Brain-computer interfaces (BCIs) allow devices to be controlled with brain-generated potentials and can increase the independence of people with impaired nervous or muscular systems, the elderly and patients weakened after surgery, while reducing the workload of medical staff. This paper presents the design and implementation of a non-invasive brain-computer interface for controlling a hospital bed. Remote control is managed through an attachment equipped with pulling electromagnets (solenoids) that press specific buttons. Its operation is enabled by a classifier that detects steady-state visual evoked potentials (SSVEPs) induced by stimulation from a graphical interface. The BCI was evaluated in one session with one healthy participant and four squares flickering at 8.57-15 Hz. Five classifiers were compared in cross-validation with folds built from whole trials (four targets and an idle class, chance level 20%). Linear discriminant analysis with 2 s windows reached the highest balanced accuracy (99.2%), and phase-locking analysis confirmed SSVEP responses to all targets. In the online test with the decision rule fixed in advance, all 40 cued selections were correct (95% confidence interval 91.2-100%, median 3.22 s from cue to command), but false activations occurred at 1.6 per minute during 10 min of idle periods. In the final interface, all 16 operations of a predefined sequence were executed correctly from the EEG to the solenoids. They were switched off after a loss of EEG data or communication, and their coil reached at most 35.1 °C under the most demanding operation allowed by the firmware. The results demonstrate technical feasibility for one user. Idle-state detection must be improved and the complete system tested in the target group before clinical use.