A cloud-management system intercepts API calls from a client application. Based on the expected amount of cloud resources required to service the API call, and given the terms of the client's billing plan, the system estimates the cost to service the call. The system embeds this cost into its response to the API call and the client application, or an IDE hosting the application, keeps a running total of API-related costs incurred by the application. The system can also run in an emulation mode, estimating the application's API resource-consumption costs without actually consuming cloud resources. In this mode, the system replaces each GET and PUT call in the intercepted API call with an equivalent HEAD call that does not consume resources. The system still, however, predicts the actual cost of the original API call by returning the cost of each original GET and PUT.