The Buccaneer software is described for automated model building in protein structures from electron-density maps. The method uses an oriented electron-density likelihood target function to identify likely $ C^\alpha $ positions and extend them into chain fragments. The process involves finding initial $ C^\alpha $ 'seed' positions, growing these into chain fragments, joining the fragments into chains, and pruning clashing chains. The approach is based on a six-dimensional search for $ C^\alpha $ positions, using a density-likelihood function that scores possible positions and orientations in the electron-density map. The likelihood function is derived from a simulated electron-density map and uses Bayes' theorem to calculate the probability of a correctly positioned $ C^\alpha $ group. The method is efficient and works well at resolutions better than 3.0 Å. The software has been tested on 58 structures from the JCSG data archive, showing good performance in terms of completeness and accuracy, especially when the initial phases are of high quality. The method is complementary to ARP/wARP, which is more sensitive to data resolution but can give results with poor phases. The Buccaneer approach is suitable for future development as a basis for automated model-building systems. The software is implemented using the Clipper crystallographic libraries and is efficient, taking minutes to an hour to process data. The method is dependent on the quality of the initial experimental phasing and phase improvement. The results show that the Buccaneer method can trace protein chains even at low resolutions, but its accuracy is strongly related to the quality of the initial phases. The method has been tested on a low-resolution data set, showing that it can trace helical regions with high precision, but non-helical regions are less accurate. The method can also overinterpret solvent as protein features in some cases, which can be identified using the 'Density fit analysis' feature of the Coot model-building program. The Buccaneer method is a promising approach for automated model building in protein structures from electron-density maps.The Buccaneer software is described for automated model building in protein structures from electron-density maps. The method uses an oriented electron-density likelihood target function to identify likely $ C^\alpha $ positions and extend them into chain fragments. The process involves finding initial $ C^\alpha $ 'seed' positions, growing these into chain fragments, joining the fragments into chains, and pruning clashing chains. The approach is based on a six-dimensional search for $ C^\alpha $ positions, using a density-likelihood function that scores possible positions and orientations in the electron-density map. The likelihood function is derived from a simulated electron-density map and uses Bayes' theorem to calculate the probability of a correctly positioned $ C^\alpha $ group. The method is efficient and works well at resolutions better than 3.0 Å. The software has been tested on 58 structures from the JCSG data archive, showing good performance in terms of completeness and accuracy, especially when the initial phases are of high quality. The method is complementary to ARP/wARP, which is more sensitive to data resolution but can give results with poor phases. The Buccaneer approach is suitable for future development as a basis for automated model-building systems. The software is implemented using the Clipper crystallographic libraries and is efficient, taking minutes to an hour to process data. The method is dependent on the quality of the initial experimental phasing and phase improvement. The results show that the Buccaneer method can trace protein chains even at low resolutions, but its accuracy is strongly related to the quality of the initial phases. The method has been tested on a low-resolution data set, showing that it can trace helical regions with high precision, but non-helical regions are less accurate. The method can also overinterpret solvent as protein features in some cases, which can be identified using the 'Density fit analysis' feature of the Coot model-building program. The Buccaneer method is a promising approach for automated model building in protein structures from electron-density maps.