Mindcare: A Multi-Agent AI Architecture for Personalized and Responsible Mental Health Support

Mental health challenges such as stress, anx- iety, depression, and loneliness are increasingly prevalent worldwide, exacerbated by stigma and limited access to professional care—especially in low-resource settings. This paper introduces MindCare, an AI-powered mental health companion chatbot designed to support individuals expe- riencing mild psychological distress. Leveraging fine-tuned large language models (LLMs), MindCare employs a mod- ular architecture with dedicated agents for intent recogni- tion, compliance checking, psychiatric response generation, memory management, correction, and empathetic interac- tion. The system is trained on curated, anonymized datasets from mental health forums and authoritative sources to ensure contextual accuracy and emotional sensitivity. Comparative evaluations show that MindCare outperforms generic conversational agents in emotional responsiveness, relevance, and user satisfaction. This study outlines the system design, data strategy, and evaluation methodology, positioning MindCare as a scalable and ethical digital tool for mental health support. Future work will focus on adding multilingual, voice-based interaction and real-time escalation to human counselors for enhanced accessibility and personalization.