ai4scholar-mcp 0.4.0

A MCP server for searching and downloading academic papers from multiple sources.

AI4Scholar MCP

English | 中文

An MCP server for academic paper search, download, and analysis. Connects LLM clients (Claude Desktop, etc.) to arXiv, PubMed, Semantic Scholar, bioRxiv, and medRxiv — with citation tracking, author lookup, and paper recommendations.

---

Tools (28)

For detailed parameters and usage examples, see Tools Reference.

Platform	Tools	Capabilities
arXiv	search_arxiv, download_arxiv, read_arxiv_paper	Search, download PDF, extract text
PubMed	search_pubmed, get_pubmed_paper_detail, get_pubmed_citations, get_pubmed_related	Search, paper detail, citations, related papers
Semantic Scholar	search_semantic, download_semantic, read_semantic_paper, get_semantic_citations, get_semantic_references, search_semantic_authors, get_semantic_author_papers, get_semantic_recommendations, get_semantic_recommendations_for_paper, search_semantic_snippets	Search, download PDF, extract text, citation graph, author search, recommendations, full-text snippet search
Google Scholar	search_google_scholar	Search with year filtering
bioRxiv	search_biorxiv, download_biorxiv, read_biorxiv_paper	Search, download PDF, extract text
medRxiv	search_medrxiv, download_medrxiv, read_medrxiv_paper	Search, download PDF, extract text
General	download_pdf_by_doi	Download PDF by DOI (supports campus network institutional access)
Auto-Cite	auto_cite	Automatically annotate academic text with citations and references
Nano Draw	nano_generate, nano_edit	AI scientific figure generation, editing, style transfer, multi-image composition

Note: PDF download tools (download_*, download_pdf_by_doi) are only available in local mode (stdio). In remote SSE mode, only search and text extraction are supported. See Local Mode Guide for setup instructions.

---

Installation

Quick Start

uv add ai4scholar-mcp

Configure Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json on Mac):

{
  "mcpServers": {
    "ai4scholar_server": {
      "command": "uv",
      "args": [
        "run",
        "--directory",
        "/path/to/your/ai4s-mcp",
        "-m",
        "ai4scholar_mcp.server"
      ],
      "env": {
        "AI4SCHOLAR_API_KEY": "your-api-key"
      }
    }
  }
}

Replace /path/to/your/ai4s-mcp with your actual path. AI4SCHOLAR_API_KEY is required for PubMed and Semantic Scholar tools.

Local Usage (stdio)

Install via pip and run locally. Supports PDF download, and on campus networks you can leverage institutional access for paywalled papers.

pip install ai4scholar-mcp

For detailed setup instructions (Cursor, Claude Desktop, Claude Code, VS Code, Gemini CLI, TRAE, etc.), see Local Mode Guide.

Remote Usage (SSE)

No installation required — connect directly to the hosted SSE server using your ai4scholar.net API key.

Cursor

Go to Settings → MCP and add a new server. Or edit .cursor/mcp.json in your project:

{
  "mcpServers": {
    "ai4scholar": {
      "url": "https://mcp.ai4scholar.net/sse",
      "headers": {
        "Authorization": "Bearer <your-ai4scholar-api-key>"
      }
    }
  }
}

TRAE

Go to MCP → Add → Manual Setup, paste the following JSON:

{
  "mcpServers": {
    "ai4scholar": {
      "type": "sse",
      "url": "https://mcp.ai4scholar.net/sse",
      "headers": {
        "Authorization": "Bearer <your-ai4scholar-api-key>"
      }
    }
  }
}

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (Mac) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "ai4scholar": {
      "url": "https://mcp.ai4scholar.net/sse",
      "headers": {
        "Authorization": "Bearer <your-ai4scholar-api-key>"
      }
    }
  }
}

Claude Code

Recommended: use stdio mode (requires installation):

pip install ai4scholar-mcp
claude mcp add ai4scholar --scope user --transport stdio -e AI4SCHOLAR_API_KEY=<your-ai4scholar-api-key> -- python -m ai4scholar_mcp.server

Note: --scope user makes the config global across all projects. Claude Code's SSE mode has a known issue, so stdio is recommended.

Cherry Studio

1. Go to Settings → MCP Servers → Add Server
2. Select type: SSE
3. URL: https://mcp.ai4scholar.net/sse
4. Headers: Authorization=Bearer <your-ai4scholar-api-key>
5. Save and enable the server

Chatbox

Edit the config file (~/.config/chatbox/config.json on Linux/Mac, %APPDATA%\chatbox\config.json on Windows), and add to mcpServers:

{
  "mcpServers": {
    "ai4scholar": {
      "url": "https://mcp.ai4scholar.net/sse",
      "headers": {
        "Authorization": "Bearer <your-ai4scholar-api-key>"
      }
    }
  }
}

Dify

1. Go to Tools → MCP → Add MCP Server (HTTP)
2. Fill in:
   • Server URL: https://mcp.ai4scholar.net/sse
   • Name: ai4scholar
   • Headers: Authorization: Bearer <your-ai4scholar-api-key>
3. Save. The tools will be available in your Workflows and Agents.

Gemini CLI

Method 1 — Run in your terminal:

gemini mcp add --transport sse ai4scholar https://mcp.ai4scholar.net/sse --header "Authorization: Bearer <your-ai4scholar-api-key>"

Method 2 — Edit ~/.gemini/settings.json (or project-level .gemini/settings.json):

{
  "mcpServers": {
    "ai4scholar": {
      "url": "https://mcp.ai4scholar.net/sse",
      "headers": {
        "Authorization": "Bearer <your-ai4scholar-api-key>"
      }
    }
  }
}

Kimi Code

Method 1 — Run in your terminal:

kimi mcp add --transport http ai4scholar https://mcp.ai4scholar.net/sse \
  --header "Authorization: Bearer <your-ai4scholar-api-key>"

Method 2 — Edit ~/.kimi/mcp.json:

{
  "mcpServers": {
    "ai4scholar": {
      "url": "https://mcp.ai4scholar.net/sse",
      "headers": {
        "Authorization": "Bearer <your-ai4scholar-api-key>"
      }
    }
  }
}

Verify the connection with kimi mcp test ai4scholar.

OpenCode

Edit your project opencode.json (or global config ~/.config/opencode/opencode.json), add to the mcp object:

{
  "mcp": {
    "ai4scholar": {
      "type": "remote",
      "url": "https://mcp.ai4scholar.net/sse",
      "enabled": true,
      "oauth": false,
      "headers": {
        "Authorization": "Bearer <your-ai4scholar-api-key>"
      }
    }
  }
}

Note: "oauth": false disables automatic OAuth detection so it uses the Bearer token from headers directly.

Verify with opencode mcp debug ai4scholar.

CC-Switch

CC-Switch is a desktop configuration manager for Claude Code / Codex / Gemini CLI with a built-in MCP management panel:

1. Click the MCP button in the top-right corner
2. Add a server and select SSE as the transport type
3. Fill in:
   • URL: https://mcp.ai4scholar.net/sse
   • Headers: Authorization: Bearer <your-ai4scholar-api-key>
4. Choose the clients to sync (Claude Code / Gemini CLI, etc.) and enable

Replace <your-ai4scholar-api-key> with your own API key from ai4scholar.net.

OpenClaw

Edit ~/.openclaw/config.json and add to mcpServers:

{
  "mcpServers": {
    "ai4scholar": {
      "url": "https://mcp.ai4scholar.net/sse",
      "headers": {
        "Authorization": "Bearer <your-ai4scholar-api-key>"
      }
    }
  }
}

For Development

git clone https://github.com/literaf/ai4s-mcp.git
cd ai4s-mcp
uv venv && source .venv/bin/activate
uv add -e .

---

Contributing

1. Fork the repository
2. Create a feature branch
3. Add new platforms in academic_platforms/, update tests in tests/
4. Submit a Pull Request

---

License

MIT License. See LICENSE for details.