The dataset covers all public procurement contracts published in the dedicated official journal in France from 2015 to 2023. A collection script first reads the raw data from the online notices and processes them into a cleaned table. Then, we use the same machine learning algorithm as in Potin et al. (2023) to identify public authorities and companies so that the notice data can be merged with individual information on the contracting parties. We obtain about one million contractual relationships covering more than 300,000 public contracts from all sectors and institutions. This comprehensive dataset is also quite accurate, as it potentially contains about 100 variables for each observation. These variables relate to contract characteristics (procedure, subject matter, award criteria, clauses...), award outcomes (award price, number of bids...), public authorities (type, geolocation, main activity...) and companies (size, legal status, main activity, age, geolocation...). The dataset is unprecedented in its accuracy and scope, providing reliable and detailed information on every advertised contract in France for nearly a decade.