A Dirac spinor is an object which transforms in the $(\frac{1}{2},0)\oplus(0,\frac{1}{2})$ representation of the Lorentz group (if you are not familiar with this notation, have a look at Chapter 6 of Weinberg's QFT I book).
The product of two Dirac spinor decomposes as: $$[(\frac{1}{2},0)\oplus(0,\frac{1}{2})]\otimes [(\frac{1}{2},0)\oplus(0,\frac{1}{2})]= [2\times (0,0)]\oplus [2\times (\frac{1}{2},\frac{1}{2})]\oplus [(1,0)\oplus (0,1)].$$
The first two objects in square brackets are two scalars (or a scalar and a pseudoscalar, if you are considering the full Lorentz group).
The second two objects are two vectors (or a vector and a pseudovector).
Finally, the third objects are a self-dual and an anti self-dual tensor (or an antisymmetric tensor, again under the full Lorentz group).
The $\Gamma$ matrices are nothing but the Clebsch-Gordan coefficient of the above tensor product decomposition, written in conventional bases for the integer spin representations (i.e. bases obtained by taking tensor products of unit orthogonal vectors in $\mathbb R ^{1,3}$).
So the answer to the question "why do the $\gamma$ matrices behave like vectors?" is: "because they are constructed to do exactly so". Notice that you could also work the other way round: you could start from an ordinary vector $V^{\mu}$ and construct a matrix $V^{\mu} \gamma _\mu$. Such a matrix would furnish a perfectly equivalent representation of the same mathematical object.