We show that the Casimir force (CF) gradient can be measured with no contact involved. Results of the CF measurement with systematic uncertainty of 3% are presented for the distance range of 100–600 nm. The statistical uncertainty is shown to be due to the thermal fluctuations of the force probe. The corresponding signal-to-noise ratio equals unity at the distance of 600 nm. Direct contact between surfaces used in most previous studies to determine absolute distance separation is here precluded. Use of direct contact to identify the origin of distances is a severe limitation for studies of the CF on structured surfaces as it deteriorates irreversibly the studied surface and the probe. This force machine uses a dynamical method with an inserted gold sphere probe glued to a lever. The lever is mechanically excited at resonant frequency in front of a chosen sample. The absolute distance determination is achieved to be possible, without any direct probe/sample contact, using an electrostatic method associated to a real time correction of the mechanical drift. The positioning shift uncertainty is as low as 2 nm. Use of this instrument to probe a very thin film of gold (10 nm) reveals important spatial variations in the measurement.