ArxivMCP

A tool for searching and retrieving academic papers from arXiv.

Prerequisites

• Python 3.10 or higher
• uv (Python package installer)

If you don't have uv installed, you can install it via pip:

pip install uv

Or follow the official installation guide.

Installation

1. Clone the repository:

   git clone <repository-url> # Replace with the actual URL
   cd paper-search
   

2. Set up the Python environment using uv: This command creates a virtual environment (.venv) if it doesn't exist and installs the dependencies listed in pyproject.toml.

   uv sync
   

3. Activate the virtual environment:

   source .venv/bin/activate
   

   (On Windows, use: .venv\\Scripts\\activate)

Usage

To run the main script:

python main.py

IDE Configuration (Cursor/MCP)

To configure an MCP-enabled client (like Cursor or Claude Desktop) to use this paper search server, you need to modify the client's configuration file. For example, in Claude Desktop, the file is typically located at ~/Library/Application Support/Claude/claude_desktop_config.json (Mac/Linux) or $env:AppData\\Claude\\claude_desktop_config.json (Windows).

Add an entry for this server under the mcpServers key like this:

{
    "mcpServers": {
        "paper-search": { // You can name this server entry
            "command": "uv", // Or provide the full path from `which uv` / `where uv`, something you might adjust it like '/opt/homebrew/bin/uv'
            "args": [
                "--directory",
                "/ABSOLUTE/PATH/TO/THIS/PROJECT/paper-search", // <-- IMPORTANT: Use the absolute path here
                "run",
                "search.py"
            ]
        }
        // ... other servers can be listed here
    }
}

Key points:

• Replace /ABSOLUTE/PATH/TO/THIS/PROJECT/paper-search with the actual absolute path to this project's directory on your system.
• If the client cannot find uv, replace "uv" in the "command" field with the full path to your uv executable (found using which uv on MacOS/Linux or where uv on Windows).
• After saving the configuration file, restart your MCP client (e.g., Claude Desktop).

Once configured and restarted, the client should recognize the server and its tools (often indicated by an icon, like a hammer 🔨 in Claude Desktop).

Dependencies

This project relies on the following main libraries:

• beautifulsoup4: For parsing HTML content.
• httpx: For making asynchronous HTTP requests.
• mcp[cli]: For MCP tooling integration.

Dependencies are managed using uv and defined in the pyproject.toml file.

TODO: Backend Code