This is along the line of Phonon's response, but I think this can be formulated more clearly.
From the angular momentum algebra $$[L_i, L_j] = i \hbar \epsilon_{ijk} L_k$$ it follows that (for $\vec L^2 = \sum_i L_i^2, L_{\pm} = L_x \pm i L_y$):
$$[\vec L^2, L_i] = 0, \quad [\vec L^2, L_\pm] = 0, \quad [L_z, L_{\pm}] = \pm \hbar L_\pm$$
Now $[\vec L^2, L_z] = 0$ implies that there is a common base of eigenvectors of $\vec L^2$ and $L_z$. Thus, there are two labels to an angular momentum eigenstate. These are not necessarily the eigenvalues themselves, but from the labels one should be able to determine the eigenvalues:
$$ \vec L^2 \vert \alpha, \beta \rangle = X_{\alpha} \vert \alpha, \beta \rangle, \quad L_z \vert \alpha, \beta \rangle = Y_{\beta} \vert \alpha, \beta \rangle$$
Now, $[L_z, L_\pm] = \pm \hbar L_\pm$ implies that $L_\pm \vert \alpha, \beta \rangle$ is again an eigenstate of $L_z$, but with a different Eigenvalue:
$$ L_z (L_\pm \vert \alpha, \beta \rangle) = L_\pm (L_z \vert \alpha, \beta \rangle) + [L_z, L_\pm] \vert \alpha, \beta \rangle = (Y_{\beta} \pm \hbar) (L_\pm \vert \alpha, \beta \rangle) $$
Thus, acting with $L_\pm$ on $\vert \alpha, \beta\rangle$ increases the $L_z$ eigenvalue $Y_\beta$ by $\hbar$. Therefore, let's try to use $m$ instead of $\beta$ and set $Y_m = m \hbar$ (As you can see, even for the $L_z$ operator, the actual eigenvalue is not equal to the label, but carries a factor of $\hbar$!).
Now for the eigenvalue w.r.t. $\vec L^2$ we need to work a bit more. Check that
$$ L_+ L_- = L_x^2 + L_y^2 + \hbar L_z, \quad L_- L_+ = L_x^2 + L_y^2 - \hbar L_z, $$$$ \Rightarrow \; \vec L^2 = L_z^2 + \frac{1}{2} (L_+ L_- + L_- L_+)$$
Also, the $\vec L^2$ eigenvalue does not change under operation with $L_\pm$ due to $[\vec L^2, L_\pm] = 0$.
Now since $L_- = L_+^\dagger$, we have that
$$ \vert L_+ \vert \alpha, m\rangle \vert^2 = \langle \alpha, m \vert L_- L_+ \vert \alpha, m \rangle \geq 0$$
$$\vert L_- \vert \alpha, m \rangle \vert^2 = \langle \alpha, m \vert L_+ L_- \vert m, \alpha\rangle \geq 0 $$
On the other hand we can use the expression of $\vec L^2$ in terms of $L_z$ and $L_\pm$ to see that
$$ X_\alpha = \langle \alpha, m \vert \vec L^2 \vert \alpha, m \rangle = \langle \alpha, m \vert L_z^2 + \frac{1}{2} ( L_+ L_- + L_- L_+) \vert \alpha, m \rangle = \\ \hbar^2 m^2 + \text{something positive or zero}$$
Thus, $X_\alpha \geq \hbar^2 m^2$. This has profound implications: $X_\alpha$ is always larger than (or equal to) $\hbar^2 m^2$. But working on a given state $\vert \alpha, m \rangle$ with $L_\pm$ increases $m$ while leaving $\alpha$ constant! If we could increase $m$ indefinitely (by operating with $L_+$ multiple times on a state), then $X_\alpha$ had to be infinite, which does not make sense physically. This means that there must be a state with highest $m$ for which $L_+ \vert \alpha, m_\text{max} \rangle = 0$.
The same chain of though produces the lowest state, for which $L_- \vert \alpha, m_\text{min}\rangle = 0$. If you do not see this, read again what I wrote, think about it, read it in books until you understand it. This is a very important result!
We can use this to construct the eigenvalue $X_\alpha$: Since $$L_+ \vert \alpha, m_\text{max} \rangle = 0$$ we have that
$$L_- L_+ \vert \alpha, m_\text{max} \rangle = 0$$ which is nothing but
$$(\vec L^2 - L_z^2 - \hbar L_z) \vert \alpha, m_\text{max} \rangle = X_\alpha - \hbar^2 m_\text{max}^2 - \hbar \hbar m_\text{max} = 0$$
This means that $X_\alpha = \hbar^2 (m_\text{max}^2 + m_\text{max})$. The same way you can show that $X_\alpha = \hbar^2 (m_\text{min}^2- m_\text{min})$. You can now characterize the $X_\alpha$ by the largest allowed $m$:
$$\vec L^2 \vert m_\text{max}, m\rangle = \hbar^2 (m_\text{max}^2 + m_\text{max}) \vert m_\text{max}, m\rangle$$ or you can just call $m_\text{max} = j$ and have
$$ \vec L^2 \vert j, m \rangle = \hbar^2 j (j+1) \vert j, m \rangle, \quad L_z \vert j, m \rangle = \hbar m \vert j, m \rangle $$
Finally, the two expressions for $X_\alpha$ that we derived from the maximal and minimal $m$ can be put together:
$$ X_\alpha = X_\alpha$$
$$ \hbar^2 (m_\text{max}^2 + m_\text{max}) = \hbar^2 (m_\text{min}^2 - m_\text{min})$$
This tells us that $m_\text{min} = - m_\text{max}$.
Since the eigenvalue of $L_z$ is increased or decreased in steps of $\hbar$, $m_\text{min}$ and $m_\text{max}$ must be separated by an integer number, telling us that $j$ (and $m$) take on either integer or half-integer numbers and it also tells us that a (half)integer $j$ only has (half)integer $m$s.
Now we derived all the basics about angular momentum from nothing but the commutation relations. Make sure you really understand what's going on here, since this is very important. A thorough understanding of these formulas and their meaning will help you understand a lot of things, like angular momentum addition, tensor operators and the Wigner-Eckart-Theorem and finally even group theory, should you ever come across it.
