose the AgroSage architecture, a modular multi-agent system integrating CropAgent, WeatherAgent, and MarketAgent, and outline a validation plan encompassing simulation and field deployment. The survey concludes with a roadmap for advancing AI-based smart agriculture, emphasizing scalable, context-aware, and farmer-centric solutions.