The paper introduces a new technique for automated tracing of protein chains in experimental electron-density maps using an oriented electron-density likelihood target function. This function is applied to identify likely Cα positions and grow them into extended chain fragments. The method is built on existing graphical and non-graphical model-building tools but incorporates a new orientation-dependent target function, which provides additional directional information and enhances sensitivity. The process involves four steps: finding initial Cα 'seed' positions, growing these positions into chain fragments, joining fragments into chains, and pruning clashing chains. The method is demonstrated through tests on 58 structures from the Joint Center for Structural Genomics (JCSG) data archive, showing reasonable performance at low resolutions (better than 3.0 Å) and sensitivity to the quality of initial phases. The software, called *Buccaneer*, is simple and fast, but lacks some advanced features like sequence docking and refinement. Initial results suggest it is a promising basis for future development in automated model building.The paper introduces a new technique for automated tracing of protein chains in experimental electron-density maps using an oriented electron-density likelihood target function. This function is applied to identify likely Cα positions and grow them into extended chain fragments. The method is built on existing graphical and non-graphical model-building tools but incorporates a new orientation-dependent target function, which provides additional directional information and enhances sensitivity. The process involves four steps: finding initial Cα 'seed' positions, growing these positions into chain fragments, joining fragments into chains, and pruning clashing chains. The method is demonstrated through tests on 58 structures from the Joint Center for Structural Genomics (JCSG) data archive, showing reasonable performance at low resolutions (better than 3.0 Å) and sensitivity to the quality of initial phases. The software, called *Buccaneer*, is simple and fast, but lacks some advanced features like sequence docking and refinement. Initial results suggest it is a promising basis for future development in automated model building.