Scientific Search MCP for Cursor

[!WARNING] Vibecoded by AI.

This is a Model Control Protocol (MCP) implementation for Cursor that provides scientific search capabilities through various academic search engines.

Features

• Integration with Cursor using the MCP protocol
• Scientific search across multiple engines:
  • DuckDuckGo (general web search)
  • Google Scholar (academic papers search)
  • PubMed (biomedical literature search)
  • arXiv (physics, mathematics, computer science, etc.)
  • Semantic Scholar (AI-powered research paper search via Tor proxy)

Installation

1. Clone this repository:

git clone https://github.com/yourusername/cursor-scientific-search.git
cd cursor-scientific-search

2. Install the required dependencies:

pip install -r requirements.txt

For specific search engines, you may need to install additional dependencies:

# For arXiv search
pip install arxiv

# For Google Scholar
pip install scholarly

# For PubMed
pip install pymed

# For Semantic Scholar (requires Tor)
pip install requests[socks]

Tor Setup for Semantic Scholar

The Semantic Scholar search uses Tor to bypass rate limits and IP blocks. To use this feature:

1. Install Tor:

   • Linux: sudo apt install tor (Debian/Ubuntu) or sudo pacman -S tor (Arch)
   • macOS: brew install tor
   • Windows: Download and install the Tor Browser which includes the Tor service

2. Start the Tor service:

   • Linux: sudo systemctl start tor or sudo service tor start
   • macOS: brew services start tor
   • Windows: The Tor Browser includes the service, or you can use Tor Expert Bundle

3. Verify Tor is running on port 9050 (default):

   nc -z localhost 9050 && echo "Tor is running"
   

4. The scientific search MCP will automatically route Semantic Scholar requests through Tor.

Usage

1. Start the MCP server:

python scientific_search_mcp.py

2. In Cursor, connect to the MCP server.

3. Use the scientific search tools:

Search Scientific

Search for scientific papers across different engines:

# Search arXiv for quantum computing papers
results = await mcp_scientific_search_search_scientific(
    query="quantum computing", 
    engine="arxiv"
)

# Search PubMed for COVID-19 research
results = await mcp_scientific_search_search_scientific(
    query="COVID-19 treatment", 
    engine="pubmed"
)

# Search Google Scholar for machine learning papers
results = await mcp_scientific_search_search_scientific(
    query="transformers neural networks", 
    engine="google_scholar"
)

# Search Semantic Scholar for AI ethics papers (via Tor)
results = await mcp_scientific_search_search_scientific(
    query="AI ethics", 
    engine="semantic_scholar"
)

Get URL Content

Extract content from a web page:

content = await mcp_scientific_search_get_url_content(
    url="https://example.com/research-paper"
)

Grep URL Content

Search for specific text patterns in a web page:

matches = await mcp_scientific_search_grep_url_content(
    url="https://example.com/research-paper",
    pattern="neural network"
)

Process PDF

Extract text and images from a PDF:

pdf_content = await mcp_scientific_search_process_pdf(
    url_or_path="https://example.com/paper.pdf",
    extract_images=True
)

Notes

• Google Scholar searches may be rate-limited if used excessively. The implementation uses free proxies to reduce the risk of blocking.
• PubMed requires an email address for usage tracking. Set the PUBMED_EMAIL environment variable or it will use a default.
• Semantic Scholar searches are routed through Tor to avoid rate limits and IP blocking.
• For best results, use specific search terms relevant to the domain.

Troubleshooting

• Semantic Scholar Connection Issues: Make sure Tor is running on localhost:9050. If you experience issues, restart the Tor service.
• Slow Responses: Tor routing may cause slower responses for Semantic Scholar searches.

License

MIT License