Scientific Paper Harvester MCP Server

A comprehensive Model Context Protocol (MCP) server that provides LLMs with real-time access to scientific papers from 6 major academic sources: arXiv, OpenAlex, PMC (PubMed Central), Europe PMC, bioRxiv/medRxiv, and CORE.

🚀 Features

Comprehensive Source Coverage

• arXiv: Computer science, physics, mathematics preprints and papers
• OpenAlex: Open catalog of scholarly papers with citation data
• PMC: PubMed Central biomedical and life science literature
• Europe PMC: European life science literature database
• bioRxiv/medRxiv: Biology and medical preprint servers
• CORE: World's largest collection of open access research papers

Advanced Capabilities

• Paper Fetching: Get latest papers from any source by category/concept
• Paper Search: Search papers by title, abstract, author, or full-text across 4 major sources
• Full-Text Extraction: Extract complete text content with intelligent fallback strategies
• Citation Analysis: Find top cited papers from OpenAlex since a specific date
• Paper Lookup: Retrieve full metadata for specific papers by ID
• Category Discovery: Browse available categories from all sources
• Smart Rate Limiting: Respectful API usage with per-source rate limiting
• DOI Resolution: Advanced DOI resolver with Unpaywall → Crossref → Semantic Scholar fallback
• Dual Interface: Both MCP protocol and CLI access
• TypeScript: Full type safety with ESM modules

📊 Coverage Statistics

• Total Sources: 6 academic databases
• Category Coverage: 100+ categories across all disciplines
• Paper Access: 200M+ papers with intelligent text extraction
• Text Extraction Success: >90% for supported paper types
• Response Time: <15 seconds average for paper fetching

🛠 Installation

npm install
npm run build

📋 MCP Client Configuration

To use this server with an MCP client (like Claude Desktop), add the following to your MCP client configuration:

For published package (available on npm):

Option 1: Using npx (recommended for AI tools like Claude)

{
  "mcpServers": {
    "scientific-papers": {
      "command": "npx",
      "args": [
        "-y",
        "@futurelab-studio/latest-science-mcp@latest"
      ]
    }
  }
}

Option 2: Global installation

npm install -g @futurelab-studio/latest-science-mcp

Then configure:

{
  "mcpServers": {
    "scientific-papers": {
      "command": "latest-science-mcp"
    }
  }
}

📖 Usage

CLI Interface

List Categories

# List arXiv categories
node dist/cli.js list-categories --source=arxiv

# List OpenAlex concepts
node dist/cli.js list-categories --source=openalex

# List PMC biomedical categories
node dist/cli.js list-categories --source=pmc

# List Europe PMC life science categories
node dist/cli.js list-categories --source=europepmc

# List bioRxiv/medRxiv categories (includes both servers)
node dist/cli.js list-categories --source=biorxiv

# List CORE academic categories
node dist/cli.js list-categories --source=core

Fetch Latest Papers

# Get latest AI papers from arXiv
node dist/cli.js fetch-latest --source=arxiv --category=cs.AI --count=10

# Get latest biology papers from bioRxiv
node dist/cli.js fetch-latest --source=biorxiv --category="biorxiv:biology" --count=5

# Get latest immunology papers from PMC
node dist/cli.js fetch-latest --source=pmc --category=immunology --count=3

# Get latest papers from CORE by subject
node dist/cli.js fetch-latest --source=core --category=computer_science --count=5

# Search by concept name (OpenAlex)
node dist/cli.js fetch-latest --source=openalex --category="machine learning" --count=3

Fetch Top Cited Papers

# Get top 20 cited papers in machine learning since 2024
node dist/cli.js fetch-top-cited --concept="machine learning" --since=2024-01-01 --count=20

# Get top cited papers by concept ID
node dist/cli.js fetch-top-cited --concept=C41008148 --since=2023-06-01 --count=10

Search Papers

# Search by keywords across all fields
node dist/cli.js search-papers --source=arxiv --query="machine learning" --count=10

# Search by paper title
node dist/cli.js search-papers --source=openalex --query="neural networks" --field=title --count=5

# Search by author name
node dist/cli.js search-papers --source=europepmc --query="John Smith" --field=author --count=10

# Search full-text content sorted by citations
node dist/cli.js search-papers --source=core --query="climate change" --field=fulltext --sortBy=citations --count=20

Fetch Specific Paper Content

# Get arXiv paper by ID
node dist/cli.js fetch-content --source=arxiv --id=2401.12345

# Get bioRxiv paper by DOI
node dist/cli.js fetch-content --source=biorxiv --id="10.1101/2021.01.01.425001"

# Get PMC paper by ID
node dist/cli.js fetch-content --source=pmc --id=PMC8245678

# Get CORE paper by ID
node dist/cli.js fetch-content --source=core --id=12345678

# Show text content with preview
node dist/cli.js fetch-content --source=arxiv --id=2401.12345 --show-text --text-preview=500

🔧 Available Tools

list_categories

Lists available categories/concepts from any data source.

Parameters:

• source: "arxiv" | "openalex" | "pmc" | "europepmc" | "biorxiv" | "core"

Returns:

• Array of category objects with id, name, and optional description

Examples:

{
  "name": "list_categories",
  "arguments": {
    "source": "biorxiv"
  }
}

fetch_latest

Fetches the latest papers from any source for a given category with metadata only (no text extraction).

Parameters:

• source: "arxiv" | "openalex" | "pmc" | "europepmc" | "biorxiv" | "core"
• category: Category ID or concept name (varies by source)
• count: Number of papers to fetch (default: 50, max: 200)

Category Examples by Source:

• arXiv: "cs.AI", "physics.gen-ph", "math.CO"
• OpenAlex: "artificial intelligence", "machine learning", "C41008148"
• PMC: "immunology", "genetics", "neuroscience"
• Europe PMC: "biology", "medicine", "cancer"
• bioRxiv/medRxiv: "biorxiv:neuroscience", "medrxiv:psychiatry"
• CORE: "computer_science", "mathematics", "physics"

Returns:

• Array of paper objects with metadata (id, title, authors, date, pdf_url)
• Text field: Empty string (text: "") - use fetch_content for full text

fetch_top_cited

Fetches the top cited papers from OpenAlex for a given concept since a specific date.

Parameters:

• concept: Concept name or OpenAlex concept ID
• since: Start date in YYYY-MM-DD format
• count: Number of papers to fetch (default: 50, max: 200)

search_papers

Searches for papers across multiple academic sources with field-specific search and sorting options.

Parameters:

• source: "arxiv" | "openalex" | "europepmc" | "core"
• query: Search query string (max 1500 characters)
• field: "all" | "title" | "abstract" | "author" | "fulltext" (default: "all")
• count: Number of results to return (default: 50, max: 200)
• sortBy: "relevance" | "date" | "citations" (default: "relevance")

Search Capabilities by Source:

• arXiv: Title, abstract, author, and general search with Boolean operators
• OpenAlex: Advanced search with relevance scoring and citation sorting
• Europe PMC: Biomedical literature with MeSH terms and full-text search
• CORE: Global academic papers with advanced query language

Example Queries:

• Keywords: "machine learning", "climate change"
• Phrases: "artificial intelligence" (use quotes for exact phrases)
• Boolean: "deep learning AND neural networks" (arXiv supports this)
• Authors: "John Smith", "Smith J"

Returns:

• Array of paper objects with metadata (id, title, authors, date, pdf_url)
• Text field: Empty string (text: "") - use fetch_content for full text

fetch_content

Fetches full metadata and text content for a specific paper by ID with complete text extraction.

Parameters:

• source: Any of the 6 supported sources
• id: Paper ID (format varies by source)

ID Formats by Source:

• arXiv: "2401.12345", "cs/0601001", "1234.5678v2"
• OpenAlex: "W2741809807" or numeric 2741809807
• PMC: "PMC8245678" or "12345678"
• Europe PMC: "PMC8245678", "12345678", or DOI
• bioRxiv/medRxiv: "10.1101/2021.01.01.425001" or "2021.01.01.425001"
• CORE: Numeric ID like "12345678"

📄 Paper Metadata Format

All tools return paper objects with the following structure:

{
  id: string;                    // Paper ID
  title: string;                 // Paper title
  authors: string[];             // List of author names
  date: string;                  // Publication date (ISO format)
  pdf_url?: string;              // PDF URL (if available)
  text: string;                  // Extracted full text content
  textTruncated?: boolean;       // Warning: text was truncated due to size limits
  textExtractionFailed?: boolean; // Warning: text extraction failed
}

🧠 Advanced Text Extraction

Multi-Source Strategy

Each source has specialized text extraction approaches:

• arXiv: HTML from arxiv.org/html with ar5iv.labs.arxiv.org fallback
• OpenAlex: HTML sources with DOI resolver fallback chain
• PMC: E-utilities API with XML/HTML extraction
• Europe PMC: REST API with multiple URL strategies
• bioRxiv/medRxiv: Direct HTML extraction with abstract fallback
• CORE: PDF/HTML with source URL fallback

DOI Resolution Chain

Advanced DOI resolver with multiple fallback strategies:

1. Unpaywall → Free full-text sources
2. Crossref → Publisher metadata and links
3. Semantic Scholar Academic Graph → Alternative access

Performance & Reliability

• Text Extraction Success: >90% for HTML-available papers
• Graceful Degradation: Always returns metadata even if text extraction fails
• Size Management: 6MB text limit with intelligent truncation
• Caching: 24-hour LRU cache for DOI resolution

🔄 Rate Limiting

Respectful API usage with per-source rate limiting:

• arXiv: 5 requests per minute
• OpenAlex: 10 requests per minute
• PMC: 3 requests per second
• Europe PMC: 10 requests per minute
• bioRxiv/medRxiv: 5 requests per minute
• CORE: 10 requests per minute (public), higher with API key

CORE API Configuration

For enhanced CORE access, set environment variable:

export CORE_API_KEY="your-api-key"

🧪 Testing

Run Test Suite

# Run all tests
npm test

# Run integration tests
npm run test -- tests/integration

# Run end-to-end workflow tests
npm run test -- tests/e2e

# Run performance benchmarks
npm run test -- tests/integration/performance.test.ts

Test Coverage

• Integration Tests: All 6 sources tested end-to-end
• Performance Tests: Response time and throughput benchmarks
• Workflow Tests: Real research scenarios across multiple sources
• Unit Tests: Core components and edge cases

🏗 Architecture

Modular Driver System

• Clean separation between sources
• Consistent interface across all drivers
• Specialized text extraction per source

Advanced Features

• DOI Resolution: Multi-provider fallback chain
• Rate Limiting: Token bucket algorithm per source
• Text Processing: HTML cleaning and normalization
• Error Handling: Structured responses with actionable suggestions
• Caching: Intelligent caching for DOI resolution

Technology Stack

• TypeScript + ESM: Modern JavaScript with full type safety
• Modular Design: Clean separation of concerns
• Graceful Degradation: Always functional even with partial failures
• Response Size Management: Automatic truncation and warnings

📊 Source Comparison

🔧 Development

Build

npm run build

Test Individual Sources

# Test specific sources
node dist/cli.js list-categories --source=arxiv
node dist/cli.js fetch-latest --source=biorxiv --category="biorxiv:biology" --count=3
node dist/cli.js fetch-content --source=core --id=12345678

# Test search functionality
node dist/cli.js search-papers --source=arxiv --query="artificial intelligence" --count=5
node dist/cli.js search-papers --source=openalex --query="quantum computing" --field=title --count=3

Performance Testing

# Run performance benchmarks
npm run test -- tests/integration/performance.test.ts

# Test memory usage
npm run test -- --reporter=verbose

🚨 Error Handling

Comprehensive error handling for all sources:

• Invalid paper IDs with format suggestions
• Rate limiting with retry-after information
• API timeouts and server errors
• Missing authentication (CORE API key)
• Network connectivity issues
• Text extraction failures with fallback strategies

🔍 Troubleshooting

Common Issues

• Rate limiting: Automatic retry with exponential backoff
• Missing papers: Try alternative sources for the same content
• Text extraction failures: Fallback to abstract or metadata
• CORE API limits: Set CORE_API_KEY environment variable

Performance Optimization

• Use appropriate count parameters (smaller for faster responses)
• Cache results when possible
• Use fetch_latest for discovery, fetch_content for detailed reading

📝 License

MIT

Ready to explore the world's scientific knowledge? Start with any of the 6 sources and discover papers across all academic disciplines! 🔬📚