Technologies are described herein for invoking API calls through voice commands. An annotated API description is received at a voice API interface. The annotated API description comprises descriptions of one or more APIs and speech annotations for the one or more APIs. The voice API interface further receives a voice API command from a client. By utilizing the annotated API description and the speech annotations contained therein, the voice API interface converts the voice API command into an API call request, which is then sent to the corresponding service for execution. Once the service returns an API call result, the voice API interface interprets the API call result and further converts it into an audio API response based on the information contained in the annotated API description and the speech annotations. The audio API response is then sent to the client.