We study the problems related to querying large, distributed XML documents. Our proposal takes the form of a new process calculus in which XML data are processes that can be queried by means of concurrent pattern-matching expressions. What we achieve is a functional, strongly-typed programming model based on three main ingredients: an asynchronous process calculus in the style of Milner’s pi-calculus and existing semantics for concurrent-ML; a model where documents and expressions are both represented as processes, and where evaluation is represented as a parallel composition of the two; a static type system based on regular expression types.