Web users frequently revisit pages that are of particular relevance to them. They also tend to have these documents on-screen and interact with them for longer periods than other pages. By recording document access frequency and activity metadata, which is based on transient user interactions within the browser, it is possible to infer the importance the user attaches to a given page. Activity metadata, access history metadata and document content can be stored in a locally held repository. This repository will help the user remember and quickly retrieve high interest documents they have accessed in the past. This paper discusses the nature of activity metadata generated and recorded during Web document use, how it relates to the document, and how it can be gathered, stored, represented and visualized for subsequent retrieval. Selected prototype implementation issues are also presented.