We present ShakeCast, a system for automatic peer-to-peer exchange of contact information between two persons who just shook hands. The accelerometer in a smartwatch is used to detect the physical handshake and implicitly triggers a setup-free information transfer between the users' personal smartphones using Bluetooth LE broadcasts. An abstract representation of the handshake motion data is used to disambiguate between multiple simultaneous transmissions and to prevent accidental data leakage.To evaluate our system, we collected individual wrist acceleration data from 130 handshakes, performed by varying combinations of 20 volunteers. We present a systematic analysis of possible data features which can be used for disambiguation, and we validate our approach using the most salient features. Our analysis shows an expected match rate between corresponding handshakes of 92.3%.