Context. Precise stellar radial velocities (RVs) are used to search for massive (Jupiter masses or higher) exoplanets around the stars of the open cluster M 67. Aims. We aim to obtain a census of massive exoplanets in a cluster of solar metallicity and age in order to study the dependence of planet formation on stellar mass and to compare in detail the chemical composition of stars with and without planets. This first work presents the sample and the observations, discusses the cluster characteristics and the RV distribution of the stars, and individuates the most likely planetary host candidates. Methods. We observed a total of 88 main-sequence stars, subgiants, and giants all highly probable members of M 67, using four telescopes and instrument combinations: the HARPS spectrograph at the ESO 3.6 m, the SOPHIE spectrograph at OHP, the CORALIE spectrograph at the Euler swiss telescope and the HRS spectrograph at Hobby Eberly Telescope. We investigate whether exoplanets are present by obtaining RVs with precisions as good as 10 m s −1 . To date, we have performed 680 single observations (Dec. 2011) and a preliminary analysis of data, spanning a period of up to eight years. We computed zero-point deviations for each spectrograph with respect to HARPS, finding that for SOPHIE and CORALIE the offsets are minimal (at −11.4 m s −1 and 26.8 m s −1 , respectively), while for our HET measurements the offset is larger, 242.0 m s −1 . After reducing all the observations to the HARPS zero point, the RV measurements for each star are used to evaluate the RV variability along the cluster color magnitude diagram (CMD). Results. Although the sample was pre-selected to avoid the inclusion of binaries, we identify 11 previously unknown binary candidates. The RV variance (including the observational error) for the bulk of stars is almost constant with stellar magnitude (therefore stellar gravity) at σ = 20 m s −1 . This number includes both the stellar intrinsic variability and the observational error, which is the major source of uncertainty for the faintest stars. Eleven stars clearly displayed larger RV variability and these are candidates to host long-term substellar companions. The average RV is also independent of the stellar magnitude and evolutionary status, confirming that the difference in gravitational redshift between giants and dwarfs is almost cancelled by the atmospheric motions. We use the subsample of solar-type stars to derive a precise true RV for this cluster, and we use asteroid observations to derive the zero point of the HARPS G star mask of 94.5 m s −1 . The true RV of the cluster is M 67 RV = 33.74 ± 0.12 km s −1 . The velocity dispersion is 0.54 km s −1 for giants and 0.68 km s −1 for dwarfs, which in both cases is substantially lower than reported in previous works. The higher velocity dispersion of the lower mass stars is confirmed by these observations, in which for the first time the RV measurement precision is much smaller than the cluster dispersion. We finally create a catalog of binaries a...