Claude Science is an AI workbench for scientific research. A "workbench" here means a single space that gathers the tools you need — the way a lab bench keeps your instruments in one place. Literature review, data analysis, drafting figures and manuscripts, and even large-scale computation: Claude Science pulls a research workflow that used to be scattered across many windows into one app. This guide explains what Claude Science is, what it does, and how (and on which plans) to get started — based on official sources.
🟢 Current lineup (September 2026): Claude Opus 5.5 / Claude Sonnet 5 / Claude Haiku 4.5 (higher tier: Claude Fable 5.1). · model notice · Fable subscription
Fable 5 and 5.1 subscription (updated September 7, 2026): Claude Fable 5.1, released September 1, 2026, is the current Fable model and Fable 5 is now legacy. Plan terms are the same for both — Max and Team Premium plans include Fable at up to 50% of the weekly usage limit; Pro and Team Standard use usage credits (
What is Claude Science?
The first thing to know is that Claude Science is an app, not a new AI "model." Anthropic did not build a separate science-only model; the app runs on top of the existing Claude models included in your plan (for example, Claude Opus 5). It was officially released on June 30, 2026 as a beta (a pre-release test version).
Researchers typically juggle dozens of databases, inconsistent file formats, and tools like PubMed, Jupyter, R, and cluster terminals. Claude Science brings these fragmented tools into a single research environment, so you can move from literature analysis to multi-step research to refining figures and manuscripts in one place.
Four core features
1. A unified environment with 60+ connectors
You talk to one coordinating agent that has access to 60+ skills and connectors pre-configured for genomics, single-cell, proteomics, structural biology, cheminformatics, and more. A "connector" is the link that lets Claude reach an external data source or tool. Examples include PubMed, Benchling, 10x Genomics, BioRender, bioRxiv/medRxiv, and ClinicalTrials.gov — you can query them in natural language and get answers with source citations.
2. A reviewer agent that catches errors
As the work runs, a reviewer agent inspects the outputs — flagging incorrect citations, numbers that can't be traced, and figures that don't match their code, and correcting them as it goes. It acts as an extra filter for mistakes that are easy to miss.
3. Reproducible artifacts
Science has to be verifiable (this is called reproducibility). When Claude Science creates a figure or manuscript, it records the code, environment, plain-language description, and full conversation history behind it — so months later you can still check whether a chart truly came from the data. It natively renders 3D protein structures, genome browser tracks, and chemical structures, and you can edit figures in plain language (for example, "switch the axis to log scale") — the agent edits its own code.
4. Compute that scales safely on your own infrastructure
Large analyses — like protein folding or a genomics pipeline over a massive dataset — need more than a laptop. Claude Science drafts a plan, asks for your approval before reaching new resources, and submits jobs to the compute your lab already uses (your own HPC cluster over SSH, or your Modal account), scaling from a single GPU to hundreds. HPC (High-Performance Computing) means many powerful machines working together on big calculations. Importantly, the app runs on your own machine (laptop, Linux box, or HPC login node), so large or sensitive datasets never leave the systems they're already on.
Which plans and platforms?
- Plans: available in beta on Claude Pro, Max, Team, and Enterprise. Team and Enterprise admins must enable Claude Science first.
- OS: a desktop app for macOS 13 or later and Linux (x64).
- Cost: no separate API token pricing — usage counts against your Claude plan.
- Academic/nonprofit discount: a discounted Team plan is available for active labs at academic and nonprofit research organizations (verified via the principal investigator).
Getting started
Per the official guide:
- Sign in with a claude.ai account on a Pro, Max, Team, or Enterprise plan (Team/Enterprise require admin enablement first).
- Install the app on macOS 13+ or Linux (x64).
- After sign-in, a setup wizard walks you through enabling connectors and skills. Manage connectors later under Customize → Connectors — featured connectors are on by default; partner connectors are added from the connector list.
AI for Science program
Anthropic is supporting up to 50 projects with up to $30,000 in credits each, and compute partner Modal is providing up to $2,000 in compute for select projects, with an early focus on biology and biomedical research. Per the official announcement, applications are open through July 15, 2026, award notifications go out by July 31, and projects run September 1 to December 1, 2026. (Dates are per the official announcement and may change — please re-check the official page before applying.)
Who is it for, and what to keep in mind
Claude Science is especially useful for researchers doing life-sciences and computational work — genomics, single-cell RNA sequencing, protein structure prediction, cheminformatics, and the like. Note that it is currently in beta and life-sciences-focused, so features and coverage may keep changing. Check the official documentation for exact requirements, remote-server setup, and troubleshooting.
FAQ
Is Claude Science a new AI model?
No. It is an app that runs on top of existing Claude models.
Can I use it for free?
It is not available on the free plan. It is available in beta on paid plans (Pro, Max, Team, Enterprise), with no separate charge — usage counts against your plan.
Does it work on Windows?
Per the official guidance it supports macOS 13+ and Linux (x64). Windows support is not confirmed in official sources.
Is my research data safe?
Because the app runs on your own machine, it is designed so that large or sensitive data never leaves the systems it is already on, and it asks for approval before reaching new resources.
This guide is based on Anthropic's official announcement (anthropic.com) and Claude's official getting-started guide (claude.com). Claude Science is in beta and details may change — check the official pages for the latest.
Related articles
- How to Write Prompts for Claude: 6 Core Principles from the Official Docs
- What Is Claude Cowork? Claude's File-Working Desktop Mode
- What Is Claude Design? Anthropic's Chat-to-Canvas Design Tool
- Claude Memory — How to Make Claude Remember and Manage It
- Claude Styles — Lock Tone, Format, and Length Your Way