arXiv Search MCP

A lightweight Model Context Protocol server that lets AI assistants query the arXiv API. It provides two powerful search approaches:

• Semantic Search: Uses vector embeddings to understand the meaning of your query, powered by ArxivSearch

  Note: Currently, semantic search only supports papers in Computer Science (cs.*) categories

• Keyword Search: Supports structured queries with multiple filters including categories, date ranges, and field-specific searches

  Note: Supports papers from all arXiv categories

Requirements

• Python 3.12 or higher
• uv package manager (dependency resolver & runner)

Getting Started

1. Clone & install dependencies

# Clone the repository
git clone https://github.com/icyclv/arxiv-semantic-search-mcp.git
cd arxiv-semantic-search-mcp

# Install all project dependencies listed in pyproject.toml
uv sync

2. Run the server locally

# Start the MCP server
uv run main.py

Editor / Client Configuration

To use this MCP server with your AI assistant tools (e.g. Claude Desktop, VS Code MCP extension), add the following configuration:

{
  "mcpServers": {
    "arxiv-semantic-search": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/arxiv-semantic-search-mcp",  // Replace with actual path
        "run",
        "main.py"
      ]
    }
  }
}

Note: Replace /path/to/arxiv-semantic-search-mcp with the actual path where you cloned the repository.

Available Functions

See src/server.py for detailed documentation and examples.

Usage Examples

Semantic Search

# Search for papers about deep learning in computer vision
search_semantic(
    query="recent advances in vision transformers for medical image segmentation",
    categories=["cs.CV", "cs.AI"]
)

Keyword Search

# Search for recent machine learning papers in multiple categories
search_keyword(
    categories=["cs.AI", "cs.LG"],
    start_date="2024-01-01",
    all_fields="transformer",
    sort_by="submittedDate",
    sort_order="descending"
)

Get Paper Details

# Get detailed information for a specific paper
get_details("2401.00001")

License

This project is released under the MIT License.