Semantic Scholar

Semantic Scholar MCP Server

A powerful wrapper for the Semantic Scholar Graph API implemented as an MCP (Multi-agent Control Protocol) server. This server provides a collection of tools for academic research agents to seamlessly access and query the Semantic Scholar database of academic papers, authors, and citations.

Semantic Scholar MCP Server

Features

• Comprehensive API Coverage: Access to key Semantic Scholar endpoints including papers, authors, citations, references, and snippets
• Batch Operations: Efficient batch retrieval of multiple papers or authors in a single request
• Advanced Search: Support for keyword search, metadata matching, and snippet search
• Authentication Support: Optional API key configuration for higher rate limits
• User-Friendly: Well-documented tools with clear parameter descriptions and validation
• Multi-agent Compatible: Designed to work seamlessly with MCP-based agent systems

Installation

Prerequisites

• Python 3.12 or higher
• pip (Python package manager)

Setup

1. Clone the repository:

git clone https://github.com/ZidongS/Semantic-Scholar-MCP-Server.git
cd Semantic-Scholar-MCP-Server

1. Install the required dependencies:

# after enter conda or uv environment
(uv) pip install -e .

This will install the following dependencies:

• mcp: For running the MCP server
• requests>=2.31: For making HTTP requests to the Semantic Scholar API
• pyyaml>=6.0: For parsing configuration files

Configuration

API Key Setup (Optional but Recommended)

The Semantic Scholar API can be used without an API key, but with stricter rate limits. To use an API key:

1. Obtain an API key from Semantic Scholar

1. Configure the API key using one of these methods:

Method 1: Environment Variable (Recommended for production)

   export SEMANTIC_SCHOLAR_API_KEY=your_api_key_here

Method 2: Configuration File Edit the config.yaml file in the project directory:

   semantic_scholar:
   api_key: your_api_key_here

Server Configuration

The server configuration is defined in the server.py file. The default settings are:

• Host: 127.0.0.1
• Port: 6666
• Transport: sse (Server-Sent Events)

You can modify these settings by editing the relevant lines in server.py:

mcp = FastMCP("Semantic-Scholar-MCP", host="127.0.0.1", port=6666)

# And later in the file
if __name__ == "__main__":
    mcp.run(transport="sse")

Usage

Starting the Server

To start the MCP server, run:

python server.py

The server will start listening for requests on the configured host and port.

Available Tools

The server provides the following tools for interacting with the Semantic Scholar API:

Paper Tools

• get_paper: Retrieve a single paper by ID
• get_papers_batch: Batch retrieve multiple papers
• search_papers: Search for papers with filters
• search_papers_bulk: Bulk search with multiple queries
• search_papers_match: Find papers matching metadata
• paper_autocomplete: Get autocomplete suggestions for paper titles
• get_paper_authors: List authors of a paper
• get_paper_citations: List papers citing a specific paper
• get_paper_references: List papers referenced by a specific paper

Author Tools

• get_author: Retrieve an author by ID
• get_authors_batch: Batch retrieve multiple authors
• search_authors: Search for authors
• get_author_papers: List papers by an author

Snippet Tools

• snippet_search: Search for relevant snippets/passages in papers

Tool Usage Examples

Get Paper Information

# Example: Get paper information by DOI
get_paper(paper_id="DOI:10.1145/3474123.3486132")

# Example: Get paper with specific fields
get_paper(
    paper_id="arXiv:1706.03762",
    fields=["title", "authors", "year", "abstract", "citationCount"]
)

Search Papers

# Example: Search for papers on machine learning
search_papers(
    query="machine learning",
    limit=10,
    year="2022-2024",
    fields_of_study=["Computer Science"]
)

Get Author Information

# Example: Get author information
get_author(author_id="123456789")

# Example: Get author's papers
get_author_papers(
    author_id="123456789",
    limit=5,
    sort="citationCount"
)

API Reference

General Parameters

Most tools accept the following common parameters:

• fields: Specify which fields to include in the response (comma-separated string or list)
• limit: Maximum number of results to return
• offset: Pagination offset for retrieving additional results

Rate Limits

Be aware of Semantic Scholar's API rate limits:

• Without API key: Approximately 100 requests per 5 minutes
• With API key: Higher rate limits based on your account tier

Error Handling

Tools will raise descriptive errors in case of invalid parameters or API failures. Common errors include:

• Invalid parameter values (empty strings, incorrect types)
• Exceeding API limits
• Resource not found (invalid IDs)
• Network connectivity issues

Development

Adding New Tools

To add a new tool to interact with the Semantic Scholar API:

1. Create a new function decorated with @mcp.tool()
2. Implement parameter validation
3. Use the _request_json helper for making API calls
4. Return JSON-serialized response

Example:

@mcp.tool()
def new_tool(param1, param2=None):
    # Validate parameters
    if not param1:
        raise ValueError("param1 is required")
    
    # Prepare request
    params = {"param1": param1}
    if param2:
        params["param2"] = param2
    
    # Make API call
    data = _request_json("GET", "/endpoint/path", params=params)
    
    # Return response
    return json.dumps(data, ensure_ascii=False)

Testing

Currently, the project doesn't include a formal test suite. Manual testing is recommended during development.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Semantic Scholar for providing the API
• MCP framework developers for the multi-agent protocol implementation

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

When contributing, please ensure:

1. Code follows the existing style and conventions
2. New tools include comprehensive documentation
3. Proper error handling is implemented
4. Parameters are properly validated