mcp-scholarly

A Model Context Protocol (MCP) server to search for accurate academic articles from arXiv and Google Scholar.

Note: This project is forked and modified from adityak74/mcp-scholarly

Features

• Search academic articles on arXiv
• Search academic articles on Google Scholar
• MCP-compliant server for integration with AI assistants

Requirements

• Python >= 3.11
• Dependencies listed in pyproject.toml

Installation

1. Clone the repository:

git clone https://github.com/LLiuZheng/scholarly.git
cd scholarly

2. Install uv if you haven't already:

pip install uv

Running the MCP Server

Using uv (Recommended)

Run the server with default settings (port 5000):

uv run mcp-scholarly

Custom Port and Options

You can customize the server with various options:

# Run on a custom port
uv run mcp-scholarly --port 8080

# Set logging level
uv run mcp-scholarly --log-level DEBUG

# Enable JSON responses instead of SSE streams
uv run mcp-scholarly --json-response

Using Python Module

Alternatively, run as a Python module:

uv run python -m mcp_scholarly

Environment Variables

You can set the default port using an environment variable:

# Set in .env file
SCHOLARLY_MCP_SERVER_PORT=5000

Available Options

• --port: Port to listen on for HTTP (default: 5000)
• --log-level: Logging level - DEBUG, INFO, WARNING, ERROR, CRITICAL (default: INFO)
• --json-response: Enable JSON responses for StreamableHTTP instead of SSE streams

License

See LICENSE file for details.