Claude Science: Anthropic’s New AI Agent for Drug Discovery and Research

Anthropic announced Claude Science on June 30, 2026, at a closed event for pharmaceutical executives, biotech founders, and researchers. The product is built to support scientific research the same way Claude Code supports software engineering: give it a high-level instruction, and it handles meaningful work on its own. It ships with tools tuned for computational biology and drug development, and it is available immediately to all paid Claude subscribers.

What happened

Anthropic used a pharma and biotech industry event on June 30, 2026, to introduce Claude Science, a new flagship product positioned alongside Claude Code in its product lineup. Like Claude Code, Claude Science is an autonomous agent: give it a concise, high-level prompt and it will carry out substantive research tasks without hand-holding.

The product ships with access to specialist tools that make it particularly useful for two areas:

• Computational biology
• Drug development and discovery

Anthropic also revealed it intends to use Claude Science internally to conduct its own research into treatments for rare and neglected diseases. This is not, according to MIT Technology Review, the company’s first move into AI for scientific work, but Claude Science represents its most direct, packaged product bet on the space so far.

Why it matters

The pattern Anthropic is following is clear. Claude Code gave software developers an autonomous coding agent with access to relevant tools. Claude Science applies that same model to research workflows. If it holds up, that means researchers could hand off tasks like literature synthesis, data analysis, or molecule screening to an AI agent rather than doing those steps manually or stitching together separate tools.

The audience at the launch event, pharmaceutical executives and biotech founders, are exactly the buyers who control large research budgets and are already evaluating AI vendors. Anthropic is making a direct pitch to that market rather than waiting for general-purpose Claude to trickle into lab workflows on its own.

The decision to use Claude Science for internal drug research on neglected diseases is notable too. It is a signal that Anthropic believes the product is capable enough to stake real research outcomes on, and it gives the company a reference case outside of commercial partnerships.

Our take

The Claude Code comparison is doing a lot of work in Anthropic’s framing here. Claude Code became credible because developers could test it on real codebases and see concrete output. The same bar applies to Claude Science: the question is whether it can actually reduce time on tasks like designing assays or filtering compound libraries, not just whether it can answer questions about those topics.

The “autonomous” label also warrants scrutiny. In practice, autonomous AI agents in high-stakes research settings still need human review at critical checkpoints. A drug candidate that looks promising to an AI still needs wet-lab validation. Businesses evaluating Claude Science should ask specifically which parts of their pipeline it handles end-to-end and which parts still require a trained researcher to approve the output.

That said, the pricing angle is interesting. Packaging this inside an existing paid Claude subscription rather than launching it as a separate, expensive enterprise tier lowers the barrier to trial for smaller biotech teams who could not justify a dedicated AI research platform budget.

What to do about it

If you are in pharma, biotech, or any research-adjacent business already paying for Claude, log in and look at what Claude Science can access. Run it against a task your team currently does manually, something with a clear, verifiable output, and measure the time difference. That is the only honest way to know whether the autonomous-agent pitch holds up for your specific workflow.

Source: MIT Technology Review