Tor is a widely used network for anonymous communication. Its users frequently experience large communication delays, due to the high user-to-relay ratio, the bandwidth-intensive BitTorrent transfers of a small fraction of the user base, and the inherent latencies from routing traffic through multiple relay hops scattered around the world. These delays significantly degrade the user experience of web browsing, a dominant use of Tor. Improving web browsing performance of Tor has been a subject of much research. Targeting the network and transport layers, prior work includes proposals to throttle bandwidth-intensive connections or to prioritize interactive traffic such as web browsing. We attack the problem at the application layer, noting that a typical web page consists of multiple resources, each of which requires a one-round-trip HTTP request-response cycle to load in the browser. Thus, for a page with many resources, these round trips are a major contributor to the page load time. We investigate PnP (for Prefetch-and-Push), where the Tor exit prefetches resources of the web page a client is visiting and pushes them to the client. Our experiments show a significant reduction in page load times as well as higher client's privacy from web page fingerprinting by a local attacker.