pyADCG: A preliminary implementation of a new parallel solver for nonconvex MINLPs in Pyomo/Python. In this talk we present pyADCG, a preliminary implementation of a new parallel decomposition method for nonconvex MINLPs in Pyomo/Python. The new optimization method, called ADCG (Alternating Direction Column Generation), is not based on the branch-and-bound approach. The basic idea of ADCG is to restrict the objective value by a target constraint and to check via a column generation based globally convergent alternating direction method if the resulting MINLP is feasible or not. Convergence is shown by using the fact that the duality gap of a general nonconvex projection problem is zero, see http://www.optimization-online.org/DB_HTML/2015/12/5233.html. We discuss algorithmic variants and report first numerical results.