Semantic Scholar MCP Server

uvx semantic-scholar-fastmcp

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

Sixteen tools cover the Semantic Scholar corpus at three levels: papers (relevance search, bulk search, exact title match, autocomplete, details), authors (search, details, their papers) and the citation graph, where paper_citations and paper_references walk the same edge in opposite directions. The two batch tools are what keep a literature review from turning into fifty round trips — paper_batch_details and author_batch_details fetch many records in one call, and at review scale that is the difference between a working session and a rate-limit wall. snippet_search is the other useful outlier, matching inside the text of papers rather than their metadata. Built on FastMCP; the API key is optional, and the client throttles itself to stay within the service's limits.