ArXiv

Provides MCP access to arXiv search, download, and read operations for academic papers.

python

GitHub Stars

python

Language

6 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "freya-lee-zou-mcp-arxiv": {
      "command": "python",
      "args": [
        "-m",
        "arxiv_mcp_server",
        "--storage-path",
        "/Users/user/Desktop/arxiv-mcp-server-main/papers"
      ],
      "env": {
        "PYTHONPATH": "/Users/you/arxiv-mcp-server-main/src",
        "ARXIV_STORAGE_PATH": "/Users/user/Desktop/arxiv-mcp-server-main/papers"
      }
    }
  }
}

The ArXiv MCP Server lets you connect AI assistants to arXiv’s research repository through the Model Context Protocol (MCP). It enables you to search for papers, download content, and access a local library of downloaded papers for rapid reference in your AI workflows.

How to use

You interact with the ArXiv MCP Server through an MCP client. Start the local server and point your client to the server using the MCP connection you configured. Use the four main operations to search for papers, download papers by arXiv ID, list your downloaded papers, and read the content of a paper you already downloaded.

How to install

Prerequisites: Python 3.11 or higher.

pip install --upgrade pip
pip install -e .

Run the server locally for testing. The server runs as a standard input/output (stdio) process and waits for MCP client connections. Stop with Ctrl+C when you are done.

Configuration and storage

Choose how you want to start the server and where to store papers. The server supports a per-user local storage path where downloaded papers are saved for quick access.

Default storage location: papers are stored under ~/.arxiv-mcp-server/papers if you do not specify a path.

Troubleshooting and notes

If you are testing manually, ensure you start the server with the correct Python interpreter from your virtual environment, and verify the storage path you pass is accessible. When you modify storage paths, keep them consistently referenced in your MCP client configuration.

API surface and prompts

This server exposes four core actions that you can invoke via your MCP client: search_papers, download_paper, list_papers, and read_paper. There is also a specialized prompt workflow for in-depth paper analysis that uses these actions to produce an executive summary, context, methodology critique, results evaluation, and future directions.

Available tools

search_papers

Query arXiv papers with optional filters such as date range and categories, returning a list of matching results.

download_paper

Download a paper by its arXiv ID and save it to local storage for offline reading.

list_papers

List all papers that have been downloaded and are available locally.

read_paper

Read the full content of a downloaded paper by its arXiv ID.

Built by

VeilStrat

AI signals for GTM teams