The World Wide Web is one of the most popular applications currently running on the Internet, and its size is growing exponentially. Web caching is an important technique that aims to reduce network traffic, server load, and user-perceived retrieval delays by replicating popular contents on proxy servers. Anycast is a new network service widely used for providing auto-configuration and load-balancing. We present an anycast-based cooperative proxy algorithm (ACPA) that brings a server "nearest" to a client to improve overall performance by selective allowing duplicates and migrate objects to other proxies based on request pattern. Analytical and experimental results show that the algorithm outperforms a hash-based algorithm in response time and hop counts.