Skip to main content
Glama
arxiv-mcp

arxiv-mcp

by reinthal

arXiv MCP Server

A FastMCP server that provides tools for converting arXiv academic papers to Markdown format using the arxiv2md.org API.

Key Features:

  • Convert arXiv papers from LaTeX to Markdown

  • Extract arXiv IDs from URLs

  • Generate arXiv URLs from paper IDs

  • Centralized rate limiting (3 requests/second across all tools)

Tools

This MCP server exposes the following tools (all subject to rate limiting):

1. convert_arxiv_to_markdown

Convert an arXiv paper to Markdown format by downloading and processing its LaTeX source.

Parameters:

  • arxiv_url (string): The arXiv URL (e.g., https://arxiv.org/abs/1706.03762)

Returns:

  • markdown (string): The converted markdown content

  • title (string, optional): Paper title from metadata

  • authors (list, optional): Paper authors from metadata

  • abstract (string, optional): Paper abstract from metadata

  • url (string): Original arXiv URL

2. extract_arxiv_id

Extract the arXiv paper ID from a URL.

Parameters:

  • arxiv_url (string): An arXiv URL in any format

Returns:

  • The extracted arXiv ID (e.g., "1706.03762")

3. build_arxiv_urls

Generate various arXiv URLs from a paper ID.

Parameters:

  • arxiv_id (string): The arXiv paper ID

Returns:

  • Dictionary with abstract, pdf, and source URLs

Related MCP server: Paper Search MCP Server

Installation

This project uses uv for fast, reliable Python package management.

Install uv (if not already installed):

curl -LsSf https://astral.sh/uv/install.sh | sh

Install project dependencies:

uv sync

That's it! uv will automatically:

  • Install the correct Python version (3.11)

  • Create a virtual environment

  • Install all dependencies including dev dependencies

Usage

Running the Server

Development mode (with auto-reload and inspector):

uv run fastmcp dev inspector server.py

This starts the server with:

  • Auto-reload when you save changes

  • MCP Inspector in your browser for interactive testing

  • Better debugging output

Production mode:

uv run python server.py

With make:

make run

Using with Claude Desktop

Add this server to your Claude Desktop configuration:

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

  • Linux: ~/.config/claude/claude_desktop_config.json

{
  "mcpServers": {
    "arxiv": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/absolute/path/to/arxiv-mcp-ng",
        "arxiv-mcp-ng"
      ]
    }
  }
}

Replace /absolute/path/to/arxiv-mcp-ng with the actual path to this directory.

After updating the configuration, restart Claude Desktop for the changes to take effect.

Using with MCP Client

A complete example client is provided in example_client.py. Run it to see the server in action:

python example_client.py

This will:

  1. Connect to the MCP server

  2. List all available tools

  3. Convert the "Attention Is All You Need" paper to Markdown

  4. Extract the arXiv ID from a URL

  5. Build URLs from an arXiv ID

You can also use the FastMCP client programmatically:

from fastmcp.client import Client
import asyncio

async def main():
    async with Client("python server.py") as client:
        # Convert a paper
        result = await client.call_tool(
            "convert_arxiv_to_markdown",
            {"arxiv_url": "https://arxiv.org/abs/1706.03762"}
        )

        paper_data = result.data
        with open("paper.md", "w") as f:
            f.write(paper_data['markdown'])

asyncio.run(main())

Rate Limiting

This server implements centralized rate limiting to prevent overwhelming the arXiv servers and ensure fair usage:

  • Limit: 3 requests per second across all tools

  • Scope: Server-wide (applies to all tool invocations combined)

  • Implementation: Uses a sliding window algorithm with async/await

  • Behavior: When the limit is exceeded, requests are automatically queued and delayed

The rate limiting is transparent to clients - your requests will simply wait if necessary. You'll see log messages indicating when rate limiting is active:

Rate limit reached (3 requests/1.0s). Waiting 0.45 seconds...

All three tools (convert_arxiv_to_markdown, extract_arxiv_id, build_arxiv_urls) share the same rate limit pool, ensuring the server never exceeds 3 requests per second regardless of which tools are called.

Limitations

Based on the arxiv2md.org API:

  • Papers without LaTeX source cannot be converted (some older papers only have PDFs)

  • Large papers may require significant processing time (up to 2 minutes)

  • Requires internet connection to access the API

Requirements

  • Python >= 3.11

  • fastmcp >= 3.0.0

  • httpx >= 0.27.0

  • pydantic >= 2.0.0

License

MIT

Available Tools

6 tools
build_arxiv_urlsBuild Arxiv UrlsA

Build various arXiv URLs from a paper ID.

Given an arXiv paper ID, this tool generates all common URL formats for the paper.

Rate Limited: This tool is subject to a server-wide rate limit of 3 requests/second.

Inputs: arxiv_id (required). Returns: object

convert_arxiv_to_markdownConvert Arxiv To MarkdownA

Convert an arXiv paper to Markdown format using the arxiv2md.org API.

This tool takes an arXiv URL and converts the paper to Markdown format via the arxiv2md.org API service.

Rate Limited: This tool is subject to a server-wide rate limit of 3 requests/second.

Inputs: arxiv_url (required), sections, remove_toc, remove_refs, +2 more. Returns: Result of arxiv paper conversion

extract_arxiv_idExtract Arxiv IdA

Extract the arXiv ID from a given arXiv URL.

This utility tool extracts the paper ID from various arXiv URL formats.

Rate Limited: This tool is subject to a server-wide rate limit of 3 requests/second.

Inputs: arxiv_url (required). Returns: result

get_author_papersGet Author PapersB

Retrieve papers by a specific author from arXiv.

This tool queries the arXiv API for papers authored by the given name. Results are returned in order of relevance/recency.

Rate Limited: This tool is subject to a server-wide rate limit of 3 requests/second.

Inputs: author_name (required), max_results. Returns: result

search_arxivSearch ArxivB

Search for arXiv papers by keyword, category, and date range.

This tool queries the arXiv API to find papers matching the given search criteria. Results are returned in order of relevance.

Rate Limited: This tool is subject to a server-wide rate limit of 3 requests/second.

Inputs: query (required), date_to, category, date_from, +1 more. Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.7/5.0

    Scored across 6 tools

    Disambiguation5/5

    Each tool has a distinct purpose: ID extraction, URL building, Markdown conversion, keyword search, author search, and related-paper discovery. The boundaries are clear despite several tools operating on arXiv URLs or IDs.

    Naming Consistency4/5

    Names are consistently snake_case and mostly follow a verb_noun pattern (extract_arxiv_id, build_arxiv_urls, search_arxiv, get_author_papers, get_related_papers). The only minor deviation is convert_arxiv_to_markdown, which uses a longer verb_object_preposition_object form.

    Tool Count5/5

    Six tools is well within the typical 3-15 range and each tool earns its place for an arXiv-focused server. The set covers URL utilities, conversion, and several retrieval modes without being bloated.

    Completeness3/5

    The server covers search, author lookup, related papers, URL building, ID extraction, and Markdown conversion, but lacks a direct get_paper_by_id or get_paper_metadata operation. Agents cannot easily retrieve a specific paper's metadata/abstract by ID, which is a notable gap.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      D
      maintenance
      Enables searching, downloading, and managing academic papers from arXiv.org through natural language interactions. Provides tools for paper discovery, PDF downloads, and local paper collection management.
      4
      1
      MIT
    • F
      license
      A
      quality
      D
      maintenance
      Enables searching, downloading, and reading academic papers from multiple platforms including arXiv, Semantic Scholar, PubMed, bioRxiv, medRxiv, IACR, Google Scholar, RePEc/IDEAS, and Sci-Hub with PDF to Markdown conversion.
      29
      7
      -
    • A
      license
      B
      quality
      B
      maintenance
      Enables searching and retrieving academic papers from arXiv with support for advanced filtering by author, category, and date, plus full paper content extraction.
      12
      16
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables searching arXiv papers and retrieving full paper content as Markdown, with smart sorting, category filtering, and pagination.
      10
      MIT