Low energy X-ray Intra-Operative Radiation Therapy (IORT) is used mostly for breast cancer treatment with spherical applicators. X-ray IORT treatment delivered during surgery (ex: INTRABEAM®, Carl Zeiss) can benefit from accurate and fast dose prediction in a patient 3D volume. However, full Monte Carlo (MC) simulations are time-consuming and no commercial treatment planning system was available for this treatment delivery technique. Therefore, the aim of this work is to develop a dose computation tool based on MC phase space information, which computes fast and accurate dose distributions for spherical and needle INTRABEAM® applicators. First, a database of monoenergetic phasespace (PHSP) files and depth dose profiles (DDPs) in water for each applicator is generated at factory and stored for on-site use. During commissioning of a given INTRABEAM® unit, the proposed Fast and Optimized Phase-Space (FOPS) generation process creates a phasespace at the exit of the applicator considered, by fitting the energy spectrum of the source to a combination of the monoenergetic precomputed phase-spaces, by means of a genetic algorithm, with simple experimental data of depth dose profiles in water provided by the user. An in-house hybrid MC algorithm which takes into account condensed history simulations of photoelectric, Rayleigh and Compton interactions for X-rays up to 1 MeV computes the dose from the optimized phase-space file. The whole process has been validated against radiochromic films in water as well as reference MC simulations performed with penEasy in heterogeneous phantoms. From the pre-computed monoenergetic PHSP files and DDPs, building the PHSP file optimized to a particular depth-dose curve in water only takes a few minutes in a single core (i7@2.5 GHz), for all the applicators considered in this work, and this needs to be done only when the X-ray source is replaced. Once the phase-space file is ready, the hybrid Monte Carlo code is able to compute dose distributions within 10 minutes. For all the applicators, more than 95% of voxels from dose distributions computed with the FOPS+hybrid code agreed within 7%-0.5 mm with both reference MC simulations and measurements. The method proposed has been fully validated and it is now implemented into radiance (GMV SA, Spain), the first commercial IORT Treatment Planning System (TPS).