<p>Task-oriented chatbots are widely used as conversational interfaces for business services, allowing users to make requests in natural language while collecting the necessary data to invoke relevant web services. However, the dynamic and multidomain nature of web services requires chatbot interactions to be highly adaptable. Current chatbots, whether rule-based or trained on specific tasks, lack flexibility to handle evolving business needs. For instance, a chatbot designed for restaurant reservations may efficiently gather standard details like date and time but struggle with specific or novel preferences such as terrace seating. To address these limitations, we propose an API-Driven chatbot generation approach that manages multiple conversation domains and varying user intentions. Based on service specifications, it conducts a slot-filling dialogue to gather the required information, identify, rank, and invoke the appropriate services, integrating pre-trained large language models at several steps. This approach enables chatbots to incorporate business-specific elements, enhancing adaptability. Additionally, it facilitates seamless integration of new services at runtime, supporting hot swapping and ensuring system extensibility. This capability makes it well-suited for dynamic environments where services continuously evolve. We developed a proof-of-concept multidomain chatbot to evaluate our proposal, obtaining highly positive results in accuracy and efficiency. Thus, our approach provides a robust alternative for more flexible, automated, and domain-adaptive chatbot development.</p>