SummaryEspecially in knowledge-rich domains such as medicine perfect access to the literature is essential for professionals. Unfortunately, especially in knowledge-rich domains it is difficult to achieve perfect access: it is too difficult and too time consuming for users to formulate queries that yield the maximum of relevant documents and a minimum of non-relevant ones. The paper first discusses the challenges of information retrieval in medicine and various existing approaches. To address the challenges two completely opposite approaches are presented. The first supports the user by means of metabrowsing: a visual way of depicting the relations between domain concepts and documents. Metabrowsing relieves the user from the formulation of queries, while leaving him in full control. The second approach aims to minimize the interaction with the user. Information needs and queries are autonomously and proactively formulated by a software-agent who remains invisible to the user. The agent uses the electronic patient record of a particular patient and domain knowledge. As a result, the agent provides the doctor with literature that is relevant with respect to the patient at hand.