The last decade has witnessed a growing interest in applying machine learning (ML) tools to traditional sciences such as chemistry. This has translated into an increasing number of publications in the field, with hardly one day passing without a new work addressing a problem in chemistry with the help of ML. Unfortunately, while many scientists open-source their code and models, reproducibility is not always granted, and their accessibility is limited to a technology-affine audience. Moreover, and most importantly, the new tools hardly ever reach the chemists whom they were designed for. Early on, our team aspired to make the ML tools developed at IBM Research available to a broad audience of chemists. To achieve this, we introduced the IBM for Chemistry website (https://rxn.res.ibm.com), a cloud-based graphical platform allowing chemists without programmatic knowledge to apply different ML models for chemical reactivity. Graphical platforms such as this one represent an intuitive way to introduce chemists to the opportunities offered by ML. To further increase the adoption of such technologies, one must go beyond an interactive use on the web browser and provide access to the underlying programmatic interfaces, allowing for (i) making predictions with published ML models without the need for specific computational environments or hardware, (ii) automating the predictions on large sets of compounds or reactions, or (iii) embedding the functionality in other tools and workflows. In addition to providing programmatic access to the IBM RXN API, we implemented an associated Python wrapper (https://github.com/rxn4chemistry), enabling easy adoption of the API. In this talk, we will present a demo of this tool and explain the challenges and outlooks of graphical and programmatic interfaces in the field of ML for chemistry.