The clockwork constructs in Mordenkainen’s Tome of Foes are a collection of machines used by rock gnomes to defend their turf. Combining trickery with extraordinary durability and disproportionate damage-dealing capacity, they share a range of condition and damage immunities along with darkvision and the ability to understand their controllers’ commands—but also a rigidity in their behavior that can only be compensated for by active, real-time control. If they’re sent off to do their work on their own, they do it mechanistically, with no adaptation to what’s going on around them.
First up is the bronze scout, which isn’t particularly strong, but it doesn’t need to be, because it’s basically a self-guided mobile land mine. The key things to note are its burrowing movement, its double proficiency in Stealth, its Earth Armor trait, its Lightning Flare action, and one more trait that’s mentioned in the flavor text but unpardonably omitted from its stat block: “telescoping eyestalks” that let it see aboveground while it burrows below. These eyestalks are crucial, because the bronze scout lacks tremorsense or any other listed way to detect the presence of creatures above it.
This combination of features makes the bronze scout the ideal ambush initiator: Using Stealth to muffle its approach, it scuttles along the ground until it sees movement, then tunnels into the earth and heads toward it. Once it’s approximately in position, it pokes its eyestalks up and looks around, checking to see if its position is correct—that is, if at least three enemies are within 15 feet of it (see “Targets in Area of Effect,” Dungeon Master’s Guide, chapter 8). If it’s not, it retracts its eyestalks and repositions. If it is, it sets off its Lightning Flare, whereupon its waiting allies launch their attack. Since it’s immune to physical damage from nonmagical, non-adamantine weapons, it can take a hell of a beating, Biting back at whatever attacks it. But if it’s seriously damaged (reduced to 7 hp or fewer), it dives back underground, provoking no opportunity attack thanks to Earth Armor, and scuttles away.
The bronze scout doesn’t have to be used in this way, though. It can be used, as its name suggests, simply as a scout, which doesn’t attack at all unless it’s discovered. In this instance, the bronze scout Readies the Lightning Flare action, with the triggering condition “when any creature winds up to make a melee attack against it.” Including the wind-up in the trigger condition is key, because it’s a perceivable circumstance allows the bronze scout to use Lightning Flare as an interrupt, occurring before the opponent follows through with the attack, whereas if the condition were “when any creature makes a melee attack,” the reaction would have to wait until after the attack either hit or missed.
A bronze scout that’s merely scouting doesn’t stick around and Bite unless an enemy manages to grapple it and thereby keep it from burrowing away, and it only Bites whoever has grappled it. As soon as the grappler lets go, it burrows away.
Let’s talk about those eyestalks for a minute, because poking them up out of the ground within the line of sight of other creatures could arguably expose a hidden bronze scout to them. This surely calls for an additional Dexterity (Stealth) check, above and beyond the one that initially determined whether the burrowing bronze scout was hidden. However, since its eyestalks are considerably smaller than the bronze scout in toto, I’m inclined to give the bronze scout advantage on this check. In fact, I think I’d have to give the bronze scout advantage on Stealth checks as it burrows underground as well, because the earth would muffle its sounds. Only aboveground would the bronze scout need to make straight Stealth checks.
The bronze scout has darkvision, optimizing it for underground or after-dark reconnaissance, either one preferable to snooping around in daylight.
The iron cobra is another mobile mine, or perhaps more properly a mobile grenade, since it doesn’t burrow. It does, however, have a high Stealth modifier, allowing it to Hide where a target might pass by it or even to approach unnoticed if visibility is poor or the target is distracted. (An alert target with clear lines of sight and nothing else to occupy its attention will see it coming—it won’t even get to make a Stealth check under such circumstances.) It, too, has darkvision, making it preferable to use it at night or underground rather than aboveground during the daytime.
With a respectable reservoir of hit points and the same immunity to physical damage from normal, non-adamantine weapons as the bronze scout, the iron cobra can keep Biting its designated target for as long as it needs to. As a construct with Intelligence of only 3, it has no capacity for independent judgment and has to follow its controller’s instructions to the letter, whether these instructions are “Bite once, then retreat,” “Bite three times, then retreat,” or “Bite everybody in the whole world!” If its instructions don’t include retreating after a certain amount of time, it doesn’t—it keeps Biting until it’s damaged beyond its ability to function.
I don’t like the random-effect aspect of the Bite, and I personally think the type of effect the Bite’s poison causes should be part of the instructions given to it, e.g., “Bite once to paralyze, then make subsequent bites to cause confusion,” or, “Bite only to cause poison damage.” Any such instructions given, however, have to be phrased in terms of the number of times the iron cobra Bites or whether the current intended target is already paralyzed or confused by a previous Bite. The iron cobra has no capacity for target assessment, either; it can’t distinguish, for instance, whether a target seems low-Constitution or high-Constitution, nor can it make decisions based on what a target is doing, unless what it’s doing is attacking the iron cobra. In any event, since its Intelligence is 3, that seems to me like a nice cap on the number of conditions it can evaluate while executing its instructions.
The controller, on the other hand, presumably knows how to get good use out of an iron cobra. Paralysis is a good effect to use on a target against whom the controller intends to follow up with an attack or an effect that requires a Dexterity save to resist. Confusion is good against supporters, who are likely to be positioned within easy reach of their allies and whose combat role can thereby be subverted. Poison damage is all-purpose.
The stone defender is a front-line bulwark that interposes itself between its allies and their enemies. It always positions itself immediately adjacent to or in front of an ally, which it defends with its Intercept Attack reaction. Lacking the capacity for independent judgment, it always Intercepts the first attack that would otherwise hit its ally; it has no way of deciding that one attack should be let through in order to prevent a potentially more dangerous attack. (That’s assuming that the stone defender is a wind-it-up-and-let-it-go device, however. If the creature it’s protecting is actively controlling it, you may decide that the controller can use its own judgment to call out, “Protect me, squire!” when it wants the stone defender to Intercept an attack and stay mum when it doesn’t.) On its own turn, it Slams any enemy within reach that has attacked it or one of its allies, or that its ally is also attacking.
There’s nothing more to the stone defender than this. However, allies of the stone defender will surely take advantage of its Slam attack by moving in and aiming melee attacks at enemies knocked prone by it. Stone defenders don’t retreat, no matter how much damage they take.
The oaken bolter is an everything-but-the-kitchen-sink artillery piece that can operate anywhere on the battlefield, including, counterintuitively, the front line. That’s because its Lancing Bolt attack is both a ranged attack (with a very long range, equal to that of a heavy crossbow) and a melee attack, so it suffers no penalty at point-blank range. That doesn’t mean it belongs on the front line, just that it can operate there if things get weird. Its ideal position, however, is between 40 feet from the nearest enemy and 100 feet from the most distant one.
An elegant exploitation of Lancing Bolt’s dual nature is found in the Multiattack that includes one Lancing Bolt attack and one Harpoon attack. Harpoon, on a hit, offers the option of pulling the target 20 feet closer using a bonus action. If the oaken bolter uses this attack first and succeeds in grappling the target, it can use its bonus action mid-turn to pull the target closer, then launch a Lancing Bolt at it regardless of distance. I can’t think of any circumstance in which a different order of operations would be better. Note, though, that Harpoon only grapples; it doesn’t restrain, so the follow-up Lancing Bolt attack roll doesn’t gain advantage from a successful harpooning.
An excellent application of Harpoon is to pull an enemy out of position—for instance, to drag a spellslinger or marksman out of the backline. Another is to keep a shock attacker from slipping away. Another is to yank a melee attacker off an ally that’s looking the worse for wear. Another is to pull an enemy past a hidden iron cobra.
Explosive Bolt is best used when four or more opponents are clustered within the area of effect (which is large—40 feet in diameter, same as a fireball), and since it’s on a 5–6 recharge, the oaken bolter takes this action whenever that criterion is met and the ability isn’t on cooldown.
Oaken bolters roll around constantly within their optimal range from the enemy; if they end up outside this range (by some means other than harpooning an opponent and dragging them nearer), they use their movement to try to return to it, potentially incurring one or more opportunity attacks. Because they share the same damage immunities that all these clockwork constructs possess, and because they too are constructs without independent judgment, they don’t concern themselves with this fact. They don’t retreat when seriously damaged.
The enhancement and malfunction tables under “Individual Designs” don’t alter these constructs’ tactics in any way, but they do make the execution of those tactics more entertaining, and I see no reason to forgo using them. In fact, I’d go so far as to suggest having a clockwork construct suffer a random malfunction whenever it’s reduced to 40 percent or less of its hit point maximum!