Context. We present additional constraints on the galaxy luminosity function at z ~ 9 based on observations carried out with ESO/VLT FORS2, HAWK-I, and X-Shooter around the lensing cluster A2667, as part of our project designed to select z ~ 7-10 candidates accessible to spectroscopy. We find that only one selected J-dropout source in this field fulfills the color and magnitude criteria. This source was recently confirmed as a mid-z interloper based on X-Shooter spectroscopy. Aims: Owing to the considerable depth and area covered by our survey, we are able to set strong constraints on the bright end of the galaxy luminosity function and hence on the star formation history at very high redshift. Methods: We used our non-detection of reliable J-dropout sources over the ~36 arcmin2 field of view towards A2667 to carefully determine the lens-corrected effective volume and the corresponding upper limit to the density of sources. Results: Our strongest limit is obtained for Φ(M1500 = -21.4 ± 0.50) < 6.70 × 10-6 Mpc-3 mag-1 at z ~ 9. A maximum-likelihood fit of the luminosity function to all available data points including the present new result yields M⋆ > -19.7 with fixed α = -1.74 and Φ⋆ = 1.10 × 10-3 Mpc-3. The corresponding star-formation rate density should be ρSFR < 5.97 × 10-3 M⊙ yr-1 Mpc3 at z ~ 9. These results are in good agreement with the most recent estimates already published for this range of redshift and luminosity domain. Conclusions: This new result confirms previously measured decreases in the density of luminous galaxies at very high redshift, hence provides strong constraints on the design of future surveys aiming to explore the very high-redshift Universe. Based on observations collected at The European Southern Observatory, Paranal, Chile, as part of the ESO 082.A-0163 and 087.A-0118.