Drug discovery scientists who use Claude can now run ADMET predictions without leaving the chat window. The connector is live. No demos.
No research previews. Inductive Bio, an AI drug discovery partner, built it.
The company announced the move Monday. ADMET stands for Absorption, Distribution, Metabolism, Excretion, and Toxicity. It is the set of properties that determines whether a candidate molecule will survive from a lab bench to a patient.
Getting ADMET wrong is the reason most drug candidates fail in clinical trials. Inductive has built global prediction models for these properties.
Now those models sit inside Claude, accessible through Anthropic’s Model Context Protocol — an open standard for giving large language models access to external tools and data. The consequences for the drug discovery industry are straightforward. Bench scientists have been slow to adopt machine learning models for ADMET prediction.
The best models live inside specialist software that bench scientists do not use day-to-day. That has been a bottleneck. Inductive’s bet is that putting the models where scientists already work — inside a chat interface they already use — will change that.
The connector handles the handshake between Claude and Inductive’s prediction engine. The scientist never sees it.
They just get the prediction. Anthropic has been quietly assembling a life sciences ecosystem around Claude. Inductive is the latest addition.
The arrangement lets Inductive surface its models without requiring scientists to learn a new platform or toggle between tools. That matters.
Drug discovery is already tool-heavy. Adding another platform, another login, another training session — that is a non-starter for most bench scientists. A chat window they already have.
A connector they do not see. Predictions they can use. What to watch next.
Inductive describes itself as an AI drug discovery partner developing virtual chemistry labs. The company has built global prediction models for the five ADMET properties.
But ADMET is not the only bottleneck in drug discovery. Solubility. Permeability.
Metabolic stability. Each has its own set of models, its own specialist software, its own adoption problem.
If the Inductive connector works — if scientists actually use it — other model builders will follow. Anthropic’s ecosystem will grow. The MCP standard will get more real-world testing.
The drug discovery industry has been slow to adopt machine learning models for ADMET prediction. That is the problem Inductive is trying to solve. The connector is the solution.
Live now. No demos.
No research previews. Any drug discovery scientist working with Claude can access state-of-the-art predictive models directly in their existing workflow, according to the company. The question is whether they will.
