arXiv MCP Server

English · 中文

English

An MCP (Model Context Protocol) server that lets LLMs like Claude search and fetch arXiv papers directly.

Features

• Fetch recent papers from any arXiv category and store them locally
• List papers by date — titles only, to keep context lean
• Keyword search across titles and abstracts
• Retrieve full abstracts on demand for papers of interest
• Database statistics by publication date

Tools

| Tool | Parameters | Description | |------|-----------|-------------| | fetch_papers | category, num_days=3, max_results=100 | Fetch recent papers from arXiv API and store in local database | | list_papers | date="", max_results=100 | List paper titles from local database, optionally filtered by date | | search_papers | query, max_results=100 | Keyword search against titles and abstracts, returns titles only | | get_paper_details | entry_ids: list | Get full details including abstracts for specific papers | | get_stats | — | Paper counts grouped by publication date |

Design principle: fetch_papers, list_papers, and search_papers return only titles and metadata to save context. Call get_paper_details when you need abstracts for specific papers.

Installation

Claude Code (recommended)

Add to your ~/.claude.json under mcpServers:

{
  "mcpServers": {
    "arxiv": {
      "command": "uvx",
      "args": ["arxiv-today-mcp"],
      "type": "stdio"
    }
  }
}

Claude Desktop

Add to your claude_desktop_config.json:

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

Usage Example

1. Browse today's papers: Ask Claude "What's new in cs.AI today?" — Claude will call fetch_papers then list_papers, returning titles grouped by topic.
2. Get details: "Tell me more about papers 2 and 5" — Claude will call get_paper_details with those entry IDs and summarize the abstracts.

Configuration

The database path can be overridden with an environment variable:

{
  "mcpServers": {
    "arxiv": {
      "command": "uvx",
      "args": ["arxiv-today-mcp"],
      "env": {
        "ARXIV_MCP_DB_PATH": "/your/custom/path/papers.db"
      }
    }
  }
}

Requirements

• Python >= 3.10
• uv or uvx

中文

一个 MCP（模型上下文协议）服务器，让 Claude 等 LLM 能够直接搜索和获取 arXiv 论文。

功能特性

• 从任意 arXiv 分类抓取最新论文并存储到本地数据库
• 按日期浏览论文——只返回标题，节省上下文
• 对标题和摘要进行关键词搜索
• 按需获取感兴趣论文的完整摘要
• 按发布日期统计数据库中的论文数量

工具列表

| 工具 | 参数 | 功能 | |------|------|------| | fetch_papers | category, num_days=3, max_results=100 | 从 arXiv API 抓取最新论文，存入本地数据库 | | list_papers | date="", max_results=100 | 列出本地数据库中的论文标题，可按日期筛选 | | search_papers | query, max_results=100 | 在标题和摘要中关键词搜索，只返回标题 | | get_paper_details | entry_ids: list | 获取指定论文的完整信息（含摘要） | | get_stats | — | 按发布日期统计论文数量 |

设计原则：fetch_papers、list_papers、search_papers 只返回标题和元数据以节省上下文。需要摘要时再调用 get_paper_details。

安装配置

Claude Code（推荐）

在 ~/.claude.json 的 mcpServers 中添加：

{
  "mcpServers": {
    "arxiv": {
      "command": "uvx",
      "args": ["arxiv-today-mcp"],
      "type": "stdio"
    }
  }
}

Claude Desktop

在 claude_desktop_config.json 中添加：

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

使用示例

1. 浏览今日论文：问 Claude "今天 cs.AI 有什么新论文？"——Claude 会调用 fetch_papers 和 list_papers，按主题分类返回标题。
2. 获取详情："介绍一下第 2、5 篇"——Claude 会调用 get_paper_details 获取摘要并进行解读。

自定义数据库路径

通过环境变量覆盖默认数据库路径：

{
  "mcpServers": {
    "arxiv": {
      "command": "uvx",
      "args": ["arxiv-today-mcp"],
      "env": {
        "ARXIV_MCP_DB_PATH": "/你的/自定义/路径/papers.db"
      }
    }
  }
}

依赖要求

• Python >= 3.10
• uv 或 uvx