Anthropic’s New Claude Science Workbench Aims to Fix One of Research’s Biggest Pain Points

Anthropic has launched Claude Science, an AI workbench for scientific research now available in public beta. Unlike a new frontier model, Claude Science is a desktop-style application built on existing Claude models that brings together literature search, data analysis, figure generation, and compute management into one environment. Early tests by researchers showed it could map an entire field for $26 and natively render 3D protein structures, all while bundling every result with its full history for reproducibility.

First announced at Anthropic’s event “The Briefing: AI for Science”, Claude Science is an “AI workbench for scientists” designed to consolidate the fragmented tools and datasets researchers juggle daily. It runs on macOS and Linux and is available to subscribers of Claude Pro, Max, Team, and Enterprise plans (Team and Enterprise require admin activation). According to Forbes, the workbench wires in more than sixty scientific databases, and researchers can bring data from disparate sources to work on them in one place.

The product is not a new model but a software layer that covers the full research workflow. It performs literature search and analysis, runs computational workflows directly from a conversation with Claude, manages computing jobs across a laptop, cluster, or rented GPUs, and generates figures, charts, and graphs that are tied to the code that produced them. It also drafts and refines manuscripts until publication-ready. For domain-specific work, Claude Science ships with pre-built skills for genomics, proteomics, structural biology, and chemistry and can natively render 3D protein structures, chemical diagrams, and genome browser tracks.

[!insight: key takeaway] Claude Science is a bet that scientists want a single environment to manage the entire research lifecycle — from hypothesis and literature search to analysis, visualization, and writing — with full auditability built in.

A major theme in both Anthropic’s messaging and early press coverage is auditability and reproducibility. Every figure generated in Claude Science carries its full history — the code, the computing environment, and the conversation — bundled so the result can be regenerated later. Yahoo Finance reports that Anthropic stresses results “are all auditable, allowing researchers to validate and rerun them.”

Anthropic is partnering with Modal to give Claude Science scalable remote compute. Users can connect the workbench to a Modal workspace, and compute workloads — including data processing pipelines, structure prediction models, and molecule design campaigns — run automatically. Modal offers $30 in free compute for new accounts.

This launch sits within Anthropic’s broader “AI for Science” strategy. The company previously launched Claude for Life Sciences, a suite of connectors and specialized tools for discovery and compliance workflows, and runs an “AI for Science Program” offering free API credits to academic and nonprofit researchers. Claude Science is the workbench that pulls those AI models, connectors, and compute capabilities into a single day-to-day environment.

Forbes frames Claude Science as an attempt to “do for laboratory research what Claude Code did for software,” referencing Anthropic’s earlier developer tool. The comparison highlights a strategic pattern: build vertical-specific workbenches on top of the same underlying models to address distinct professional workflows.

[!insight: key takeaway] By bundling reproducibility tools directly into the interface, Claude Science could help address the replicability crisis in computational research — but adoption will depend on whether scientists trust a black-box AI to track their work.

Yahoo Finance explicitly links the Claude Science launch to Anthropic’s business strategy: “Anthropic launches Claude Science in bid to expand revenue streams ahead of IPO.” The product is a paid offering layered on existing subscriptions, targeting scientists and research organizations. It is part of Anthropic’s effort to grow recurring revenue from specialized vertical tools, similar to the Claude Code product for developers.

For researchers and organizations, the practical benefit is integration. Instead of switching between literature databases, coding environments, plotting libraries, and word processors, Claude Science consolidates these into a single session. Early adopter and Forbes contributor John Drake reported that the tool “wired into more than sixty scientific databases,” rendered protein structures and chemical diagrams inline, and managed compute across different backends. Another LinkedIn user, Chris Reid, described the experience as “literature synthesis, analysis, figure generation, and manuscript drafting happening in a single session, reducing the need for browser tabs and external tools.”

[!insight: practical takeaway] For life sciences organizations and academic labs, Claude Science could reduce the overhead of tool-switching and improve reproducibility. However, it requires buy-in at the Team or Enterprise level and a willingness to route compute through Anthropic’s ecosystem. The $30 Modal credit lowers the barrier for initial testing.

Claude Science is in public beta, and Anthropic’s Scott White has invited users to “try it, tell us where it breaks.” How quickly the tool gains adoption in academic labs and pharmaceutical R&D will be a key signal. If successful, Anthropic may expand the workbench model to other scientific domains — or even to non-science verticals — as it builds a portfolio of specialized AI products ahead of a potential IPO.