The p = kruskalwallis(x) test works on columns of x. If all of your tables have the same number of rows, you just have to extract the values from each table into a matrix. If the tables only have 1 variable, as you described, it's as easy as,
T1 = table(rand(20,1));
T2 = table(rand(20,1)+1);
T3 = table(rand(20,1)+.5);
p = kruskalwallis([T1{:,:},T2{:,:},T3{:,:}])
% or
p = kruskalwallis([T1.Var1,T2.Var1,T3.Var1])
If the tables do not have equal rows, you must concatenate them into a single column and use a grouping variable p = kruskalwallis(x,group)
x = [T1.Va1;T2.Var1;T3.Var1];
group = [1 + zeros(size(T1.Var1)); 2 + zeros(size(T2.Var1)); 3 + zeros(size(T3.Var1))];
p = kruskalwallis(x,group)
