Let's start by considering a spy-bot in the form of a quadcopter drone. Its job is to move a camera and transmitter around. Right now, you can do this with a reasonable amount of autonomous intelligence in a package about the size and weight of cell phone. 200 grams, including the batteries, motors, and electronics.
Now consider a gun. Not even a huge gun. Let's consider a .22 caliber pistol. Just the pistol and a firing mechanism weighs in at almost half a kilo. The electronics won't add much to that, but you'll need that much again for motors and batteries. Let's call this a 1kg package that gets you half a dozen shots. This disparity only gets worse with higher tech levels.
Consider loiter time. The ability to hang out and wait for something to happen is very important in the drone business. If you're keeping a gun aloft, then your loiter time drops significantly. You can have both with a much larger vehicle, but that only works if your drones have a huge field of view, as with Raptors. If they're anywhere near the ground, you lose that.
Consider numbers. Don't ask yourself "is a spybot with a gun better than a spybot." Ask yourself "is a spybot with a gun better than a dozen spybots."
Now that you're thinking in terms of numbers, think of how much effort it would take your opponent to take down a dozen spybots that are smaller and harder to see. You could take out a weaponized quadcopter with a shotgun. You could fire a shotgun into a cloud of spybots and you'd be lucky to get half of them.
All of this changes when you shift to something with legs. Comparatively speaking, legs add a lot more weight, but also increase carrying capacity. At that point, you'd have to make something a hopper like a frog or a flea for it to be effective as a spybot. On the other hand, you could make a walking/hopping gun networked with a half dozen spybot sub-drones and have a versatile and cheap combat unit.
So, to reiterate, don't think this or that, think this and that, in the proper proportions.