返回 MCP 目录
public公开dns本地运行

paperpal:论文伙伴

MCP扩展程序为大型语言模型(LLMs)提供对arXiv和Hugging Face论文的访问,使用户能够通过自然对话讨论论文、搜索新研究和组织文献综述。

article

README

paperpal

MCP Extension to aid you in searching and writing literature reviews

Check out this conversation with Claude to see what it can do

How it works

paperpal gives your LLMs access to arxiv and Hugging Face papers. You can then have a natural conversation with your favourite LLMs (e.g. Claude) and have it guide you.

You can:

  • Discuss papers
  • Look for new papers
  • Organize ideas for liteature reviews
  • etc.

Of course, this tool is as good as the sum of its parts. LLMs can still hallucinate, and semantic search is never perfect.

Quickstart

There are many different ways with which you can interact with an MCP server.

Claude Desktop App

If this is your first time using an MCP server for Claude Desktop App, see https://modelcontextprotocol.io/quickstart/user

First, clone this repository locally:

git clone https://github.com/jerpint/paperpal

Next, add the extension to your app. Open your configuration file (on macOS this should be ~/Library/Application Support/Claude/claude_desktop_config.json) and and add the following to the extension:

For example on MacOS:

{
  "mcpServers": {
    "paperpal": {
      "command": "uv",
      "args": [
        "--directory",
        "/Users/<username>/paperpal",
        "run",
        "paperpal.py"
      ]
    }
  }
}

Restart your Claude Desktop App and you should see it appear.

Cursor

If this is your first time using an MCP server for Cursor, see https://docs.cursor.com/context/model-context-protocol#remote-development

First, clone this repository locally:

git clone https://github.com/jerpint/paperpal

Add this to the root of the project in a .cursor/mcp.json file:

{
  "mcpServers": {
    "paperpal": {
      "command": "/Users/jeremypinto/.cargo/bin/uv",
      "args": [
        "--directory",
        "/Users/jeremypinto/paperpal",
        "run",
        "paperpal.py"
      ]
    }
  }
}
help

运行方式说明

cloud

托管运行

托管运行通常表示这个 MCP Server 由服务方环境承载,用户一般按页面提供的连接方式或授权流程接入,不需要在本地长期启动一个 MCP 进程

  1. 打开服务方连接页
  2. 完成授权或复制端点
  3. 在 MCP 客户端中连接
terminal

本地运行 / 其它方式

本地运行通常需要用户在自己的电脑或服务器上安装依赖,把 server_config 复制到 MCP 客户端,并按 env_schema 补齐环境变量、密钥或其它配置

  1. 复制 server_config
  2. 安装所需依赖
  3. 补齐环境变量后重启客户端