Organic wastes originating from livestock, agro-industry or urban activities may represent true resources when recycled for new uses, for example, as soil improvers, organic fertilizers or bioenergy sources. The compositional characteristics of these organic resources (ORs) can vary considerably depending on origin, nature, processing, stage, and state. Despite being of potential interest to different stakeholders in a circular economy, the variability in OR characteristics and the difficulty of accessing reliable, fast and inexpensive analysis methods may curb the recycling of OR in the agriculture or bioenergy sectors. As is the case in other low-income countries, scarcity of data on OR characteristics and the difficulty in assessing these data (due to cost and the sparsity of laboratories) is particularly acute in Madagascar, thus impairing the rational utilization of OR in the agricultural or bioenergy sectors. Visible-near infrared spectroscopy (VIS-NIR) has proven to be suitable for the fast, reliable and low-cost determination of the composition of different ORs, usually through the development of calibration models based on one type of OR by single research or lab groups. It is challenging to develop VIS-NIR models based on several types of ORs encompassing a wide range of target characteristics. Another challenging issue is the extension of databases containing spectra acquired on different spectrometers to increase model genericity. In both cases, standardization can be performed to resolve the problem of developing models for diverse ORs whose spectra originate from different laboratories. To assess the ability to develop VIS-NIR models with as much genericity as possible, we built a large database containing a wide diversity of ORs produced in Madagascar. The data presented in this paper were obtained by chemical and spectral analyses of 1,0 0 0 ORs collected from five districts in Madagascar. The data are accompanied by fine-grained metadata defined by 32 descriptors of ORs, including origin (animal, agroindustrial, and urban); nature (manure, agro-industrial waste, and compost); farm type (smallholder and agricultural factory); exploitation type (smallholder farm, factory farm, on farm compost facility, and town compost facility); diversity of animal feed, litter, sex, and age; and diversity of bedding material. The chemical properties (including the organic nitrogen, organic carbon, organic matter, inorganic matter, phosphorus, potassium, calcium, magnesium, zinc, copper, nickel, chromium, cadmium, and lead and soluble, hemicellulose, cellulose, lignin and cutin fractions) were analyzed following laboratory standards. The number of analyses performed ranged from 39 to 180 depending on the chemical property. VIS-NIR spectra were acquired using a Labspec spectrometer. To facilitate the merging of spectral data or the development of VIS-NIR models based on broad datasets, the spectra were presented in raw form and after standardization. The dataset is original in terms of sources and width. This dataset should be of particular interest to chemometricians, biogeochemists, agronomists, energy planners, hygienists and other professionals involved in recycling ORs for various new purposes in low-income countries and elsewhere.