Aims. We present a 69 arcmin 2 ALMA survey at 1.1 mm, GOODS-ALMA, matching the deepest HST-WFC3 H -band part of the GOODS-South field. Methods. We tapered the 0″24 original image with a homogeneous and circular synthesized beam of 0″60 to reduce the number of independent beams – thus reducing the number of purely statistical spurious detections – and optimize the sensitivity to point sources. We extracted a catalog of galaxies purely selected by ALMA and identified sources with and without HST counterparts down to a 5 σ limiting depth of H  = 28.2 AB (HST/WFC3 F160W). Results. ALMA detects 20 sources brighter than 0.7 mJy at 1.1 mm in the 0″60 tapered mosaic (rms sensitivity σ ≃ 0.18 mJy beam −1 ) with a purity greater than 80%. Among these detections, we identify three sources with no HST nor Spitzer -IRAC counterpart, consistent with the expected number of spurious galaxies from the analysis of the inverted image; their definitive status will require additional investigation. We detect additional three sources with HST counterparts either at high significance in the higher resolution map, or with different detection-algorithm parameters ensuring a purity greater than 80%. Hence we identify in total 20 robust detections. Conclusions. Our wide contiguous survey allows us to push further in redshift the blind detection of massive galaxies with ALMA with a median redshift of z  = 2.92 and a median stellar mass of M ⋆  = 1.1 × 10 11   M ⊙ . Our sample includes 20% HST-dark galaxies (4 out of 20), all detected in the mid-infrared with Spitzer -IRAC. The near-infrared based photometric redshifts of two of them ( z  ∼ 4.3 and 4.8) suggest that these sources have redshifts z  >  4. At least 40% of the ALMA sources host an X-ray AGN, compared to ∼14% for other galaxies of similar mass and redshift. The wide area of our ALMA survey provides lower values at the bright end of number counts than single-dish telescopes affected by confusion.