ArXiv

Provides access to arXiv metadata, PDF downloads, and article searches via natural language MCP endpoints.

python

34

GitHub Stars

python

Language

6 months ago

First Indexed

2 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": {
    "prashalruchiranga-arxiv-mcp-server": {
      "command": "uv",
      "args": [
        "--directory",
        "/ABSOLUTE/PATH/TO/PARENT/FOLDER/arxiv-mcp-server/src/arxiv_server",
        "run",
        "server.py"
      ],
      "env": {
        "DOWNLOAD_PATH": "/ABSOLUTE/PATH/TO/DOWNLOADS/FOLDER"
      }
    }
  }
}

You can interact with arXiv content through a dedicated MCP server that translates natural language requests into arXiv API actions. This lets you look up article details, fetch metadata, search for papers, download PDFs, and load articles into a large language model context with simple prompts and minimal manual steps.

How to use

To work with this MCP server, you connect your MCP client to the local server using the provided runtime configuration. You can ask it to retrieve article details by title, fetch downloadable metadata, search the arXiv catalog, download PDFs, or load an article into your LLM context. Typical usage patterns include: asking for metadata of a specific paper, downloading the PDF to your machine, or listing papers by a given author. Use natural language prompts to trigger the corresponding tools, such as getting article details, downloading PDFs, or loading the article content into your model’s context.

How to install

Prerequisites you need before installation: a Python-compatible runtime and a MCP runtime for running stdio servers. You will also need a terminal or shell with permission to install packages and clone repositories.

Additional steps

Install MacOS or Windows dependencies, clone the project, and prepare the environment as described below. The server is started via a stdio MCP command that runs the Python server script using the UV runtime. The installation flow includes creating a virtual environment, syncing development dependencies, and then starting the server with the specified command.

Available tools

get_article_url

Retrieve the URL of an arXiv article based on its title.

download_article

Download the arXiv article as a PDF to your local machine.

load_article_to_context

Load the article into the context of a language model for use in prompts.

get_details

Fetch metadata for an arXiv article by its title.

search_arxiv

Search the arXiv database with various parameters and return matching article metadata.