In this paper, we investigate the problem of assembling fragments from different graphs to build an answer to a user query. The goal is to be able to provide an answer, by aggregation, when a single graph cannot satisfy all the query constraints. We provide the underlying basic algorithms and a relational framework to support aggregated search in graph databases. Our objective is to provide a flexible framework for the integration of data whose structure is graph-based (e.g., RDF). The idea is that the user has not to specify a join operation between fragments. The way the fragments can be combined is a discovery process and rests on a specific algorithm. We also led some experiments on synthetic datasets to demonstrate the effectiveness of this approach.