111

zhuxiaoxia/111

Hosted

0 Stars 240 次浏览更新于 2026-08-23

一个通过arXiv API提供访问权限的MCP服务器，允许LLM客户端（如Claude Desktop）根据标题和摘要内容搜索arXiv上的科学论文，获取论文元数据和摘要，并访问可用论文格式（PDF/HTML）的链接。

学术研究

MCP 服务配置

复制以下 JSON 到 OPClaw 或其他 MCP 客户端的配置文件中即可使用

{
  "mcpServers": {
    "simple-arxiv": {
      "args": [
        "-m",
        "mcp_simple_arxiv"
      ],
      "command": "C:\\Users\\YOUR_USERNAME\\AppData\\Local\\Programs\\Python\\Python311\\python.exe"
    }
  }
}

可用工具（4 个）

该服务在 MCP 协议中暴露的工具，AI 可按需调用

search_papers 2 个参数需填 1 项 Search for papers on arXiv by title and abstract content. You can use advanced search syntax: - Search in title: ti:"search terms" - Search in abstract: abs:"search terms" - Search by author: au:"author name" - Combine terms with: AND, OR, ANDNOT - Filter by category: cat:cs.AI (use list_categories tool to see available categories) Examples: - "machine learning" (searches all fields) - ti:"neural networks" AND cat:cs.AI (title with category) - au:bengio AND ti:"deep learning" (author and title)

Search for papers on arXiv by title and abstract content. You can use advanced search syntax: - Search in title: ti:"search terms" - Search in abstract: abs:"search terms" - Search by author: au:"author name" - Combine terms with: AND, OR, ANDNOT - Filter by category: cat:cs.AI (use list_categories tool to see available categories) Examples: - "machine learning" (searches all fields) - ti:"neural networks" AND cat:cs.AI (title with category) - au:bengio AND ti:"deep learning" (author and title)

必填参数：query

get_paper_data 1 个参数需填 1 项 Get detailed information about a specific paper including abstract and available formats.

Get detailed information about a specific paper including abstract and available formats.

必填参数：paper_id

list_categories 1 个参数 List all available arXiv categories and how to use them in search.

List all available arXiv categories and how to use them in search.

该工具无需必填参数，直接调用即可

update_categories Update the stored category taxonomy by fetching the latest version from arxiv.org

Update the stored category taxonomy by fetching the latest version from arxiv.org

该工具无需必填参数，直接调用即可

服务介绍

mcp-simple-arxiv

An MCP server that provides access to arXiv papers through their API.

Features

This server allows LLM clients (like Claude Desktop) to:

Search for scientific papers on arXiv by title and abstract content
Get paper metadata and abstracts
Access links to available paper formats (PDF/HTML)

To use the web version just add this connector to Claude.ai https://mcp.andybrandt.net/arxiv .
You can also install & use it locally.

Installation and Deployment

This server can be run in two modes: as a local stdio server for desktop clients or as a network-accessible web server.

Installing via Smithery

To install Simple Arxiv for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install mcp-simple-arxiv --client claude

Manual Installation

pip install mcp-simple-arxiv

Usage with Claude Desktop

Add this configuration to your claude_desktop_config.json:

(Mac OS)

{
  "mcpServers": {
    "simple-arxiv": {
      "command": "python",
      "args": ["-m", "mcp_simple_arxiv"]
      }
  }
}

(Windows version):

{
  "mcpServers": {
    "simple-arxiv": {
      "command": "C:\\Users\\YOUR_USERNAME\\AppData\\Local\\Programs\\Python\\Python311\\python.exe",
      "args": [
        "-m",
        "mcp_simple_arxiv"
      ]
    }
  }
}

After restarting Claude Desktop, the following capabilities will be available:

Searching Papers

You can ask Claude to search for papers using queries like:

Can you search arXiv for recent papers about large language models?

The search will return basic information about matching papers including:

Paper title
Authors
arXiv ID
Publication date

Getting Paper Details

Once you have a paper ID, you can ask for more details:

Can you show me the details for paper 2103.08220?

This will return:

Full paper title
Authors
Publication and update dates
Journal reference (if available)
Paper abstract
Links to available formats (PDF/HTML)

For web deployment see DEPLOYMENT.md.

Development

To install for development:

git clone https://github.com/andybrandt/mcp-simple-arxiv
cd mcp-simple-arxiv
pip install -e .

arXiv API Guidelines

This server follows arXiv API usage guidelines:

Rate limiting to max 1 request per 3 seconds
Single connection at a time
Proper error handling and retry logic

License

MIT

查看源码项目主页