While there is likely a more conceptual argument, there is a fairly direct argument using an explicit construction of this cocompletion. We can realize this extension of $\mathcal{C}$ as a full subcategory of $\mathrm{Pr}(\mathcal{C})$, namely the smallest full subcategory of $\mathrm{Pr}(\mathcal{C})$ closed under $\kappa$-small colimits and containing the image of $y : \mathcal{C} \to \mathrm{Pr}(\mathcal{C})$.
It suffices to show that this subcategory is essentially small, but it can be realized as a union of a chain of $\kappa^+$-many essentially small full subcategories of $\mathrm{Pr}(\mathcal{C})$. Explicitly, we start with the image of $y$ (essentially small as it is $\mathcal{C})$, then take the full subcategory of objects built out of $\kappa$-small colimits of representables (small because there is only a set's worth of $\kappa$-small colimits in any small category) and then proceeding inductively.