7 major academic databases, one search covers them all#

Supports 7 databases: arXiv, DBLP, Semantic Scholar, Crossref, OpenAlex, PubMed, and Baidu Scholar. One query automatically deduplicates, scores, and sorts the results, saving you the hassle of comparing across platforms.

Automatic Chinese/English detection#

Enter a Chinese query and the system automatically detects it, searching both Baidu Scholar and English databases at the same time, returning results mixed at a 50/50 ratio. English input takes an English-database-first strategy. No need to switch languages manually — search in whatever you type.

LLM-driven query planning#

This is the smartest part — it calls a large language model via the Groq API to convert natural language queries into precise boolean search expressions. For example, if you type “medical imaging diagnosis based on deep learning”, the LLM automatically generates a search statement like "deep learning" AND "medical image" AND "diagnosis", greatly improving retrieval quality.

No Groq API key? No problem — a built-in Chinese academic dictionary serves as a fallback, so it still works. However, for Chinese searches, the results are far worse than with a Groq API.

Automatic routing for medical queries#

When biomedical-related keywords are detected, PubMed is queried first automatically, ensuring medical literature isn’t missed.

Intelligent scoring and ranking#

Results are scored and ranked by comprehensively considering title relevance, source weight, publication year, citation count, and other dimensions. arXiv and PubMed carry the highest weight, and papers from the last 3 years get extra points.

One-click export#

Search results can be exported in both CSV and BibTeX formats. BibTeX can be imported directly into LaTeX for writing papers, which is very convenient.

1. Deploy to Cloudflare Workers (optional)#

1
git clone https://github.com/Tokisaki-Galaxy/bibliographySearchMCP.git
2
cd bibliographySearchMCP
3
npm install
4
npx wrangler login
5
npx wrangler secret put GROQ_API_KEY  # 可选，用于 LLM 查询规划
6
npx wrangler secret put BAIDU_API_KEY  # 可选，用于百度学术
7
npm run deploy

2. Configure in an MCP client#

Using Claude Desktop as an example, add the following to the configuration file:

1
{
2
  "mcpServers": {
3
    "bibliography-search": {
4
      "url": "https://bib.api.tski.uk/mcp"
5
    }
6
  }
7
}

3. Search directly in natural language#

Once configured, just say it to your AI assistant:

Find me the latest papers on transformer for time series prediction

The AI automatically calls the search_papers tool and returns a structured list of papers, including title, authors, year, abstract, PDF link, and more.

Don’t want to deploy? Use API key passthrough#

If you don’t want to deploy, you can pass your own API keys in the MCP request headers:

1
x-groq-api-key: gsk_your_key_here
2
x-baidu-api-key: your_baidu_key