Let a class of objects be enriched in the category of abelian groups, i.e. for each ordered pair of objects, an abelian group is done, with an operation $+$, identities and compositions are morphisms coming from the monoidal structure on $\operatorname{Ab}$ given by tensor product of $\mathbb{Z}$-modules and satisfy diagrams for associativity and unity.
My question is: is it always true in such a context that $(f+g)\circ h=f\circ h+g\circ h$ and $k\circ (f+g)=k\circ f +k\circ g$? In other words, monoidal structure is enough to ensure bilinearity of composition or I need to assume it?
Given $f: A \to B$, $g: A \to B$ and $h:A' \to A$, the expression $(f+g)\circ h$ is a notation for $c_{A,B,C}((f+g)\otimes h)$ where $$ c_{A,B,C} : \hom(A,B) \otimes \hom(A',A) \to \hom(A',B) $$ is the morphism of composition. As you said, it is a morphism of abelian groups, and in the category of abelian groups morphisms $X\otimes Y \to Z$ corresponds bi-univocally to bilinear maps $X\times Y\to Z$. More precisely there is a bilinear map $\eta : X\times Y \to X\otimes Y$ such that for every bilinear $f: X\times Y \to Z$ there exists a unique $\hat f: X \otimes Y \to Z$ such that $\hat f \eta = f$. In particular, you can see that $$\hat f ((x+x')\otimes y) = \hat f(x\otimes y) + \hat f(x'\otimes y)$$
Coming back to your problem, we have that $$ c_{A,B,C} ((f+g)\otimes h) = c_{A,B,C} (f\otimes h) + c_{A,B,C} (g\otimes h) $$ But the latter is also denoted $f \circ h + g \circ h$.
The same goes for the other linearity, so in the end bilinearity of composition is part of the definition of an $\mathbf{Ab}$-enriched category.