Aims Prolonged cardiac monitoring after stroke is recommended though there is no consensus on optimal methods. Short-term ECG recordings with a "thumb-ECG" device have shown promising preliminary results regarding effectiveness and cost benefit. We aimed to examine the performance of thumb-ECG and five days' Holter monitoring in a prospective trial. A secondary endpoint was the inter-observer agreement of the thumb-ECG. Methods Patients older than 65 years with no history of atrial fibrillation who suffered an acute stroke or transient ischemic attack of unknown origin were prospectively included. Patients were monitored for atrial fibrillation with five days' Holter and concurrent 30 s thumb-ECG twice daily, the latter continuing for 30 days. Inter-observer agreement for the thumb-ECG was determined. Results One hundred patients were included and 95 patients were analyzed. Paroxysmal atrial fibrillation was diagnosed in 20 patients with the thumb-ECG recordings and 17 patients on the Holter monitoring. Only 10 were diagnosed with both methods. The difference between the detection rates of the two devices was not significant ( p = 0.63). The inter-observer agreement of the thumb-ECG had a kappa value of 0.65. Conclusion Thirty days' thumb-ECG recordings twice daily for 30 s detect a high proportion of paroxysmal atrial fibrillation in a stroke or transient ischemic attack cohort. The proportion was comparable to five days' Holter monitoring but the agreement between the two methods was poor and the trial was not powered to detect a minor difference between the devices. The inter-observer agreement for the thumb-ECG was substantial. www.clinicalTrials.gov UI: NCT02261766.