A quadratic simplex algorithm tailored to the primal optimization over the vertices of zero-one polytopes is presented. It computes a locally optimum basic feasible solution and it so generalizes over local improvement heuristics for corresponding applications such as unconstrained binary quadratic programming, maximum cut, or the quadratic assignment problem.