We propose a confidence envelope for false discovery control when testing multiple hypotheses of association simultaneously. The method is valid under arbitrary and unknown dependence between the test statistics and allows for an exploratory approach when choosing suitable rejection regions...