Automatic differentiation (AD) has proven its interest in many fields ofapplied mathematics, but it is still not widely used. Furthermore, existingnumerical methods have been developed under the hypotheses that computingprogram derivatives is not affordable for real size problems. Exact derivativeshave therefore been avoided, or replaced by approximations computed by divideddifferences. The hypotheses is no longer true due to the maturity of AD addedto the quick evolution of machine capacity. This encourages the development ofnew numerical methods that freely make use of program derivatives, and willrequire the definition and development of new AD strategies. AD tools mustbe extended to produce these new derivative programs, in such a modular waythat the different sub-problems can be solved independently from one another.Flexibility assures the user to be able to generate whatever specificderivative program he needs, with at the same time the possibility to generatestandard ones. This paper sketches a new model of modular, extensible andflexible AD tool that will increase tenfold the DA potential for appliedmathematics. In this model, the AD tool consists of an AD kernel namedKAD supported by a general program transformation platform.