We constrain cosmological parameters by analysing the angular power spectra of the Baryon Oscillation Spectroscopic Survey DR12 galaxies, a spectroscopic follow-up of around 1.3 million SDSS galaxies over 9,376 deg$^2$ with an effective volume of $sim 6.5$ (Gpc $h^{-1}$)$^3$ in the redshift range $0.15 leq z < 0.80$. We split this sample into 13 tomographic bins ($Delta z = 0.05$); angular power spectra were calculated using a Pseudo-$C_{ell}$ estimator, and covariance matrices were estimated using log-normal simulated maps. Cosmological constraints obtained from these data were combined with constraints from Planck CMB experiment as well as the JLA supernovae compilation. Considering a $w$CDM cosmological model measured on scales up to $k_{max} = 0.07h$ Mpc$^{-1}$, we constrain a constant dark energy equation-of-state with a $sim 4%$ error at the 1-$sigma$ level: $w_0 = -0.993^{+0.046}_{-0.043}$, together with $Omega_m = 0.330pm 0.012$, $Omega_b = 0.0505 pm 0.002$, $S_8 equiv sigma_8 sqrt{Omega_m/0.3} = 0.863 pm 0.016$, and $h = 0.661 pm 0.012$. For the same combination of datasets, but now considering a $Lambda$CDM model with massive neutrinos and the same scale cut, we find: $Omega_m = 0.328 pm 0.009$, $Omega_b = 0.05017^{+0.0009}_{-0.0008}$, $S_8 = 0.862 pm 0.017$, and $h = 0.663^{+0.006}_{-0.007}$ and a 95% credible interval (CI) upper limit of $sum m_{ u} < 0.14$ eV for a normal hierarchy. These results are competitive if not better than standard analyses with the same dataset, and demonstrate this should be a method of choice for future surveys, opening the door for their full exploitation in cross-correlations probes.