semantic-scholar-mcp 0.3.0

MCP server for Semantic Scholar API integration with Claude Desktop

Semantic Scholar MCP Server

Access millions of academic papers from Semantic Scholar using the Model Context Protocol (MCP). Works with Claude Code, Claude Desktop, Cursor, VS Code, and other MCP-compatible editors.

Response Schema (Contract)

All tools return a compact JSON object shaped around a top-level data key. Pagination metadata is included only when applicable:

• Always: data
• Paginated endpoints (e.g., search_papers, search_authors, search_snippets): total, offset, limit, has_more
• Count-only endpoints (e.g., recommendations, batch ops, datasets list): count
• Non-paginated single-item endpoints (e.g., get_paper, get_author, dataset info): only data

This keeps payloads predictable and compact for MCP clients.

Instruction Templates (SSOT)

Tool guidance (“Next Steps”) is injected from YAML files and treated as the single source of truth (SSOT):

• Location: src/semantic_scholar_mcp/resources/tool_instructions/**/*.yml
• Server auto-injects guidance into tool descriptions and responses
• Markdown templates are kept only for backward compatibility

Features

• Smart Search: Search papers with filters for year, fields of study, and sorting
• Full Paper Details: Get abstracts, authors, citations, and references
• Author Profiles: Explore researcher profiles and their publications
• Citation Network: Analyze citation relationships and impact
• AI-Powered: Get paper recommendations and research insights
• Fast & Reliable: Built-in caching, rate limiting, and error recovery
• PDF Conversion: Turn open-access PDFs into Markdown or semantic chunks with optional image capture

Installation

One-Command Setup

Claude Code (recommended):

claude mcp add semantic-scholar -- uvx semantic-scholar-mcp

Manual Configuration (if needed)

Basic setup:

{
  "mcpServers": {
    "semantic-scholar": {
      "command": "uvx",
      "args": ["semantic-scholar-mcp"]
    }
  }
}

With API key for higher limits:

{
  "mcpServers": {
    "semantic-scholar": {
      "command": "uvx",
      "args": ["semantic-scholar-mcp"],
      "env": {
        "SEMANTIC_SCHOLAR_API_KEY": "your-api-key-here"
      }
    }
  }
}

Get your free API key at: https://www.semanticscholar.org/product/api

Usage

Ask in natural language:

• "Find recent papers on transformer architectures in NLP"
• "Show me details about paper with DOI 10.1038/nature14539"
• "Find papers by Yoshua Bengio from 2020 onwards"
• "Get recommendations based on the BERT paper"
• "Create a literature review on quantum computing"

Available Tools

📄 Paper Tools

Tool	Description	Example
search_papers	Search papers with filters	"Search for deep learning papers from 2023"
get_paper	Get detailed paper info	"Get full details for paper ID abc123"
get_paper_fulltext	Convert open-access PDFs to Markdown or chunks	"Convert the PDF for paper abc123 into Markdown"
get_paper_citations	Get papers citing this paper	"Find papers that cite the attention paper"
get_paper_references	Get papers this paper cites	"Show references from the BERT paper"
get_paper_authors	Get detailed author info for paper	"Show authors of paper abc123"
batch_get_papers	Get multiple papers efficiently	"Get details for papers: abc123, def456, ghi789"
bulk_search_papers	Advanced search with filters	"Search ML papers from 2020-2023 with 50+ citations"
search_papers_match	Search by exact title match	"Find paper with title 'Attention Is All You Need'"

👤 Author Tools

Tool	Description	Example
search_authors	Search for researchers	"Find authors working on computer vision"
get_author	Get author profile	"Get profile for author ID 12345"
get_author_papers	List author's publications	"Show papers by Geoffrey Hinton"
batch_get_authors	Get multiple authors efficiently	"Get details for authors: 123, 456, 789"

🤖 AI Tools

Tool	Description	Example
get_recommendations_for_paper	AI-powered paper recommendations	"Recommend papers similar to GPT-3"
get_recommendations_batch	Advanced ML recommendations	"Get recommendations from positive/negative examples"

🔍 Advanced Search Tools

Tool	Description	Example
autocomplete_query	Get search suggestions	"Complete query 'machine lear...'"
search_snippets	Search text snippets	"Find papers mentioning 'transformer architecture'"

🧠 Semantic Analysis Tools

Tool	Description	Example
get_paper_with_embeddings	Get paper with vector embeddings	"Get paper with SPECTER embeddings"
search_papers_with_embeddings	Search with semantic vectors	"Find semantically similar papers"

📊 Dataset Tools

Tool	Description	Example
get_dataset_releases	List available datasets	"Show available dataset releases"
get_dataset_info	Get dataset information	"Get info for dataset release 2023-01"
get_dataset_download_links	Get download links	"Get download links for papers dataset"
get_incremental_dataset_updates	Get dataset updates	"Get updates between releases"

Advanced Features

📚 Resources (Direct Access)

Resource	Description	Usage
papers/{paper_id}	Direct paper data access	Auto-populated in conversations
authors/{author_id}	Direct author profile access	Auto-populated in conversations

🎯 AI Prompts (Smart Templates)

Prompt	Description	Usage
literature_review	Generate comprehensive literature reviews	"Create a literature review on machine learning"
citation_analysis	Analyze citation networks and impact	"Analyze citations for the transformer paper"
research_trend_analysis	Identify emerging research trends	"Analyze trends in NLP over the last 5 years"

Practical Examples

Search and Explore

"Find recent papers on graph neural networks published after 2022"
"Show me the most cited papers in computer vision from 2023"
"Search for papers about attention mechanisms with more than 100 citations"

Deep Analysis

"Get full details including citations and references for paper DOI 10.1038/nature14539"
"Show me all papers by Yann LeCun from the last 3 years"
"Find papers that cite 'Attention Is All You Need' and analyze their impact"

AI-Powered Research

"Based on the GPT-4 paper, recommend 5 related papers I should read"
"Create a literature review covering the evolution of transformer architectures"
"Analyze citation patterns for deep learning papers in the last decade"

Features

✅ 22 research tools covering all Semantic Scholar API endpoints
✅ Smart AI prompts for literature reviews and citation analysis
✅ Fast & reliable with built-in caching and error recovery
✅ Free to use - no API key required (optional for higher limits)

License

MIT License - see LICENSE for details.

⚠️ The get_paper_fulltext tool relies on PyMuPDF4LLM, which is AGPL licensed. Commercial usage of the PDF conversion feature may require a commercial PyMuPDF license.

PDF Markdown Tips

• Chunk-only output

  uv run semantic-scholar-mcp --tool get_paper_fulltext --argument '{
    "paper_id": "649def34f8be52c8b66281af98ae884c09aef38b",
    "output_mode": "chunks"
  }'
  

• Include extracted images

  uv run semantic-scholar-mcp --tool get_paper_fulltext --argument '{
    "paper_id": "649def34f8be52c8b66281af98ae884c09aef38b",
    "output_mode": "both",
    "include_images": true
  }'
  

• Manual cache cleanup (respects PDF_PROCESSING__ARTIFACT_TTL_HOURS)

  uv run python -c "from semantic_scholar_mcp.pdf_processor import cleanup_pdf_cache; cleanup_pdf_cache()"
  

  Cached artifacts auto-expire after the configured TTL, and you can trigger cleanup manually with the command above.

Acknowledgments

• Semantic Scholar for the academic graph API
• Anthropic for the MCP specification
• The academic community for making research accessible

---

Built for researchers worldwide 🌍