Résumé : We have designed a powerful new algorithm to detect stellar streams in an automated and systematic way. The algorithm, which we call the STREAMFINDER, is well suited for finding dynamically cold and thin stream structures that may lie along any simple or complex orbits in Galactic stellar surveys containing any combination of positional and kinematic information. In the present contribution we introduce the algorithm, lay out the ideas behind it, explain the methodology adopted to detect streams and detail its workings by running it on a suite of simulations of mock Galactic survey data of similar quality to that expected from the ESA/Gaia mission. We show that our algorithm is able to detect even ultra-faint stream features lying well below previous detection limits. Tests show that our algorithm will be able to detect distant halo stream structures $>10^{\circ}$ long containing as few as $\sim 15$ members ($\Sigma_{\rm G} \sim 33.6\, {\rm mag \, arcsec^{-2}}$) in the Gaia dataset.