A conversation, whether with a computer or human, starts with a request and response, where the request can take various shapes such as voice phrases or actions on an app, and the response can be simple text or more complex media. The Voiceflow API model mimics this series of requests and responses, using API calls to adapt it into an interface. The test tool uses the same API, allowing for improvements over its functionality. When calling the API, key information includes versionID, API Key, userID, request type (JSON object), and response format (JSON array). The Voiceflow API comes in two flavors: stateful and stateless, with the former being easier to work with but requiring user session data storage, while the latter passes current state back and forth without storing it.