Protocols enable unambiguous and smooth interactions among agents. Multiagent systems need to host a rich variety of interaction protocols if they are to model the needs of real-world applications such as business processes. Commitments allow flexible execution of protocols, while at the same time providing a basis for compliance checking. We propose conceptual abstractions for commitmentbased protocols similar to those used in information models, specifically (1) refinement: a subprotocol may satisfy the requirements of a superprotocol but support additional properties; and (2) aggregation: a protocol may combine existing protocols. We develop an algebra of protocols that can be used to compose protocols by refining and merging existing ones at a level of abstraction high enough to be useful for real-world applications.