Semantic Scholar

Semantic Scholar MCP Server

A FastMCP server implementation for the Semantic Scholar API, providing comprehensive access to academic paper data, author information, and citation networks.

Looking for Claude Code skills? Check out semantic-scholar-skills — the next-generation toolkit that bundles this MCP server with ready-to-use Claude Code skills (/expand-references, /trace-citations, /paper-triage) and a Python workflow engine.

Project Structure

The project has been refactored into a modular structure for better maintainability:

semantic-scholar-server/
├── semantic_scholar/            # Main package
│   ├── __init__.py             # Package initialization
│   ├── server.py               # Server setup and main functionality
│   ├── mcp.py                  # Centralized FastMCP instance definition
│   ├── config.py               # Configuration classes
│   ├── utils/                  # Utility modules
│   │   ├── __init__.py
│   │   ├── errors.py           # Error handling
│   │   └── http.py             # HTTP client and rate limiting
│   ├── api/                    # API endpoints
│       ├── __init__.py
│       ├── papers.py           # Paper-related endpoints
│       ├── authors.py          # Author-related endpoints
│       └── recommendations.py  # Recommendation endpoints
├── run.py                      # Entry point script

This structure:

• Separates concerns into logical modules
• Makes the codebase easier to understand and maintain
• Allows for better testing and future extensions
• Keeps related functionality grouped together
• Centralizes the FastMCP instance to avoid circular imports

Features

• Paper Search & Discovery

  • Full-text search with advanced filtering
  • Title-based paper matching
  • Paper recommendations (single and multi-paper)
  • Batch paper details retrieval
  • Advanced search with ranking strategies

• Citation Analysis

  • Citation network exploration
  • Reference tracking
  • Citation context and influence analysis

• Author Information

  • Author search and profile details
  • Publication history
  • Batch author details retrieval

• Advanced Features

  • Complex search with multiple ranking strategies
  • Customizable field selection
  • Efficient batch operations
  • Rate limiting compliance
  • Support for both authenticated and unauthenticated access
  • Graceful shutdown and error handling
  • Connection pooling and resource management

System Requirements

• Python 3.10+
• Environment variable for API key (optional)

Installation

Quick Install (PyPI)

pip install semantic-scholar-fastmcp

Or run directly with uvx:

uvx semantic-scholar-fastmcp

MCP Client Configuration

For any MCP client that supports uvx:

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

Installing via Smithery

To install Semantic Scholar MCP Server for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install semantic-scholar-fastmcp-mcp-server --client claude

Manual Installation

1. Clone the repository:

git clone https://github.com/YUZongmin/semantic-scholar-fastmcp-mcp-server.git
cd semantic-scholar-fastmcp-mcp-server

2. Install in development mode:

pip install -e ".[dev]"

3. Run the server:

semantic-scholar-mcp-server

API Key (Optional)

To get higher rate limits and better performance:

1. Get an API key from Semantic Scholar API
2. Add it to your FastMCP configuration as shown above in the env section

If no API key is provided, the server will use unauthenticated access with lower rate limits.

Configuration

Contributing

See CONTRIBUTING.md.

Environment Variables

• SEMANTIC_SCHOLAR_API_KEY: Your Semantic Scholar API key (optional)
  • Get your key from Semantic Scholar API
  • If not provided, the server will use unauthenticated access

HTTP Bridge (Built-in)

This repository includes a small HTTP bridge (semantic_scholar.bridge) that exposes a minimal REST API for common workflows.

Default listening port: 8000.

Bridge configuration via environment variables:

• SEMANTIC_SCHOLAR_ENABLE_HTTP_BRIDGE (default: 1) — set to 0 to disable
• SEMANTIC_SCHOLAR_HTTP_BRIDGE_HOST (default: 0.0.0.0)
• SEMANTIC_SCHOLAR_HTTP_BRIDGE_PORT (default: 8000)

Available endpoints:

• GET /v1/paper/search?q=... — paper search (params: fields, offset, limit)
• GET /v1/paper/{paper_id} — paper details (param: fields)
• POST /v1/paper/batch — batch paper details (JSON: { "ids": [ ... ] })
• GET /v1/author/search?q=... — author search (params: fields, offset, limit)
• GET /v1/author/{author_id} — author details (param: fields)
• POST /v1/author/batch — batch author details (JSON: { "ids": [ ... ] })
• GET /v1/recommendations?paper_id=... — recommendations for a paper

The bridge reuses the package's HTTP utilities (semantic_scholar.utils.http) so rate limits, API key handling and connection pooling remain consistent with the MCP tools.

Example:

curl 'http://localhost:8000/v1/paper/search?q=machine+learning&limit=5'

Rate Limits

The server automatically adjusts to the appropriate rate limits:

With API Key:

• Search, batch and recommendation endpoints: 1 request per second
• Other endpoints: 10 requests per second

Without API Key:

• All endpoints: 100 requests per 5 minutes
• Longer timeouts for requests

Note: Rate limits are subject to change. See Semantic Scholar API for the latest information.

Semantic Scholar API Terms

This project uses the Semantic Scholar Academic Graph API, provided by the Allen Institute for AI (AI2). Please review the API License Agreement before use.

Available MCP Tools

The server currently exposes 16 MCP tools.

Note: All tools are aligned with the official Semantic Scholar API documentation. Please refer to the official documentation for detailed field specifications and the latest updates.

Paper Search Tools

• paper_relevance_search: Search for papers using relevance ranking

  • Supports comprehensive query parameters including year range and citation count filters
  • Returns paginated results with customizable fields

• paper_bulk_search: Bulk paper search with sorting options

  • Similar to relevance search but optimized for larger result sets
  • Supports sorting by citation count, publication date, etc.

• paper_title_search: Find papers by exact title match

  • Useful for finding specific papers when you know the title
  • Returns detailed paper information with customizable fields

• paper_details: Get comprehensive details about a specific paper

  • Accepts various paper ID formats (S2 ID, DOI, ArXiv, etc.)
  • Returns detailed paper metadata with nested field support

• paper_batch_details: Efficiently retrieve details for multiple papers

  • Accepts up to 1000 paper IDs per request
  • Supports the same ID formats and fields as single paper details

• paper_authors: Get the authors associated with a specific paper

  • Returns paginated author results for the paper
  • Supports author field customization with offset and limit controls

• paper_autocomplete: Get paper title suggestions for a partial query

  • Returns minimal paper metadata for interactive search completion
  • Truncates overly long queries to the API-supported length

• snippet_search: Search within paper snippets and excerpts

  • Returns relevant text matches from titles, abstracts, and body text
  • Supports filters for paper IDs, authors, venue, year, and field of study

Citation Tools

• paper_citations: Get papers that cite a specific paper

  • Returns paginated list of citing papers
  • Includes citation context when available
  • Supports field customization and sorting

• paper_references: Get papers referenced by a specific paper

  • Returns paginated list of referenced papers
  • Includes reference context when available
  • Supports field customization and sorting

Author Tools

• author_search: Search for authors by name

  • Returns paginated results with customizable fields
  • Includes affiliations and publication counts

• author_details: Get detailed information about an author

  • Returns comprehensive author metadata
  • Includes metrics like h-index and citation counts

• author_papers: Get papers written by an author

  • Returns paginated list of author's publications
  • Supports field customization and sorting

• author_batch_details: Get details for multiple authors

  • Efficiently retrieve information for up to 1000 authors
  • Returns the same fields as single author details

Recommendation Tools

• get_paper_recommendations_single: Get recommendations based on a single paper

  • Returns similar papers based on content and citation patterns
  • Supports field customization for recommended papers

• get_paper_recommendations_multi: Get recommendations based on multiple papers

  • Accepts positive and negative example papers
  • Returns papers similar to positive examples and dissimilar to negative ones

Usage Examples

Basic Paper Search

results = await paper_relevance_search(
    context,
    query="machine learning",
    year="2020-2024",
    min_citation_count=50,
    fields=["title", "abstract", "authors"]
)

Paper Recommendations

# Single paper recommendation
recommendations = await get_paper_recommendations_single(
    context,
    paper_id="649def34f8be52c8b66281af98ae884c09aef38b",
    fields="title,authors,year"
)

# Multi-paper recommendation
recommendations = await get_paper_recommendations_multi(
    context,
    positive_paper_ids=["649def34f8be52c8b66281af98ae884c09aef38b", "ARXIV:2106.15928"],
    negative_paper_ids=["ArXiv:1805.02262"],
    fields="title,abstract,authors"
)

Batch Operations

# Get details for multiple papers
papers = await paper_batch_details(
    context,
    paper_ids=["649def34f8be52c8b66281af98ae884c09aef38b", "ARXIV:2106.15928"],
    fields="title,authors,year,citations"
)

# Get details for multiple authors
authors = await author_batch_details(
    context,
    author_ids=["1741101", "1780531"],
    fields="name,hIndex,citationCount,paperCount"
)

Error Handling

The server provides standardized error responses:

{
    "error": {
        "type": "error_type",  # rate_limit, api_error, validation, timeout
        "message": "Error description",
        "details": {
            # Additional context
            "authenticated": true/false  # Indicates if request was authenticated
        }
    }
}