Massive Multiple Input Multiple Output (mMIMO) technology is essential for massive machine type communication (mMTC), where identifying active devices from a large pool via multiple access points (APs) is crucial. In emerging cell-free mMIMO (CF-mMIMO) systems, device activity detection is a critical task. This paper introduces a proximal gradient algorithm for detecting device activity in grant-free random-access scenarios within CF-mMIMO networks, specifically targeting two dominant APs. Each active device sends non-orthogonal pilot sequences to the APs, which then relay the received signals to a central processing unit (CPU) for joint activity detection. The proposed clustering-based proximal gradient algorithm significantly improves performance and efficiency in activity detection, offering reduced execution time compared to existing methods.