We offer an alternative approach to the existing methods for intuitionistic formalization of reasoning about probability. In terms of Kripke models, each possible world is equipped with a structure of the form \(\langle H, \mu \rangle \) that needs not be a probability space. More precisely, though _H_ needs not be a Boolean algebra, the corresponding monotone function (we call it measure) \(\mu : H \longrightarrow [0,1]_{\mathbb {Q}}\) satisfies the following condition: if \(\alpha \), \(\beta \), \(\alpha \wedge \beta \), (...) \(\alpha \vee \beta \in H\), then \(\mu (\alpha \vee \beta ) = \mu (\alpha ) + \mu (\beta ) - \mu (\alpha \wedge \beta )\). Since the range of \(\mu \) is the set \([0,1]_{\mathbb {Q}}\) of rational numbers from the real unit interval, our logic is not compact. In order to obtain a strong complete axiomatization, we introduce an infinitary inference rule with a countable set of premises. The main technical results are the proofs of strong completeness and decidability. (shrink)