Robots are steadily becoming one of the significant 21st century learning technologies that aim to improve education within both formal and informal environments. Such robots, called Robots for Learning, have so far been utilized as constructionist tools or social agents that aided learning from distinct perspectives. This thesis presents a novel approach to Robots for Learning that aims to explore new added values by means of investigating uses for robots in educational scenarios beyond those that are commonly tackled: We develop a platform from scratch to be "as versatile as pen and paper", namely as composed of easy to use objects that feel like they belong in the learning ecosystem while being seamlessly usable across many activities that help teach a variety of subjects. Following this analogy, we design our platform as many low-cost, palm-sized tangible robots that operate on printed paper sheets, controlled by readily available mobile computers such as smartphones or tablets. From the learners' perspective, our robots are thus physical and manipulable points of hands-on interaction with learning activities where they play the role of both abstract and concrete objects that are otherwise not easily represented. We realize our novel platform in four incremental phases, each of which consists of a development stage and multiple subsequent validation stages. First, we develop accurately positioned tangibles, characterize their localization performance and test the learners' interaction with our tangibles in a playful activity. Second, we integrate mobility into our tangibles and make them full-blown robots, characterize their locomotion performance and test the emerging notion of moving vs. being moved in a learning activity. Third, we enable haptic feedback capability on our robots, measure their range of usability and test them within a complete lesson that highlights this newly developed affordance. Fourth, we develop the means of building swarms with our haptic-enabled tangible robots and test the final form of our platform in a lesson co-designed with a teacher. Our effort thus contains the participation of more than 370 child learners over the span of these phases, which leads to the initial insights into this novel Robots for Learning avenue. Besides its main contributions to education, this thesis further contributes to a range of research fields related to our technological developments, such as positioning systems, robotic mechanism design, haptic interfaces and swarm robotics.