Scinaut · AI Materials Science Workstation

AI Research Paper Analysis — Read, Summarize, Discover
Upload papers and get instant AI summaries, knowledge Q&A, and materials science computation. Built for researchers.

📤Upload

📄Read Papers

🧪Design Experiments

📝Publish

📄 Literature Summary💬 Knowledge Q&A🔬 Critical Analysis

Analyze →Get Started

⚡ Interactive Demo

See it in action.

Paste any paper abstract or experiment protocol — AI analyzes it live.

Scinaut Lab · Live Demo

Input

Try:

AI Analysis

🧪

Paste text and click Analyze to see results

Scinaut Lab · Demo Mode

Summary Modes

Three ways to understand a paper

From quick extraction to deep critique — choose the lens that fits your research workflow.

Quick Overview

Core Extraction

Goals, methods, results, conclusions structured in seconds.

Structured output
Instant results

Deep Dive

Critical Analysis

Evaluate methodology, detect biases, assess evidence quality.

Bias detection
Evidence grading

Review

Study Cards

10–15 Q&A cards for exam prep and quick concept review.

Q&A format
10–15 cards

Context

Follow-up Q&A

Ask anything about the paper — AI answers with citations.

Source citations
Multi-turn

Not Just Search

Answers. Not links.

Regular RAG gives you a pile of relevant snippets to piece together. Our synthesis layer delivers a complete, cited, self-questioning answer — like a review written by a principal scientist.

Trust

Every sentence has a source

Factual claims are followed by clickable citation badges [1][3]. Click to jump to the source document. Expert judgments marked [expert], uncertainty marked [?]. No black box — every claim is traceable.

Inline citation badges
Click-to-source
Expert/uncertainty markers

Critical

Tells you what your paper doesn't

AI proactively identifies what's MISSING from your references — synthesis temperatures, characterization data, long-term stability. And tells you what to look for next. Search engines can't do this.

Knowledge gap detection
Severity ratings
Remediation suggestions

Expert

Polymer? Ceramic? Semiconductor? It knows.

No domain assumptions. Upload a perovskite paper → photovoltaic framework. Upload a MOF paper → catalysis framework. Won't force battery logic onto non-battery materials. 40 years of expertise across all materials disciplines.

10+ materials disciplines
Auto domain detection
Zero preset frameworks

Preview: a real knowledge Q&A

What material was studied and how was it synthesized?

## Expert Analysis

This work employs in-situ synthesis to prepare silicate-coated iron hydroxyphosphate for Cd(II) removal. Synthesis parameters estimated from similar literature: iron source FeCl₃·6H₂O (0.1–0.5 M), phosphorus source H₃PO₄...12expert

🔍 Knowledge Gaps

🟡Synthesis temperature, pH, stirring speed not reported — section "2.2 Material preparation" lacks critical parameters
🟡No control experiment with uncoated samples — cannot quantify coating benefit
🟢Tested only in deionized water — real wastewater co-ions (Ca²⁺, Mg²⁺) would severely interfere

📚 References

1S1385894724105281-main.pdf › Body

Try it in workspace →

NEW — Materials Science Engine

Not Just Another AI Chat

Built-in world-class Materials Science & Engineering AI院士 system. 40 years of experience distilled into diagnostic rules, multi-scale analysis frameworks, and real database queries — every answer is sourced, every number has a method.

🧬

Materials Database

Direct query to Materials Project, AFLOW, OQMD. Search by formula, elements, or properties. Get crystal structures, band gaps, stability. Local caching avoids repeated API calls.

100,000+ materials · instant query

⚡

Computation Engine

Battery voltage/capacity/energy density. Phase diagrams. Defect formation energies. Diffusion barriers. Dopant screening. Pourbaix diagrams. No DFT installation needed.

6 compute modes · sub-second response

🕸️

Knowledge Graph

Material → Process → Property → Failure relationship graph. Seeded with Li-ion, solid electrolytes, perovskite PV, electrocatalysis. Query in natural language.

50+ nodes · 60+ edges · growing

Capability Matrix8 capabilities · always on

·🔬Multi-Scale AnalysisAtomic → nano → micro → macro all at once. Not dimension by dimension.·💎Crystal StructureSymmetry, surface slabs, XRD simulation, coordination, diffusion channels.·📐Battery EstimationTheoretical capacity, voltage, energy density. Coin-cell data skepticism built in.·⚠️Failure DiagnosisSymptom → root cause mapping. 7 battery, 7 PV, 7 catalysis failure modes.·🔬Multi-Scale AnalysisAtomic → nano → micro → macro all at once. Not dimension by dimension.·💎Crystal StructureSymmetry, surface slabs, XRD simulation, coordination, diffusion channels.·📐Battery EstimationTheoretical capacity, voltage, energy density. Coin-cell data skepticism built in.·⚠️Failure DiagnosisSymptom → root cause mapping. 7 battery, 7 PV, 7 catalysis failure modes.

·🏭IndustrializationCost models, TRL assessment, process windows, manufacturing barriers. No hype.·🧪Experimental DataAuto-parse cycling, EIS, XRD, TGA data. Multi-instrument format support.·📊Dopant ScreeningIonic radius mismatch, solubility prediction, known effects. Auto-ranked candidates.·🌐Molecular ComputationRDKit descriptors, pKa prediction, solvation energy, electrochemical windows.·🏭IndustrializationCost models, TRL assessment, process windows, manufacturing barriers. No hype.·🧪Experimental DataAuto-parse cycling, EIS, XRD, TGA data. Multi-instrument format support.·📊Dopant ScreeningIonic radius mismatch, solubility prediction, known effects. Auto-ranked candidates.·🌐Molecular ComputationRDKit descriptors, pKa prediction, solvation energy, electrochemical windows.

IV.How It Works•3 步 · 持续迭代

How It Works · Nº 01

Open in Your Project、Ask or Compute、Get Verifiable Answers.

每一步都是即时的、视觉化的、零配置的 —— 用可组合的文件，而不是不透明的提示词。

Open in Your Project→

Enter any research project, click the 🔬 Materials Science tab. No setup needed.

📂

Ask or Compute→

Search databases, run battery estimates, query the knowledge graph — or just ask materials questions in Q&A, AI auto-activates materials science skills.

🔬

Get Verifiable Answers→

Every answer cites data sources. Every computation is reproducible. Not LLM fiction. It's computed.

✅

从上传文献到获得可验证结果，三步完成。

research-ai/scinaut·MIT

Open Workspace →→

Comparison

Scinaut vs the alternatives

Purpose-built for research. Not a general chatbot, not a manual slog.

Capability	Scinaut	ChatGPT	Manual Reading
Structured summaries	3 modes built-in	Prompt engineering needed	Hours per paper
Source citations	Every answer cited	Hallucination risk	Manual notes
Critical analysis	Bias + methodology audit	Surface-level only	Requires expertise
Multi-document Q&A	RAG across projects	Context window limits	Cross-ref manually
PDF parsing	Drag & drop	Copy & paste	N/A
🔬 Materials Database	100K+ materials DFT data query	Search web manually	Check handbooks
⚡ Computation Engine	Battery/doping/defect estimation	Install DFT code / can't use	Manual calc
🕸️ Knowledge Graph	Material→property→failure graph	Not available	Memory-based
🧪 Lab Assistant	Protocol analysis + morphology prediction	Generic unverifiable answers	Literature-based
Speed (per paper)	~30 seconds	~2 min (prompting)	30-60 min

FAQ

Frequently Asked Questions

What can Scinaut do?+

Upload research papers and get AI-powered summaries, critical analysis, and interactive Q&A. Built for researchers who need to process papers efficiently.

How is this different from ChatGPT?+

Scinaut has structured summary modes (extraction, analysis, study cards), built-in critical thinking, source citations, and multi-document RAG. ChatGPT requires manual prompt engineering for each of these.

What file formats are supported?+

PDF documents are fully supported. You can also paste text directly for analysis.

Is my data private?+

Yes. Your documents are stored in your own Supabase database. We do not train on your data.

AI Research Paper Analysis — Read, Summarize, DiscoverUpload papers and get instant AI summaries, knowledge Q&A, and materials science computation. Built for researchers.