Oh, you mean TVS diode. Sure, you can use those, but note their voltage curves are pretty lame, i.e., you need to pick one with nominal voltage = at least maximum battery voltage, and then the clamping voltage will be some 30-50% higher.
Even if they were infinitely quick, the voltage still needs to rise to that clamping voltage for anything to happen. And the capacitance of the bus limits voltage rise.
For catching "battery input suddenly disconnects and the controller regeneratively dumps momentum-stored mechanical energy into the DC bus" case, TVS is great because this would require quite a lot of capacitance. TVS can take quite a lot of heat before it blows up. Of course, proper action is to prevent any regeneration in this case, but this is up to the implementation of the controller.
For catching "battery input suddenly disconnects and the controller does not regenerate still but dumps motor inductance-stored energy into the DC bus" case, electrolytics usually are enough and not too big of a solution, but in some cases, a TVS might be better.
For EMI, it does nothing (except for the effect of its parasitic capacitance). The capacitors do that. If electrolytics are not "fast enough" - they do have maybe 10nH of inductance, and ESR also limits the voltage ripple absorption, add small ceramics (for really high frequencies).
Remember that EMI originates from the energy stored in layout loop inductance resonating with the parasitic capacitance of open MOSFET. It does not require a massive amount of capacitance to absorb that, the problem is minimizing the impedance, which is caused by L and R of the capacitance. Electrolytics are used to dampen the remaining oscillation, here the key is to have enough R so that change in voltage per time causes actual heating loss, removing the energy instead of putting it back to the oscillation. Although too small of electrolytics, even if they had enough C for the job, have too much R again. Hence it's typically some combination such as 10pcs of 1uF ceramics right in the layout, plus 1000uF of low-ESR electrolytic or 2200uF of "traditional" electrolytic for damping. (This is chosen so that Celectrolytic > 2-3 times Cceramic, and then chosen for the correct amount of ESR.)
If you are having EMI problems, adding a bit of differential-mode LC plus common mode filtration could be inevitable, but this sucks if you need to minimize the solution. Tight layout and proper ceramic caps right in that layout is the key for a successful yet small design. A TVS could work but it should be in that tight layout to have any chance of catching the voltage overshoot.
Randomly placed TVS does not hurt, as long as the nominal voltage (not clamping voltage) is above maximum battery voltage, but as it likely doesn't help either, I would start by verifying what problem you have, if any, then try to solve that.
If you can't analyze/modify the controller, adding electrolytics outside does not hurt (except for size/weight) because their ESR is always large enough it's almost impossible to cause oscillation by adding electrolytics. So it costs nothing to try adding a beefy cap and measure what happens. Add some 1uF small ceramic in parallel and see again. If you really see spikes exceeding 30-40% the battery voltage on the oscillosscope screen (probed right from the DC wires, without creating a loop neither with the DC wires, nor with the scope ground clip), and these spikes are more than a few dozen nanoseconds, a TVS can possibly help.