PaperMCP 智能学术论文检索系统

欢迎使用 PaperMCP 智能学术论文检索系统！这是一个基于 Model Context Protocol (MCP) 的高级学术论文搜索服务器，专为研究员和教授设计。通过 OpenAlex API 和智能算法，为 AI 助手提供精准的学术文献检索能力，并额外提供论文 PDF 下载能力。

🌟 Features

📚 Comprehensive Paper Search

Search academic papers with flexible filtering options:

• Keyword Search - Find papers by title, abstract, or full-text content
• Country Filter - Limit results to papers from specific countries (CN, US, GB, etc.)
• Year Filter - Search papers from specific publication years
• Result Limit - Control the number of results (up to 50 papers)
• Sort Options - Sort by citation count, publication date, or relevance
• Open Access Filter - Find only freely accessible papers

📄 Browser-driven PDF Download

Download paper PDFs with better compatibility for ACM / Cloudflare-protected websites:

• Browser Driver First - Uses puppeteer-core with local Microsoft Edge / Google Chrome
• Three-field Input - url, local_path, file_name
• Auto Path Assembly - Combines local path and file name into the final PDF path
• Auto .pdf Suffix - Appends .pdf automatically when missing
• Better Compatibility - Uses browser warmup, cookies and browser-context fetch for stricter sites
• No HTML Debug File by Default - Avoids writing debug HTML files to disk

📊 Rich Paper Information

Get comprehensive details for each paper:

• Basic Info - Title, authors, publication year, document type
• Abstract - Full abstract text with intelligent reconstruction from inverted index
• Publication Details - Journal/venue, DOI, URLs
• Citation Data - Citation count and related works
• Institutional Info - Author affiliations and institutions
• Subject Classification - Topics, subfields, fields, and domains
• Open Access Status - OA status and APC (Article Processing Charge) information

🤖 MCP Integration

Seamless integration with MCP-compatible clients (like Claude) for intelligent academic research and PDF acquisition.

🚦 Requirements

Before getting started, please ensure you have:

1. Node.js:

   • Requires Node.js version >= 18
   • Download and install from nodejs.org

2. pnpm:

   • This project uses pnpm as the package manager
   • Install via npm install -g pnpm

3. Email Address:

   • Provide a valid email address for OpenAlex API access
   • OpenAlex requires an email for rate limiting and contact purposes
   • No API key needed - OpenAlex is free to use!

4. Local Browser:

   • Install Microsoft Edge or Google Chrome locally
   • paper_pdf_download will automatically detect a supported browser on Windows

🛠️ Installation & Setup

Install via Smithery (Recommended)

If you're using Claude Desktop, you can quickly install via Smithery:

npx -y @smithery/cli install @guangxiangdebizi/paper-mcp --client claude

Manual Installation

1. Get the code:

   git clone https://github.com/guangxiangdebizi/PaperMCP.git
   cd PaperMCP

2. Install dependencies:

   pnpm install

3. Configure Email Address:

   • Create a .env file in the project root directory
   • Add the following content:

     [email protected]

   • Or set it directly in the src/config.ts file

4. Build the project:

   pnpm run build

🚀 Running the Server

There are two ways to start the server:

Method 1: Using stdio mode (Direct run)

node build/index.js

Method 2: Using Supergateway (Recommended for development)

npx supergateway --stdio "node build/index.js" --port 3100

📝 Configuring MCP Clients

To use this server in Claude or other MCP clients, you need the following configuration:

Claude Configuration

Add the following to Claude's configuration file:

{
  "mcpServers": {
    "paper-search-server": {
      "url": "http://localhost:3100/sse",
      "type": "sse",
      "disabled": false,
      "autoApprove": [
        "paper_search",
        "paper_pdf_download"
      ]
    }
  }
}

If using stdio mode directly (without Supergateway), configure as follows:

{
  "mcpServers": {
    "paper-search-server": {
      "command": "C:/path/to/PaperMCP/build/index.js",
      "type": "stdio",
      "disabled": false,
      "autoApprove": [
        "paper_search",
        "paper_pdf_download"
      ]
    }
  }
}

🧰 Available Tools

paper_search

Intelligent academic paper retrieval powered by OpenAlex.

paper_pdf_download

Download a paper PDF to local disk through a browser-driven flow.

Input parameters

| Parameter | Type | Description | Example | |-----------|------|-------------|---------| | url | string | Remote PDF URL (required) | https://dlnext.acm.org/doi/pdf/10.1145/3718735 | | local_path | string | Local absolute directory path (required) | E:\\TMP | | file_name | string | Output file name; .pdf will be appended if missing | acm-3718735.pdf |

Behavior

• Final output path is built from local_path + file_name
• If file_name does not end with .pdf, the server auto-appends it
• Uses puppeteer-core + local Edge/Chrome for higher site compatibility
• Does not save HTML debug files by default

💡 Usage Examples

Here are some example queries using the PaperMCP server:

1. Basic Paper Search

You can ask Claude:

General Search:

"Search for papers about machine learning published in 2024"

2. Download a PDF

"Use paper_pdf_download to save https://dlnext.acm.org/doi/pdf/10.1145/3718735 into E:\\TMP with file name acm-3718735"

3. Research-focused Queries

"Help me find recent papers about transformer architectures for my literature review"

This server can use paper_search to retrieve paper metadata and paper_pdf_download to save accessible PDFs locally.

📊 Supported Search Parameters

The paper_search tool supports the following search parameters:

| Parameter | Type | Description | Example | |-----------|------|-------------|---------| | query | string | Search keywords (required) | "machine learning", "deep learning" | | country_code | string | Filter by country code | "CN" (China), "US" (USA), "GB" (UK) | | year | number | Filter by publication year | 2024, 2023 | | num_results | number | Number of results (max 50) | 10, 20, 50 | | sort_by | string | Sort method | "cited_by_count", "publication_date", "relevance_score" | | open_access | boolean | Filter open access papers | true, false |

📈 Data Sources

This server uses the OpenAlex API, which provides:

• 260M+ papers from across all disciplines
• Real-time updates with new publications
• Comprehensive metadata including citations, authors, institutions
• Open access information and APC data
• Subject classification at multiple levels
• Institution and country data for geographic analysis

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.

👨‍💻 Author

• Name: Xingyu_Chen
• Email: [email protected]
• GitHub: guangxiangdebizi

🙏 Acknowledgments

This project uses the OpenAlex API, a free and open catalog of scholarly papers, authors, institutions, and more. Special thanks to the OpenAlex team for providing this invaluable resource to the research community.