Triceratops is an example of a very large mediportal animal: an animal that is large, but adapted for relatively rapid locomotion. This can be seen by the structure of its limbs, which unlike an elephant or a more contemporary sauropod, are not columnar, but semi-flexed, a design that allows rapid acceleration. This would fit with its horns: when defending against a predator, it would need to be able to turn quickly and lunge quickly in order to stab at its foe with its horns. The robustness of the leg bones even in comparison to an elephant or sauropod suggests that this is likely. A side effect of being able to accelerate quickly is being able to also run quickly.
Now, we must consider that Triceratops is estimated to have a body mass of 6 to 12 metric tons.
If we consider modern horses, it can be seen from their skeletons and lifestyle that they are a large cursorial species, adapted to running. However, being cursorial does not necessarily equate to maximal load-bearing or -pulling capability: horses' legs are slender at the bottom and relatively far up the limb, and then bulk out near the body, which are classic adaptations to minimising the mass of the limb in order to maximise the rapidity of the gait cycle.
However, Triceratops has far more bulky, robust limbs that are proportionally shorter than a horse's. Despite its longer legs in absolute terms, it is unlikely that it would have been anywhere near as fast as a horse. As I have pointed out, Triceratops seems evolved for making rapid turns and lunges, probably so that it can use its horns as weapons.
So, if a 1 metric ton draft horse can pull a 2 metric ton cart, scaling up the creature to a 6 to 12 ton Triceratops would naively give a hauling capacity of 12 to 24 tons... but this does not take the creature's musculature into consideration.
If we compare a horse with a Triceratops, the comparison is similar to that between a marathon runner and a weight lifter. Even if they weighed the same, the weight lifter would run less quickly and for a lesser duration, but could pull a much greater weight.
A human Marathon runner might be able to lift his own body weight, or perhaps a bit more. However, his training has been more toward lightening his body and increasing his endurance. This is in contrast to a weightlifter, who might be able to lift 2.3 times his own body mass. The comparison between a horse and a Triceratops is that extreme, perhaps even more so.
So, with Triceratops being one of the weight lifters of the animal kingdom, we can expect that they could pull a much greater load than an equivalent-sized horse. If we go with a not unreasonable figure of 2.5 as the greater weight-pulling capacity of a Triceratops as compared to a horse, then we must conclude that a Triceratops's pulling capacity would be on the order of 5 times its body mass... so for a 6 to 12 metric ton Triceratops, it might be able to pull a 30 to 60 ton wagon, provided that the friction of the horse-sized and Triceratops-sized wagons was equivalent.
I mentioned elephants in passing. Elephants have relatively slender limbs for such a large animal, and they are also relatively slow, despite their relatively long legs. A rhinoceros of the same mass has shorter, stronger limbs, and is also faster, as it can gallop, bringing all of its feet off the ground for part of its gait cycle, while an elephant cannot gallop at all.
So, since a Triceratops is equivalent to a physiologically even stronger rhinoceros rather than an elephant, the pulling capabilities of an elephant are a poor example for comparison.
Now, having said that a Triceratops can likely pull vast masses, I would not expect that it could pull them particularly quickly, but it is likely that they could pull a load for a considerable amount of time. Lunging with its horns at a predator would require anaerobic fitness, rather than aerobic fitness. However, Triceratops has a deep, broad rib cage which would be able to house a large heart and a large set of lungs, so it is likely that it could keep up a fairly decent walking pace for a considerable amount of time.
Triceratops dentition consists of around 800 teeth arranged to slice up very fibrous food. This would mean that their preferred diet may have been something like fern fronds or even tree-fern trunks. They would need to spend a considerable amount of time eating in order to meet their energy needs.
The fact that Triceratops has a vast battery of teeth implies that its digestive system is more reminiscent of a horse than a cow, in that it must chew its food thoroughly before swallowing it, and that it does not have the opportunity to re-chew its food as cows do. Cows are forgut-fermenting herbivores, while horses are hindgut-fermenting herbivores.
The significance of this is that a Triceratops could likely eat rich, high-energy food (as can a horse), while a cow cannot, lest it suffer from a digestive imbalance that can be lethal. This would tend to make domestication of Triceratops for purposes of haulage feasible in that eating rich food gives it the ability to gain the nutrients it needs in less time than eating its usual poor-quality food, therefore increasing the time that it can spend working.
On the matter of carts, ball- or roller- bearing axles would reduce the friction of the wood-on-wood bearing surfaces of pre-modern carts, as would having carts with flanged steel wheels running on metal rails, as both these measures decrease rolling resistance. Rubber tyres would have the effect of providing a smoother ride, but at the cost of increasing rolling resistance.
However, as typically happened in pre-modern times when the only power was muscle power, if more power was needed, more animals were added. However, with a single Triceratops likely being able to pull a 30 to 60 metric ton cart, it would be difficult to build a cart that could carry that much, let alone twice that much.
In fact, the extreme strength of a Triceratops would suggest that the more likely arrangement of animal and load would be a train of carts pulled by a single animal, since the individual carts may need to be able to be moved by lesser beings on occasion, as well as the cost involved with producing a single 20 to 40 ton capable cart as compared to the cost of 10 to 20 two-ton capable carts.