We present a fully synchronized multimodal dataset of electroencephalography (EEG), functional near-infrared spectroscopy (fNIRS), galvanic skin response (GSR), heart rate (HR), RR intervals, and photoplethysmography (PPG) collected from 8 healthy adults across five sessions each (40 recordings). Each session followed a three-phase design: 1 min resting baseline, 3 min guided breathing meditation, and 1 min recovery. All modalities were acquired simultaneously with unified timestamps, standardized placement, and harmonized sampling (EEG/fNIRS 250 Hz; PPG 250 Hz; GSR 10 Hz; HR/RR resampled to 10 Hz). The release provides raw and minimally processed signals, channel maps, synchronized event markers, signal-quality flags, and preprocessing logs. Data are distributed as a 250 Hz MATLAB .mat file per session (time axis, fNIRS HbR/HbO, EEG, PPG, markers), a ZIP export with CSV files time-stamped by Unix time (GSR, HR, HRV/RR, events), and a metadata.json describing configuration and checksums. The dataset supports studies of meditation-related physiology, central–autonomic coupling, and benchmarking of multimodal fusion or arousal-detection methods without prescribing specific models.