The recent explosion of social media sites is creating an unprecedented opportunity to build radically new types of learning environments, which use all the capabilities of social media but augmented with learning agents that can foster collaborative learning by doing. Music is a potentially very attractive domain to test this idea. There is an enormous need for education in music but personal music tuition has become unaffordable or impossible due to the large increase in demand. Very recently, social music media sites have sprung up, some attracting close to a million amateurs and professional musicians. Users share what they are doing within a community of trusted peers, give comments, create pieces together, and establish commercial and social relations. PRAISE proposes to go beyond what is currently available by building learning agents to act as guardians and tutors in a community of music practice who constantly monitor the student providing ongoing personalised praise and feedback. The realisation of this vision requires breakthroughs at two levels: pedagogical and technological. At a pedagogical level we need to find new principles that can work in these novel environments, such as personalised reflexive feedback, and ways in which elements from the standard educational curricula can be embedded seamlessly, triggered by personal needs. At a technological level we need significant advances in real-time content, gesture and linguistic analysis, community infrastructure, student modelling and educational interaction design. The project will not only perform the research needed to build a fully operational system but also carry out exciting case studies targeting the acquisition of cognitively challenging skills for Jazz and popular music both with children and advanced learners. Giving praise will be the commonly agreed mechanism for providing feedback so that the overall system stimulates the motivation and enhances the confidence of learners.