The work aims to solve the problems of low efficiency of milling in traditional micro-groove machining process for micro-fluidic chip. The machining process for tungsten carbide die was studied by femtosecond laser. Scanning power, scanning speed and scanning layer were adopted for orthogonal test to study their influence on surface quality and size precision. Roughness of the machined surface was measured by a white light interferometer, and machined depth was measured by a VHX-1000. Through a comprehensive optimization to those results, optimal process parameters were obtained. In selected 25 sets of parameters, the variation of each parameter had little effect on the surface roughness. The factor having the greatest influence on the machined depth was the scanning speed, and the factor having the minimal influence was the scanning power. After comprehensive optimization, the best combination of process parameters could be obtained when the power was 20 W, the scanning speed was 200 mm/s, and the scanning layer was 25. The machined depth can experience a remarkable growth as the femtosecond laser power increases, the scanning speed decreases, or the canning layer increases. The width of upper surface of the mold obtained by the optimum processing parameters is 152 μm, the edges are neat, the depth is 42.41 μm, and the bottom surface roughness is 0.164 μm. The results of the study provide some reference for tungsten carbide microfluid grooves by femtosecond laser machining.