Google Antigravity Adds Agentic Cheminformatics Skills

Google has introduced a suite of Cheminformatics Skills for Google Antigravity, a set of agentic tools designed to help chemists spend less time preparing and organizing molecular data and more time developing scientific hypotheses. The release is aimed at both computational chemists and experimental bench chemists, providing a collaborative science workbench with access to local chemistry tooling inside the Antigravity environment. The tools address a persistent challenge in modern chemistry research. Chemical knowledge is spread across millions of structures found in journal articles, vendor catalogs, patents, and databases, but the same molecule may appear in many different forms, including IUPAC names, Markush structures, and structure-data files. Because chemical meaning can change with even one atom or bond, automated systems must handle structure representations with very high precision. The new skills are intended to unify these sources while reducing manual errors in workflows that support drug discovery, materials science, and broader life and environmental sciences. Under the hood, the system combines language-model-based reasoning with deterministic chemistry software. Google says the underlying model handles workflow orchestration, database navigation, and scientific reasoning, while RDKit, the long-established open-source cheminformatics toolkit, performs molecular calculations. The skills also connect to curated chemistry resources including ChEMBL and PubChem, allowing the agent to access and interpret existing chemical knowledge using established domain practices. The announcement highlights rigidification as an example application. Rigidification is a medicinal chemistry technique used to improve drug potency by reducing the number of rotatable bonds in a molecule. Flexible drug candidates can adopt many conformations in solution, and each additional degree of freedom can create an entropic penalty when the molecule binds to its biological target. By adding bonds that lock important substructures into place, chemists can reduce that penalty and improve binding affinity. The article cites crizotinib and lorlatinib as examples, noting that lorlatinib was cyclized to reduce the effective number of rotatable bonds and increase potency. In the demonstration, the agent retrieved thousands of compounds, removed irrelevant entries such as buffers and solvents, and produced a documented report identifying candidate structures and source information. The system also provided early analytical insight, including that macrocycles in the retrieved set were largely predicted to be brain-penetrant, a relevant property for lung cancer research because brain metastasis is a major clinical concern. The Cheminformatics Skills are available today through Google Antigravity. Users can access them by installing the Science plugin in the Antigravity App. Enterprise users can sign in using a business account with Google Cloud credentials. To accelerate adoption in research communities, Google is also offering the highest usage tier of Google Antigravity free of charge for six months to 10,000 eligible scientists, with onboarding handled on a rolling basis subject to capacity and selection.