Autistic children’s emotional meltdown is a common challenge faced by their caregivers, including their teachers and families. Communicating their problems can be extremely challenging for many autistic children, making it difficult for caregivers to identify the triggers of these meltdowns. Previous studies have revealed that autistic children find it easier to communicate with non-human objects, such as comfort dolls or social robots. In light of this, we have designed Kodi, a mobile virtual conversational agent aimed at bridging the communication gap between autistic children and their caregivers during distressing situations. Kodi’s appearance can be personalized to match the child’s favorite characters, such as a bunny or a dinosaur. To provide additional comfort, Kodi comes with furry phone cases that offer haptic stimulation. When the child faces stressful situations, caregivers can prompt Kodi on children’s mobile device, allowing the child to express their emotions with various input modalities (emojis, text, voice) and listen to soothing stories. Caregivers can tailor these stories to suit each child’s unique needs. Furthermore, Kodi will employ keyword analysis to summarize conversations, identify potential triggers for emotional meltdowns (e.g., loud noises in a shopping mall, unexpected changes at school), and alert caregivers of potential problems, such as bullying. The data collected through Kodi holds the potential to be leveraged for predicting emotional meltdowns using machine learning techniques. This paper discusses the implications of the Kodi platform and outlines future research directions in this area.