Climate-driven salinity intrusion poses significant threats to coastal agriculture and sustainable food systems in developing nations. By examining farmer preferences for climate adaptation policies, this study aims to inform the design of sustainable policies that align with farmers' risk perceptions and decision-making. Using Best–Worst Scaling (BWS) methodology, we analyzed preferences across 20 climate adaptation policy options among 200 coastal rice farmers in Bangladesh. Data were collected through face-to-face interviews across four coastal districts. A conditional logit model examined how risk attitudes and socio-economic factors influence adaptation decisions across technology-based adaptation, production diversification, water management, and economic support strategies. Results show that salinity-tolerant rice varieties ranked highest (standardized BW score: 0.215), followed by high-yielding varieties (0.125), pest-resistant varieties (0.098), and farm mechanization (0.087), indicating that farmers strongly favor technology-based climate adaptation strategies. Risk attitude significantly influences policy preferences: extremely risk-averse farmers avoid sophisticated innovations like integrated floating agriculture but favor diversification techniques like double cropping, while moderately risk-averse farmers strongly favor proven adaptation technologies. Fair output pricing emerged as a key prerequisite for adoption. Key policy implications include: (1) prioritizing salt-tolerant rice varieties as the primary adaptation strategy, (2) using proven technologies as stepping stones for broader adoption, and (3) designing risk-sensitive policy tailored to different farmer groups. This study demonstrates how BWS methodology can systematically reveal farmer preferences to inform adaptation policy design. The results provide practical guidance for developing nations facing similar coastal climate challenges, emphasizing that successful adaptation requires aligning context-specific policies with farmer preferences rather than top-down solutions.