Social-emotional skills, such as empathy and emotion regulation, are critical bricks in the foundation for children’s long-term well-being. However, early childhood educators often lack the time, resources, and (consequently) expertise to deliver high-quality social-emotional training. Even then, existing trainings are delivered with rudimentary hand puppets and one-size-fits-all manuals that may not engage children with varying developmental needs and cultural backgrounds. In this demonstration paper, we introduce PeACE (Pedagogical Agent for Children’s Emotions), an interactive digital tutor powered by a version of OpenAI’s Generative Pre-Trained Transformer (GPT) series that we prompt with evidence-based pedagogical and emotional support frameworks. PeACE is designed to deliver a pre-validated social-emotional training in a typical classroom setting alongside educators, complementing their pedagogical skills, ethical discernment, and trust with children. We describe the components of PeACE: an agent and digital environment that appeal to young children, audio channels for educators and children to converse with the agent in most languages, a user interface for educators to monitor and manage training content, and an integrated back-end system to process spoken language, maintain contextual understanding, and provide real-time responses aligned with prompting frameworks. We briefly conclude with next steps for evaluating our design goals to help more children reach their full social-emotional potential.