Adolescent mental health is a pressing global issue, with many of adolescents suffering from mental health disorders and a significant shortage of professional counselors in schools. To address this gap, we propose MindMate, an intelligent mental health consultation system for adolescents powered by Retrieval-Augmented Generation (RAG) technology. MindMate uses large language models (LLMs) to generate a high-quality knowledge base tailored to adolescent mental health, which is refined through quality and diversity filtering. The RAG framework enhances the system’s performance by integrating this knowledge base with LLMs, ensuring accurate, reliable, and empathetic responses. Evaluations show significant improvements in accuracy, practicality, risk response, and empathy compared to baseline LLMs. MindMate demonstrates strong potential to provide effective psychological support for adolescents. Future work will focus on dynamic knowledge base updates, handling complex queries, large-scale user studies, and ensuring robust privacy mechanisms.