literature-discovery-mcp 0.1.0

MCP server for scholarly literature discovery via arXiv and Semantic Scholar

literature-discovery-mcp

Private MCP server for scholarly literature discovery. It covers the discovery/metadata side of the LIST literature-search workflow and intentionally delegates web-page/PDF crawling to an external fetch_url MCP.

Capabilities

• arXiv discovery:
  • search_arxiv_papers
  • get_arxiv_paper
  • search_arxiv_by_author
  • search_arxiv_by_category
  • get_arxiv_urls
• Semantic Scholar discovery:
  • search_semantic_scholar_papers
  • get_paper_metadata
  • get_paper_citations
  • get_paper_references
  • get_related_papers
  • search_authors
  • get_author
• Citation export:
  • get_bibtex

Returned arXiv and Semantic Scholar search records use a normalized schema with fields such as:

{
  "id": "arXiv:2402.03300",
  "source": "arxiv",
  "title": "...",
  "authors": ["..."],
  "year": 2024,
  "abstract": "...",
  "url": "https://arxiv.org/abs/...",
  "html_url": "https://arxiv.org/html/...",
  "pdf_url": "https://arxiv.org/pdf/...",
  "doi": "...",
  "citation_count": 42,
  "access": "abstract_only",
  "caveats": []
}

Use your external fetch_url MCP with url, html_url, or pdf_url to retrieve full pages or PDFs.

Install and run

uv sync
uv run literature-discovery-mcp

MCP client configuration example:

mcp_servers:
  literature_discovery:
    command: "uvx"
    args: ["--from", "git+https://github.com/pidgeos/literature-discovery-mcp.git", "literature-discovery-mcp"]
    env:
      SEMANTIC_SCHOLAR_API_KEY: "optional-key"

For local development:

mcp_servers:
  literature_discovery:
    command: "uv"
    args: ["--directory", "/path/to/literature-discovery-mcp", "run", "literature-discovery-mcp"]

Configuration

• SEMANTIC_SCHOLAR_API_KEY / S2_API_KEY: optional Semantic Scholar API key.
• ARXIV_RATE_LIMIT_RPS: default 0.33 to respect arXiv's courtesy limit.
• SEMANTIC_SCHOLAR_RATE_LIMIT_RPS / S2_RATE_LIMIT_RPS: default 1.0.
• ARXIV_PAGE_SIZE: default 100.
• SEMANTIC_SCHOLAR_TIMEOUT / S2_TIMEOUT: default 30 seconds.
• LITERATURE_MAX_ATTEMPTS: default 3.
• LITERATURE_RETRY_BACKOFF_SECONDS: default 1.0.

Rate limits are process-local, not distributed across multiple MCP processes.

Local quality gates

uv lock --check
uv sync --locked --dev
uv run ruff format --check .
uv run ruff check .
uv run coverage run -m pytest -q
uv run coverage report --fail-under=95
uv build