Massive parallel sequencing (MPS) technologies have paved the way into new areas of research including individualized medicine. However, sequencing of trace amounts of DNA or RNA still remains a major challenge, especially for degraded nucleic acids like circulating DNA. This together with high cost and time requirements impedes many important applications of MPS in medicine and fundamental science. We have established a fast, cheap and highly efficient protocol called 'Capture and Amplification by Tailing and Switching' (CATS) to directly generate ready-to-sequence libraries for MPS from nanogram and picogram quantities of both DNA and RNA. Furthermore, those DNA libraries are strand-specific, can be prepared within 2-3 h and do not require preliminary sample amplification steps. To exemplify the capacity of the technique, we have generated and sequenced DNA libraries from hundred-picogram amounts of circulating nucleic acids isolated from human blood plasma, one nanogram of mRNA-enriched total RNA from cultured cells and few nanograms of bisulfite-converted DNA. The approach for DNA library preparation from minimal and fragmented input described here will find broad application in diverse research areas such as translational medicine including therapy monitoring, prediction, prognosis and early detection of various human disorders and will permit high-throughput DNA sequencing from previously inaccessible material such as minute forensic and archeological samples.