Computational biology labs
Coordinate genomics, single-cell, proteomics, structural biology, and cheminformatics tools from one research session.
Independent tool overview
Claude Science at a glance
Claude Science is Anthropic's beta research workbench for running scientific analysis, connecting databases and lab tools, managing compute, and producing versioned artifacts with provenance.
Visit the official Claude Science site ↗
- Product type
- Desktop scientific research workbench
- Status
- Public beta
- Platforms
- macOS 13+ and Linux x64
- Eligible plans
- Pro, Max, Team, and Enterprise
- Scientific access
- 60+ databases and domain-specific models
- Compute routes
- Local, SSH/Slurm HPC, cloud VM, or Modal
- Workflow languages
- Python, R, and shell
- Organization setup
- Team and Enterprise admin must enable it
- Data boundary
- Raw data may stay local; prompt content reaches Anthropic
- Last reviewed
- August 29, 2026
Overview
What Claude Science is
Claude Science is a desktop research environment that pairs Claude with scientific tools, code execution, connectors, specialist agents, and compute orchestration. It is an app built around the Claude models included with a subscription, not a separate scientific foundation model.
Researchers can ask it to search literature and databases, inspect files, write and run Python, R, and shell workflows, build figures and manuscripts, and coordinate multi-step analyses. Its strongest initial coverage is in life-science workflows such as genomics, single-cell analysis, proteomics, structural biology, and cheminformatics.
The app runs on macOS or Linux wherever the data and compute live: a laptop, lab workstation, cloud VM, or HPC login node. It can submit work to a Slurm cluster over SSH or use a researcher's Modal account, asking before reaching new resources and allowing prior decisions to be reviewed or revoked.
Claude Science saves figures, tables, and other results as versioned artifacts with the code, environment, plain-language explanation, and conversation that produced them. A reviewer agent checks citations, calculations, provenance, and consistency between figures and underlying code, but this does not replace independent reproduction, peer review, or domain-expert sign-off.
Raw datasets and compute can remain on the researcher's infrastructure, but content placed into prompts and model responses is processed by Anthropic under the account's applicable retention terms. Labs handling patient, controlled, unpublished, export-restricted, or commercially sensitive data should approve the exact plan, connectors, compute path, retention, and access controls before use.
Use cases
Who Claude Science is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Literature-heavy research
Search and synthesize many papers while preserving citations, extracted evidence, and reviewer checks.
Labs with established pipelines
Read, run, and extend existing Python, R, and shell scripts without rebuilding validated methods around a new interface.
HPC and GPU workflows
Plan and submit long analyses to local machines, Slurm clusters, cloud VMs, or Modal compute.
Reproducible figure generation
Keep a figure's code, environment, inputs, explanations, revisions, and conversation attached to the artifact.
Custom scientific agent systems
Create reusable skills, connectors, and specialist agents around a lab's own models, databases, ELNs, and internal APIs.
Capabilities
Core Claude Science features
Generalist coordinating agent
Plans research tasks, delegates work to specialists, integrates results, and maintains context across a running session.
Scientific specialist agents
Provides domain-oriented capabilities for genomics, single-cell, proteomics, structures, chemistry, and related analysis.
Database and model connectors
Connects to more than 60 scientific databases and domain models, including life-science resources and BioNeMo tools.
Sandboxed code execution
Writes and runs analysis code while limiting file access to folders and resources the researcher provides.
Existing-pipeline support
Reads, runs, and builds on established Python, R, and shell workflows rather than requiring a proprietary pipeline format.
Compute orchestration
Builds environments and submits work to local compute, SSH-accessible Slurm clusters, cloud machines, or Modal.
Native scientific renderers
Displays proteins, molecular structures, genome tracks, chemical structures, figures, tables, and manuscripts in the workbench.
Versioned artifacts
Preserves multiple states of figures and manuscripts so researchers can compare, fork, restore, and refine results.
Full provenance
Attaches code, environment, description, and conversation history to generated artifacts for later inspection and reproduction.
Reviewer agent
Checks citations, calculations, traceability, and agreement between presented results and the code beneath them.
Session forking
Branches a running research session to compare methods without losing the original analysis path.
Reusable skills and connectors
Turns trusted lab tools and procedures into capabilities that can be inherited by later sessions.
Process
How the Claude Science workflow works
Step 1
Approve the data class and plan
Decide what data is permitted, select the needed retention and compliance terms, and enable Claude Science through an organization admin where required.
Step 2
Install near the data
Run the app on a supported Mac, Linux workstation, HPC login node, or cloud VM where the files and compute already reside.
Step 3
Grant narrow file access
Expose only the project folders, scripts, databases, credentials, and connectors needed for the specific study.
Step 4
Define the scientific question
State the hypothesis, dataset, accepted methods, exclusion criteria, endpoints, statistical assumptions, and definition of a valid result.
Step 5
Review the proposed plan
Inspect data transformations, package choices, resource requests, external calls, and cluster jobs before execution.
Step 6
Run against trusted tools
Prefer the lab's validated pipelines, reference databases, containers, and environment locks over newly invented analysis code.
Step 7
Inspect every artifact
Trace claims back to code and inputs, verify citations and calculations, and compare plots with raw and intermediate data.
Step 8
Reproduce independently
Rerun the analysis from a clean environment or by a second researcher and compare outputs before relying on the result.
Step 9
Record human approval
Keep domain experts, statisticians, safety officers, and principal investigators accountable for interpretation, publication, and downstream action.
Cost
Claude Science pricing and free plan
Claude Science is included with paid Claude plans rather than sold as a separate add-on. Compute used through a lab, cloud provider, HPC system, or Modal can cost extra. Anthropic also operates a capacity-limited discounted Team program for verified academic and nonprofit labs.
Pro
$20/month or $200/year
The lowest generally available individual plan that includes Claude Science.
- $17 effective monthly rate when $200 is billed annually
- Claude Science beta included
- Usage limits apply
- Includes Claude Code, Cowork, Design, projects, and Research
Max
From $100/month
For individual researchers who need substantially more Claude usage.
- Includes all Pro features
- 5x or 20x usage options
- Higher output limits
- Priority access at busy times
- Compute-provider charges remain separate
Team
$20-$125/seat/month
For labs and organizations needing shared administration, SSO, and connector controls.
- Standard: $20 annually or $25 monthly per seat
- Premium: $100 annually or $125 monthly per seat
- 2 to 150 users
- Claude Science must be enabled by an owner
- Content is not used for model training by default
Enterprise
$20/seat plus usage
For organizations that need advanced identity, retention, network, audit, or regulated-data options.
- Usage billed at API rates
- SSO, SCIM, custom roles, audit logs, and spend controls
- Custom retention and network controls
- HIPAA-ready offering is available
- Sales-assisted terms may differ
Team plan for scientists
$0 standard or $15 premium
A one-year program for eligible academic and nonprofit labs within the announced 10,000-seat allocation.
- Principal investigator or equivalent must verify eligibility
- Standard seats are free
- Premium seats provide 5x usage for $15 per month
- Availability is capacity-limited
- For-profit R&D teams use standard Team or Enterprise plans
Pricing checked . Check current pricing at the source ↗
Assessment
Claude Science strengths and limitations
Where it stands out
- Combines scientific databases, code, local files, compute, artifacts, and conversation in one research environment
- Works with established Python, R, shell, ELN, API, and pipeline investments
- Raw datasets can remain on lab-controlled infrastructure
- Local, cluster, and on-demand compute routes support both small and large analyses
- Artifact provenance makes AI-assisted results easier to inspect and reproduce
- Reviewer agent adds a systematic check for citations, calculations, and untraceable claims
- Specialists and connectors provide unusually deep life-science coverage out of the box
- Academic and nonprofit lab program can make team access substantially cheaper for eligible researchers
What to consider
- Claude Science is still beta and its interfaces, limits, and behavior can change
- There is no documented Windows version; current support is macOS 13+ and Linux x64
- Raw datasets may stay local, but prompt and response content is still processed by Anthropic
- The reviewer agent can miss incorrect methods, fabricated citations, statistical errors, or subtle scientific misinterpretation
- Paid Claude usage limits and separate compute costs can constrain long or multi-agent analyses
- Life sciences receive the richest preconfigured coverage, while other disciplines may need more custom setup
- Connecting SSH, HPC, internal APIs, ELNs, and proprietary pipelines expands the permission and credential attack surface
- AI-generated analyses, hypotheses, figures, and manuscripts still require independent reproduction and expert accountability
Compare
Claude Science alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Science
SciSpace BioMed Agent
A more narrowly biomedical research agent for literature, multi-omics, phenotype, and protocol reasoning.
Explore SciSpace BioMed Agent →Data Analysis
Elicit
A simpler option focused on systematic literature search, paper screening, extraction, and evidence synthesis.
Explore Elicit →Students
Gemini Notebook (formerly NotebookLM)
A lower-complexity source-grounded research notebook for understanding and discussing an uploaded document collection.
Explore Gemini Notebook (formerly NotebookLM) →Questions
Claude Science FAQs
What is Claude Science?
Claude Science is Anthropic's beta desktop workbench for scientific research. It combines Claude with code execution, scientific databases, specialist agents, lab connectors, compute orchestration, and reproducible artifacts.
Is Claude Science a separate AI model?
No. It is an application that uses the Claude models available on the user's plan and adds scientific tools, connectors, environments, agents, renderers, compute access, and provenance.
Who can use Claude Science?
It is available on Pro, Max, Team, and Enterprise plans. Team and Enterprise owners must enable it first. The app currently supports macOS 13 or later and Linux x64.
How much does Claude Science cost?
There is no separate Claude Science add-on price. The lowest general-access plan is Pro at $20 monthly or $200 annually. Max, Team, and Enterprise provide additional usage and administration. External compute may cost extra.
Does research data stay private?
The app can keep raw datasets and computation on the researcher's infrastructure, but any content included in prompts and model responses is processed by Anthropic under the account's retention terms. Labs should approve the exact data path before use.
Can Claude Science run on an HPC cluster?
Yes. Anthropic describes installing it on an HPC login node or connecting to Slurm compute over SSH. It can also run locally, on a cloud VM, or through a researcher's Modal account.
What does provenance mean in Claude Science?
Generated artifacts retain the code, execution environment, plain-language description, and conversation that produced them. This makes the analysis easier to inspect, rerun, edit, and defend later.
Does the reviewer agent guarantee scientific accuracy?
No. It can flag citation, calculation, provenance, and code-to-figure issues, but it does not replace independent replication, statistical review, peer review, ethics review, or accountable domain experts.
Bottom line
Our Claude Science verdict
Claude Science is one of the most complete AI research workbenches for computational labs that want a single agentic layer over literature, code, scientific tools, local data, and cluster compute. Its provenance model is genuinely useful, but beta software and automated review do not make results valid by default. The best deployments will pair narrow permissions and approved pipelines with independent reproduction and named human owners.
Visit Claude Science website ↗Similar Science Tools
View all Science AI tools →
TRIBE v2
TRIBE v2 - Meta's predictive foundation model that simulates human brain responses to sights, sounds, and language
Leanstral 1.5
Leanstral 1.5 - Mistral's open model for writing verified math proofs
Gemini 3.1 Deep Think
Gemini 3 Deep Think - Google's upgraded AI reasoning mode now solving PhD-level math and science research problems
Kosmos
Kosmos - Next-generation AI scientist
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.