scholar-mcp 0.2.1

MCP server for academic paper search, powered by Semantic Scholar (214M+ papers)

scholar-mcp

MCP server for academic paper search, powered by Semantic Scholar (214M+ papers, 2.49B citations). Search, cite, download, and read papers directly from Claude Code or any MCP client.

Quick Start

One-liner for Claude Code:

claude mcp add scholar -- uvx scholar-mcp

Or with an API key for higher rate limits:

claude mcp add scholar -e S2_API_KEY=your_key -- uvx scholar-mcp

For Claude Desktop, add to your config:

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

[!NOTE] Requires Python 3.10+ and uv installed. No API key needed for basic use (100 requests / 5 minutes free).

Features

• Search across 214M+ papers with filters (year, venue, field of study, citation count, open access)
• Paper details with TLDR summaries, BibTeX, venue metadata
• Citation graph traversal (who cites this paper, what does it reference)
• Recommendations for similar/related papers
• Author search with h-index, affiliations, paper counts
• PDF download with smart fallback chain (Semantic Scholar -> arXiv -> CORE -> bioRxiv/medRxiv)
• Full-text extraction from downloaded PDFs
• Fallback search via arXiv, CORE (250M+ open access papers), PubMed (36M+ biomedical), and Google Scholar

Tools

Tool	Description
search_papers	Search 214M+ papers with year, venue, field, citation filters. Falls back to arXiv, CORE, PubMed, then Google Scholar
get_paper	Paper details by Semantic Scholar ID, DOI, ArXiv ID (ArXiv:xxxx), or PMID (PMID:xxxx)
get_citations	Papers that cite a given paper (up to 1000)
get_references	Papers referenced by a given paper (up to 1000)
recommend_papers	Similar/related papers via S2 recommendation engine (up to 500)
search_authors	Find researchers with h-index, affiliations, paper/citation counts
download_paper	Download PDF: tries S2 open access, arXiv, CORE, bioRxiv/medRxiv
read_paper	Download + extract full text from PDF (with optional page limit)

Configuration

All configuration is via environment variables (all optional):

Variable	Default	Description
S2_API_KEY	—	Semantic Scholar API key for higher rate limits
CORE_API_KEY	—	CORE API key for institutional repository search (free)
SCHOLAR_DOWNLOAD_DIR	./downloads	Directory for downloaded PDFs
S2_TIMEOUT	30	API request timeout in seconds

Rate limits: Free tier allows 100 requests per 5 minutes. With an API key: ~100 requests per second.

Examples

Search with filters:

search_papers("transformer architecture", year="2020-2024", venue="NeurIPS", min_citations=100)

Look up specific papers:

get_paper("ArXiv:1706.03762")      # by arXiv ID
get_paper("10.1038/nature12373")   # by DOI
get_paper("PMID:19872477")         # by PubMed ID

Explore the citation graph:

get_citations("ArXiv:1706.03762", limit=20)   # who cites this?
get_references("ArXiv:1706.03762", limit=20)  # what does it cite?
recommend_papers("ArXiv:1706.03762")           # find similar work

Download and read:

download_paper("ArXiv:1706.03762")
read_paper("ArXiv:1706.03762", max_pages=5)

Development

git clone https://github.com/Liyux3/scholar-mcp.git
cd scholar-mcp
uv venv && uv pip install -e ".[dev]"
uv run pytest tests/

How It Works

Search:     S2 (214M) -> arXiv (preprints) -> CORE (institutional) -> PubMed (biomedical) -> Google Scholar (scraping)
Download:   S2 open access -> arXiv direct -> CORE (by DOI/title) -> bioRxiv/medRxiv -> fail

License

MIT