Bibliography Search MCP

A multi-source academic bibliography search MCP (Machine Context Protocol) server running on Cloudflare Workers. Supports both Chinese and English literature retrieval across 7 academic databases with automatic language detection, query planning, and scoring.

Features

• Multi-source search — arXiv, DBLP, Semantic Scholar, Crossref, OpenAlex, PubMed, Baidu Xueshu

• Automatic language detection — chooses appropriate sources and search strategies for Chinese vs English queries

• LLM-powered query planning — uses Groq API to build optimal boolean search queries from natural language

• Dictionary fallback — built-in Chinese academic vocabulary dictionary when LLM is unavailable

• Medical query detection — auto-routes biomedical queries to PubMed

• Deduplication & scoring — relevance-based ranking across all sources

• Export formats — CSV and BibTeX support

Related MCP server: scholar-mcp

Usage (MCP)

Configure as an MCP server in your MCP client:

{
  "mcpServers": {
    "bibliography-search": {
      "url": "https://bibliography-search-mcp.example.workers.dev/mcp"
    }
  }
}

Passing your own API keys

Send your Groq key via request header — it takes precedence over the environment variable:

x-groq-api-key: gsk_your_key_here

Similarly for Baidu Xueshu:

x-baidu-api-key: your_baidu_key

This is useful when the server is shared but each client has their own quota.

Tools

Environment

Deployment

1. Clone and install

git clone https://github.com/your-username/bibliographySearchMCP.git
cd bibliographySearchMCP
npm install

2. Login to Cloudflare

npx wrangler login

3. Set environment variables

npx wrangler secret put GROQ_API_KEY
# paste your Groq API key when prompted

npx wrangler secret put BAIDU_API_KEY
# paste your Baidu Xueshu API key when prompted

4. Deploy

npm run deploy

Your worker will be available at https://bibliography-search-mcp.your-subdomain.workers.dev/mcp.

Updating

npx wrangler deploy

Development

npm install
npm run dev        # local dev with wrangler
npm run typecheck  # type check
npm run deploy     # deploy to Cloudflare Workers

Architecture

User query → [LLM / Dict tokenizer] → search query → [arXiv | DBLP | Semantic Scholar | Crossref | OpenAlex | PubMed | Baidu]
                                                                                                                          ↓
                                                                                              [Deduplicate → Score → Rank]
                                                                                                                          ↓
                                                                                                              Result (JSON / CSV / BibTeX)