Recommendation System (RS) is software that suggests similar items to a purchaser based on his/her earlier purchases or preferences. RS examines huge data of objects and compiles a list of those objects which would fulfil the requirements of the buyer. Nowadays most ecommerce companies are using Recommendation systems to lure buyers to purchase more by offering items that the buyer is likely to prefer. Book Recommendation System is being used by Amazon, Barnes and Noble, Flipkart, Goodreads, etc. to recommend books the customer would be tempted to buy as they are matched with his/her choices. The challenges they face are to filter, set a priority and give recommendations which are accurate. RS systems use Collaborative Filtering (CF) to generate lists of items similar to the buyer’s preferences. Collaborative filtering is based on the assumption that if a user has rated two books then to a user who has read one of these books, the other book can be recommended (Collaboration). CF has difficulties in giving accurate recommendations due to problems of scalability, sparsity and cold start. Therefore this paper proposes a recommendation that uses Collaborative filtering with Jaccard Similarity (JS) to give more accurate recommendations. JS is based on an index calculated for a pair of books. It is a ratio of common users (users who have rated both books) divided by the sum of users who have rated the two books individually. Larger the number of common users higher will be the JS Index and hence better recommendations. Books with high JS index (more recommended) will appear on top of the recommended books list.