Optical microscopy analyses using thin sections is one of the most standard techniques in geology. It permits identification of most rock-forming minerals and it is essential for micropaleontology since observation in transmitted light is the only way to observe individual, or colonies of, microfossils in their mineralogical and textural context. Thin sections are also used for other techniques such as Raman spectroscopy. In the framework of exploration of the Martian surface and of the search for potential relics of microbial life, the possibility to observe petrographic thin sections in situ would be revolutionary. Nevertheless, although optical microscopes can be readily designed for space exploration, thin section preparation is not easy to do in situ on extraterrestrial rocky bodies due to the generally harsh environmental conditions and to the difficulty to automate a protocol that normally requires several human interventions. The aim of the LithoSpace project, supported by the French Space Agency (CNES) since 2014, is to work on the development of an automated system permitting preparation of petrographic thin sections on extraterrestrial bodies, such as Mars, the Moon or asteroids. In this paper, we describe the approaches used to study rocks in situ during space missions and compare them to those used on Earth. Then, the relevance of making thin sections during space exploration is highlighted from the description of the data that could be collected from the analysis of these specific samples, using new instruments and instruments already sent to Mars. The standard preparation protocol is described as well as previously envisioned or existing in situ sample preparation systems. We then propose a protocol for automation of the preparation. Particular attention is paid to demonstrating the feasibility of making thin sections in the absence of liquid water and with energy consumption compatible with automated space probes. Tests are then carried out to control the quality of the prepared samples. On the basis of the demonstrated feasibility, an automated system is proposed as a conceptual all-in-one system. Finally, a “proof-of-concept” model developed with the help of students at different educational levels is presented.