DOI Citation Verifier MCP Server

A Model Context Protocol server that prevents citation hallucination by verifying academic citations against multiple authoritative databases. Forked from tfscharff/doi-mcp and trimmed for atmospheric science / geophysical research workflows.

The Problem

Large language models sometimes hallucinate academic citations — fabricating papers that don't exist, misattributing titles to wrong authors, or inventing DOIs. During thesis writing with AI assistance, this is a serious risk.

This MCP server addresses the first layer of citation integrity: does the paper actually exist?

What This Tool Does (and Doesn't)

Does:

• Verifies that a paper with given title/author/year/DOI exists in real databases
• Returns the verified DOI so you can add the paper to your reference manager
• Searches for real papers on a topic when you need citations

Doesn't:

• Verify that a paper says what you're citing it for (that requires reading the paper)
• Replace your reference manager (Zotero, etc.)
• Fetch or manage BibTeX entries

Complementary Tools

This MCP complements local library tools rather than replacing them:

Tool	Purpose	Searches
doi-mcp	Verify a paper exists, get DOI	External databases (CrossRef, OpenAlex, Semantic Scholar)
semq	Find conceptually related papers you already have	Your Zotero library (semantic embeddings)
qoll-fd	Find exact phrases/terms in papers you already have	Your Zotero PDFs (full-text search)

Typical workflow:

1. AI agent suggests citing "Smith et al. (2019)" during writing
2. doi-mcp → verifyCitation({title: "...", authors: ["Smith"], year: 2019}) → confirmed, DOI returned
3. Add DOI to Zotero
4. semq/qoll-fd → verify the paper actually supports the specific claim

Databases

Three databases are queried in parallel (~1 second total):

• CrossRef — 150M+ scholarly publications, the DOI registry
• OpenAlex — 250M+ scholarly works, broad coverage
• Semantic Scholar — 200M+ papers, AI-powered search

These three cover essentially all published academic work across all disciplines. No API keys required.

Tools

verifyCitation

Verifies a citation exists before it can be mentioned.

Input:

• title (string, optional): Paper title (partial matches accepted)
• authors (array, optional): Author names (last names sufficient)
• year (number, optional): Publication year
• doi (string, optional): DOI if known
• journal (string, optional): Journal name

Returns: verified: true/false, with DOI, metadata, and confidence level if verified.

batchVerifyCitations

Verify multiple citations in a single call.

findVerifiedPapers

Search for real papers on a topic.

Input:

• query (string): Search query
• source (string, optional): "all" (default), "crossref", "openalex", or "semanticscholar"
• limit (number, optional): Results per source (1-20, default: 5)
• yearFrom / yearTo (number, optional): Publication year range

Installation

With npx (from this fork)

{
  "mcpServers": {
    "doi-mcp": {
      "command": "npx",
      "args": ["-y", "github:nbCloud91/doi-mcp"]
    }
  }
}

Local clone

git clone git@github.com:nbCloud91/doi-mcp.git
cd doi-mcp
npm install
npm run build

Config:

{
  "mcpServers": {
    "doi-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/doi-mcp/dist/index.js"]
    }
  }
}

Changes from Upstream

• Removed 6 irrelevant APIs: ERIC (education), DBLP (computer science), INSPIRE-HEP (particle physics), PubMed (biomedical), zbMATH (mathematics), HAL (French archives). Reduces latency and noise for non-biomedical/non-CS research.
• Improved match scoring: Replaced crude 30-character substring match with token-overlap scoring (normalize titles, tokenize, compute overlap ratio). More robust against minor title variations.
• Kept: CrossRef, OpenAlex, Semantic Scholar — sufficient coverage for atmospheric science and most other fields.

License

MIT