home / mcp / arxiv mcp server

arXiv MCP Server

Provides an MCP server to search arXiv, download papers, and read sections via MCP clients.

Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
  "mcpServers": {
    "annasusu-arxiv-mcp": {
      "command": "uvx",
      "args": [
        "arxiv-paper-mcp-server"
      ],
      "env": {
        "ARXIV_STORAGE_DIR": "~/.arxiv-mcp/papers"
      }
    }
  }
}

You run an MCP server that lets language models search arXiv, fetch papers, and extract sections for reading. It caches downloaded papers locally for fast reuse and works with many MCP clients out of the box, giving you practical access to scientific literature during conversations or automated workflows.

How to use

To leverage the arXiv MCP Server, connect it to an MCP client you use. You can search for papers by title, keywords, author, or arXiv ID, download full papers, and extract specific sections like the abstract, introduction, methods, or conclusion. You can also list papers you have cached locally and reuse them without re-downloading.

How to install

Prerequisites you need before starting: install the MCP runtime that your client uses. In this case, you will work with the uvx runtime or equivalent, so you should install uvx first.

Install the arXiv MCP server configuration in your MCP client setup by adding the following configuration snippet. This tells your client how to locate and start the arXiv MCP server using the local runtime.

{
  "mcpServers": {
    "arxiv": {
      "command": "uvx",
      "args": ["arxiv-paper-mcp-server"]
    }
  }
}

Start the server locally

After you have the config in place, run the MCP server locally using the runtime. This starts the arXiv MCP server so your MCP clients can connect and issue tool calls.

uv run arxiv-paper-mcp-server

Available tools

search

Search arXiv papers by title, keywords, or arXiv ID (e.g., 2401.12345). Returns paper ID, title, authors, date, and abstract preview.

get_paper

Download and read the full text of a paper with optional section filtering (abstract, introduction, method, conclusion, or all).

list_downloaded_papers

List all papers that have been downloaded and cached locally.