In this paper an adaptive LZW data compression algorithm using changeable MCL (Maximum-Code-Length) is proposed. It is an enhancement for the conventional LZW (proposed by Lempel, Ziv and Welch [4]) algorithm. In the conventional LZW algorithm, the MCL is a fixed value (e.g. 16-bit). That is, when the code value was increased to 2^MCL the compressor has to clear the dictionary even if it offers a high hit rate. On the other hand, the compressor cannot reset the dictionary with a low hit rate until the code value reaches 2^MCL. To overcome this disadvantage, we employ an adaptive MCL mechanism in our algorithm. According to the adaptive MCL, the compressor can decide when to clear the dictionary intelligently. Because this algorithm can choose the better opportunity to clear the dictionary, so the new algorithm can make better use of the dictionary and obtain better compression ratio than the conventional LZW. If the threshold value (the maximum MCL chosen according to the memory size) was properly chosen, this algorithm can be applied in SOC (System-On-a-Chip) testing or other applications in embedded systems. Moreovel; the notable advantage of this algorithm is that it can be adopted alone with other variations of the LZW