In order to domesticate wild animals, you must necessarily complete a number of criteria (docility, can't be picky eaters, conforms to a social hierarchy, etc), but above all the most important criteria is the reproductive control step which separates taming from domestication. Most domestication candidates, like fish, crustaceans, mollusks, sessile creatures, etc, have tiny planktonic larvae that are mobile, freely swimming in the water column. Even demersal eggs where the parents watch over the eggs are, once hatched, tiny pelagic independent plankton that disperse away. That means that, once hatched, the larvae are extremely hard to see (or sense) and are difficult to catch, and it also means controlling wanted & unwanted traits is also difficult which makes true domestication impossible since they'll always be wild.
Quite aside from the difficulty of caging fish, broadcast spawning is easy as heck at sea (which applies as much to agriculture as pastoralism). How would a sapient aquatic race control the reproduction of a tamed wild animal that has planktonic pelagic larvae?
Assume neolithic era technology for these reef dwelling sapients. They use long strands of seaweed for rope, and corals (instead of stone) for building material and making cutting tools & spears. The rope could be used to haul larger quantities of stuff than one could carry, similar to how we use the wheel to push stuff around. It could also be further weaved into baskets and nets, and other similar items.
They use animal bladders filled with low density oil or gas for balloons to lift heavy objects. Oil is very common in underwater creatures and is incredibly easy to harvest in a primitive way (leave dead fish in container, oil is squeezed out & automatically floats upwards), parallel to paleolithic humans creating alcohol very primitively.