Platform	Search	Fetch	Advanced	Rate Limit
arXiv	✅	✅	Query syntax	3 req/s
Web of Science	✅	✅	WQL support	API key required
PubMed	✅	✅	Boolean ops	3 req/s
Google Scholar	✅	❌	Auto-retry	IP-based
Semantic Scholar	✅	✅	GraphQL	100 req/5min
Sci-Hub	❌	✅	Mirror mgmt	Legal caution ⚠️
ScienceDirect	✅	✅	Facets	API key required
Springer Nature	✅	✅	Dual API	API key required
Wiley	✅	✅	Full metadata	API key required
Scopus	✅	✅	Advanced	API key required
Crossref	✅	✅	Works API	Open access
bioRxiv	✅	✅	Pre-prints	Open access
medRxiv	✅	✅	Medical pre-prints	Open access
IACR ePrint	✅	✅	Cryptography	Open access

Prerequisites

Node.js 18+
Globally installed: npm install -g paper-search-mcp-nodejs
API keys for premium platforms (optional)

Quick Start

Add to Claude Code (MCP Integration)

# The package is already globally installed via:
# npm install -g paper-search-mcp-nodejs

# Add MCP server to Claude Code
claude mcp add paper-search npx paper-search-mcp-nodejs

# Verify installation
npx paper-search-mcp-nodejs --version  # Should show v0.2.5

Configure API Keys

Create .env file in your working directory:

# Optional: Premium platform API keys
WOS_API_KEY=your_web_of_science_key
PUBMED_API_KEY=your_pubmed_key
SCIENCEDIRECT_API_KEY=your_elsevier_key
SPRINGER_API_KEY=your_springer_key
WILEY_API_KEY=your_wiley_key
SCOPUS_API_KEY=your_scopus_key

# Optional: Sci-Hub mirror override
SCIHUB_MIRROR=https://sci-hub.se

Available MCP Tools (17 Total)

Universal Search Tools

1. `search_papers`

Search across any supported platform with unified interface.

// Example: Search arXiv
{
  "platform": "arxiv",
  "query": "quantum computing",
  "maxResults": 10
}

// Example: Search PubMed with filters
{
  "platform": "pubmed",
  "query": "CRISPR gene editing",
  "filters": {
    "year": "2024",
    "species": "human"
  },
  "maxResults": 20
}

// Example: Web of Science advanced search
{
  "platform": "wos",
  "query": "TS=(machine learning) AND PY=(2023-2024)",
  "filters": {
    "documentType": "article"
  }
}

2. `fetch_paper`

Retrieve full paper metadata or PDF by identifier.

// By DOI
{
  "doi": "10.1038/s41586-019-1666-5",
  "platform": "crossref"
}

// By arXiv ID
{
  "arxivId": "2301.07041",
  "platform": "arxiv"
}

// By PubMed ID
{
  "pmid": "12345678",
  "platform": "pubmed"
}

Platform-Specific Tools

3. `search_arxiv`

Advanced arXiv search with category filtering.

{
  "query": "ti:transformer AND cat:cs.LG",
  "maxResults": 50,
  "sortBy": "submittedDate",
  "sortOrder": "descending"
}

4. `search_pubmed`

PubMed/MEDLINE search with MeSH terms.

{
  "query": "(cancer[MeSH]) AND (immunotherapy[Title])",
  "filters": {
    "publicationType": "Clinical Trial",
    "minDate": "2020/01/01",
    "maxDate": "2024/12/31"
  }
}

5. `search_google_scholar`

Google Scholar search with citation data.

{
  "query": "deep learning natural language processing",
  "filters": {
    "yearLow": 2020,
    "yearHigh": 2024
  }
}

⚠️ Rate Limit Warning: Google Scholar enforces strict IP-based limits. Use sparingly.

6. `search_semantic_scholar`

Semantic Scholar with citation graphs.

{
  "query": "attention is all you need",
  "fields": ["title", "authors", "citationCount", "influentialCitationCount"],
  "limit": 100
}

7. `search_wos` (Web of Science)

Advanced WoS search with citation analytics.

{
  "query": "TS=(climate change) AND CU=(China)",
  "filters": {
    "databaseId": "WOS",
    "timespan": "2020-2024"
  }
}

8. `search_sciencedirect`

Elsevier ScienceDirect full-text search.

{
  "query": "CRISPR Cas9",
  "filters": {
    "date": "2024",
    "contentType": "Journal",
    "openAccess": true
  }
}

9. `search_springer`

Springer Nature search across journals and books.

{
  "query": "quantum cryptography",
  "filters": {
    "subject": "Physics",
    "year": "2024"
  }
}

10. `search_wiley`

Wiley Online Library search.

{
  "query": "protein folding",
  "filters": {
    "publicationYear": "2024"
  }
}

11. `search_scopus`

Scopus citation database search.

{
  "query": "TITLE-ABS-KEY(machine learning healthcare)",
  "filters": {
    "doctype": "ar",  // Article
    "pubyear": "2024"
  }
}

12. `search_crossref`

Crossref DOI search and metadata.

{
  "query": "neural networks",
  "filters": {
    "type": "journal-article",
    "from-pub-date": "2024-01-01"
  },
  "rows": 100
}

13. `search_biorxiv`

bioRxiv preprint search (biology).

{
  "query": "COVID-19 vaccine",
  "filters": {
    "category": "immunology"
  }
}

14. `search_medrxiv`

medRxiv preprint search (medicine).

{
  "query": "cancer therapy",
  "maxResults": 50
}

15. `search_iacr`

IACR ePrint archive (cryptography).

{
  "query": "zero knowledge proof",
  "maxResults": 30
}

Utility Tools

16. `fetch_scihub_pdf`

Download PDF via Sci-Hub (legal caution required).

{
  "doi": "10.1016/j.cell.2024.01.001"
}

⚠️ Legal Warning: Sci-Hub access may violate copyright laws in many jurisdictions. Use only for:

Papers you have legal access to
Fair use/educational purposes where permitted
Jurisdictions where Sci-Hub is legal

17. `get_paper_recommendations`

Get citation-based paper recommendations.

{
  "doi": "10.1038/nature12373",
  "platform": "semantic_scholar",
  "limit": 10
}

Security Features (v0.2.5)

1. DOI Validation

Automatic validation of DOI formats to prevent injection attacks.

2. Query Sanitization

SQL injection and command injection protection for all queries.

3. Sensitive Data Masking

API keys automatically masked in logs and error messages.

4. Rate Limiting

Built-in token bucket rate limiting for all platforms.

Common Workflows

Literature Review

// 1. Broad search across multiple platforms
await search_papers({
  platform: "semantic_scholar",
  query: "transformer models NLP",
  maxResults: 100
});

// 2. Filter by citations
const highImpact = results.filter(p => p.citationCount > 100);

// 3. Get related papers
for (const paper of highImpact) {
  await get_paper_recommendations({
    doi: paper.doi,
    limit: 5
  });
}

Systematic Review

// Search across multiple databases
const platforms = ["pubmed", "wos", "scopus", "crossref"];

for (const platform of platforms) {
  await search_papers({
    platform,
    query: "(systematic review) AND (machine learning diagnosis)",
    filters: { year: "2020-2024" }
  });
}

Citation Analysis

// Get paper with citations
const paper = await search_semantic_scholar({
  query: "BERT: Pre-training of Deep Bidirectional Transformers",
  fields: ["citationCount", "influentialCitationCount", "references", "citations"]
});

// Analyze citation network
console.log(`Total citations: ${paper.citationCount}`);
console.log(`Influential citations: ${paper.influentialCitationCount}`);

PDF Retrieval

// Try open access first
let pdf = await fetch_paper({
  doi: "10.1038/nature12373",
  platform: "crossref"
});

// Fallback to Sci-Hub if needed (legal caution)
if (!pdf.openAccess) {
  pdf = await fetch_scihub_pdf({
    doi: "10.1038/nature12373"
  });
}

Platform-Specific Advanced Features

Springer Dual API Support

Springer supports both REST and GraphQL APIs:

// REST API (default)
await search_springer({
  query: "quantum computing",
  api: "rest"
});

// GraphQL API (more flexible)
await search_springer({
  query: "quantum computing",
  api: "graphql",
  fields: ["title", "creators", "doi", "abstract"]
});

Web of Science Advanced Search (WQL)

// Topic + Author + Year
await search_wos({
  query: "TS=(deep learning) AND AU=(LeCun) AND PY=(2015-2024)"
});

// Citation count filter
await search_wos({
  query: "TS=(CRISPR) AND TC >= 100"  // 100+ citations
});

Sci-Hub Mirror Management

// Check mirror status
const mirrors = [
  "https://sci-hub.se",
  "https://sci-hub.st",
  "https://sci-hub.ru"
];

// Tool automatically tries mirrors in order
// Override via SCIHUB_MIRROR env variable

Rate Limits and Best Practices

Platform	Limit	Recommendation
arXiv	3 req/s	Batch queries, cache results
PubMed	3 req/s	Use E-utilities efficiently
Google Scholar	IP-based	Minimize usage, add delays
Semantic Scholar	100 req/5min	Use GraphQL for efficiency
Web of Science	API key dependent	Check your plan limits
Crossref	No strict limit	Be respectful, ~50 req/s max
Sci-Hub	Varies by mirror	Use sparingly, legal caution

Best Practices:

Cache results: Store search results locally to avoid re-querying
Batch operations: Group related queries together
Use filters: Narrow searches to reduce API calls
Respect rate limits: Add delays between requests
Check open access: Try open platforms before premium APIs
API key management: Store keys in .env, never commit to git

Data Model

All platforms return standardized Paper objects:

interface Paper {
  // Identifiers
  doi?: string;
  arxivId?: string;
  pmid?: string;
  pmcid?: string;

  // Core metadata
  title: string;
  authors: Author[];
  abstract?: string;

  // Publication info
  journal?: string;
  volume?: string;
  issue?: string;
  pages?: string;
  year?: number;
  publishedDate?: string;

  // Access
  pdfUrl?: string;
  openAccess?: boolean;

  // Metrics
  citationCount?: number;
  influentialCitationCount?: number;

  // Platform-specific
  platform: string;
  sourceData?: Record<string, any>;
}

Error Handling

try {
  const results = await search_papers({
    platform: "arxiv",
    query: "quantum computing"
  });
} catch (error) {
  if (error.code === 'RATE_LIMIT_EXCEEDED') {
    // Wait and retry
    await sleep(5000);
    retry();
  } else if (error.code === 'INVALID_API_KEY') {
    // Check API key configuration
    console.error('API key invalid or missing');
  } else if (error.code === 'PLATFORM_UNAVAILABLE') {
    // Try alternative platform
    await search_papers({ platform: "crossref", query });
  }
}

Troubleshooting

Issue: API key not recognized

# Check environment variables
echo $WOS_API_KEY
echo $PUBMED_API_KEY

# Verify .env file location (should be in working directory)
ls -la .env

# Test API key directly
curl -H "Authorization: Bearer $WOS_API_KEY" "https://api.clarivate.com/..."

Issue: Rate limit exceeded

# Add delays between requests
# Use batch operations
# Cache results locally
# Consider upgrading API plan

Issue: Google Scholar blocking

# Google Scholar is IP-sensitive:
# - Reduce request frequency
# - Use rotating proxies (if permitted)
# - Consider Semantic Scholar as alternative
# - Wait 24-48 hours if blocked

Issue: Sci-Hub mirror down

# Tool automatically tries multiple mirrors
# Override with specific mirror:
export SCIHUB_MIRROR=https://sci-hub.st

# Check mirror status manually
curl -I https://sci-hub.se

Issue: Empty results

# Check query syntax for platform
# Verify filters are not too restrictive
# Try broader search terms
# Check if platform is accessible (API key, network)

Compliance and Legal Notice

Sci-Hub Usage

⚠️ Important: Sci-Hub access may violate copyright laws in many countries. Use responsibly:

Only for papers you have legal access to
Educational/fair use purposes where permitted
Check local laws and institutional policies
Consider legal alternatives first (open access, institutional access)

Google Scholar Terms

⚠️ Important: Google Scholar prohibits automated scraping:

Use minimally and responsibly
Add delays between requests
Consider official APIs or Semantic Scholar as alternatives
Respect robots.txt and rate limits

API Terms of Service

Each platform has its own terms:

Web of Science: Check your license agreement
Scopus: Institutional access required
Elsevier/ScienceDirect: API agreement needed
Springer Nature: Check API usage limits

Additional Resources

GitHub: https://github.com/your-repo/paper-search-mcp-nodejs
Documentation: Full API docs in package README
npm Package: https://www.npmjs.com/package/paper-search-mcp-nodejs
Version: v0.2.5 (latest)
MCP Protocol: https://modelcontextprotocol.io/

Updating the Tool

# Update to latest version
npm update -g paper-search-mcp-nodejs

# Reinstall MCP server (if needed)
claude mcp remove paper-search
claude mcp add paper-search npx paper-search-mcp-nodejs

# Verify version
npx paper-search-mcp-nodejs --version

Related Skills

agentdb-vector-search: For semantic search over downloaded papers
reasoningbank-intelligence: For analyzing research patterns
github-code-review: For reviewing code in academic software

Support

For issues or questions:

Check the troubleshooting section
Review platform-specific documentation
Check GitHub issues
Verify API keys and configuration
Test with minimal examples

Pro Tips:

Start with open platforms (arXiv, Crossref, bioRxiv) - no API keys needed
Use Semantic Scholar for citation analysis - excellent free API
Cache results locally to minimize API calls
Combine multiple platforms for comprehensive coverage
Always check open access status before using Sci-Hub

Academic Researcher

Academic Researcher

What This Skill Does

Supported Platforms