Motor Imagery (MI) based Brain-Computer Interfaces (BCI) can facilitate very early motor rehabilitation during the acute post-stroke period that is characterised by both greater impairment and, paradoxically, greater plasticity. Current Electroencephalography (EEG) based approaches require substantial time and effort to learn to use. We suggest that a multimodal and multiphase approach may allow stroke survivors to learn to use the BCI more rapidly, enabling early bedside intervention.In phase one, for two sessions, 17 participants learned to perform MI with the help of Transcranial Magnetic Stimulation-derived Neurofeedback (TMS-NF) of their muscle responses. In phase two, participants conducted EEG BCI for three sessions. We compared the classification accuracy on the EEG BCI of the TMS-NF group to a control group (N=17) learning for five sessions with only an EEG BCI.The TMS group achieved better control and did so faster than the EEG group. Further, the spatial distribution of imagery-based neural activity in the TMS group more closely resembled patterns seen for movement execution.Priming using two days of TMS-NF may assist EEG-BCI users to develop robust mental imagery strategies that transfer well to an EEG BCI, enabling better control within a shorter timeframe. Given that EEG BCI is more portable and cost effective, this may enable a two-phase approach where TMS-NF is used at the bedside in the early days after stroke followed by self-directed EEG BCI rehabilitation at home using wireless wearable technology.