The mental health and well-being of children and adolescents is a growing concern in our country, due to digital overexposure, the impact of social media and the consequences of the COVID-19 pandemic, which has led to a steady increase in organisations such as the ANAR Foundation. Due to the growing demand and saturation of help channels, Support-B Chat is proposed as an innovative solution to optimise the efficiency of these services. The suggested system automates the first interactions with the user, filters the incorrect use of the channel and prioritises the cases that require immediate attention. To achieve a fluent and safe interaction, a modular architecture has been implemented based on technologies such as Streamlit, to provide a clear and accessible environment for minors, and Langchain and Ollama to execute the language model (LlaMA 3.2) in a local and controlled way. The chatbot conducts empathic and guided conversations through a flow of states and validations, allowing the collection of essential information from the user in a structured and personalized approach. All conversations are stored in MongoDB in real time and classified according to their urgency and appropriateness, to detect improper uses on the platform and to ensure professional follow-up. Furthermore, in order to evaluate the behavior of the model used, its performance has been compared with other current language models in the classification of simulated conversations. This has allowed us to verify its suitability for the purpose of the system and to establish technical criteria to guide future improvements and integrations. Support-B Chat has demonstrated that Artificial Intelligence technologies, designed under ethical and control criteria, can function as an additional support in childcare, optimizing the efficiency of the service without replacing the professional. The system not only provides a practical and scalable solution, but also serves as a basis for future applications of AI-based systems.e