p-value associated with Kruskal Wallis with multiple comparison?

#1
I have a data set that shows the differences between 3 different controllers of a biomechanical system on the same performance metric (specifically, I measure each controller's performance on a set of N tasks, and record the percentage of tasks that achieve success (i.e. the controlled segment reaches its specified target position). Success values range from 0% to 100%.

Because these data sets have failed Levene's test for homogeneity of variances and the Kolmogorov-Smirnov test for normality, instead of using 1-way ANOVA analysis to compare these 3 controllers' performance, I use the nonparametric Kruskal-Wallis ANOVA test with multiple comparison.

Now, I'm creating a boxplot figure displaying the data of the 3 different controllers. I want to add asterisks to indicate the significant differences present, but the software that I'm using requires that I indicate the p-value for each pair of controllers (in order to determine how many asterisks to add for each pair).

The problem: when I perform the Kruskal-Wallis test with multiple comparison, I get a *single* p-value as a result, which simply indicates whether there are *any* significant differences between controllers, but doesn't provide info about which controllers differ. I use the confidence intervals resulting from the multiple comparison to judge whether or not each pair of controllers differs significantly (if the confidence interval contains 0, there is no significant difference). However, these confidence intervals do *not* give me a p-value for each controller pair comparison, which I need in order to add asterisks to my boxplot figure.

Is there a way to derive a p-value for *each* comparison of 2 controllers, when using a Kruskal-Wallis ANOVA with multiple comparison? Thanks for your help!