Legal contract review automation faces significant challenges in processing unstructured texts and performing complex legal reasoning, particularly in Chinese legal contexts where manual methods remain inefficient and error-prone. We present PEAT-LLM4LCR, a specialized automated contract review tool that addresses these challenges through the integration of prompt engineering and multi-agent collaboration technology. Unlike existing solutions such as CUAD, which focus on English contracts, or general-purpose LLMs that lack domain expertise, our system employs the Agent-GoTFlow architecture to enable dynamic workflow adaptation and multi-task coordination. The system comprises three core review modules coordinated through 30 chain-of-thought reasoning templates distributed across three specialized agents, processing natural language requirements and generating structured risk reports with actionable recommendations. Experimental evaluation with 18 participants across different expertise levels demonstrates significant improvements compared to manual review without tool assistance: 34–77% reduction in review time across different expertise levels (novices: 50%, professionals: 77%, experts: 34%) and 50% improvement in accuracy for non-expert users. The system democratizes professional-grade legal analysis, benefiting legal professionals, business stakeholders, and resource-constrained small-medium enterprises alike. The tool is available at https://github.com/Hongbin-Xiao/PEAT-LLM4LCR-TOOL , with a demonstration video at https://www.youtube.com/watch?v=eyIikQDkv0E .
Xiao et al. (Thu,) studied this question.