Crossref MCP — Academic DOI & Literature Metadata

Type: Apify MCP Actor (JavaScript)
Purpose: AI agent access to 120M+ academic papers, books, and conference proceedings via Crossref API
Stack: Apify SDK, Node.js 20+, MCP protocol, standby mode

Quick Start

cd ~/Projects/apify-actors/crossref-mcp
apify run          # Local development
apify push         # Deploy to Apify

Key Files

• src/main.js — MCP handler entry point with handleTool export
• .actor/actor.json — Standby mode enabled (usesStandbyMode: true)
• Dockerfile — Node.js 24 runtime
• package.json — Dependencies (apify ^3.6.0)

Architecture

• Standby MCP via handleRequest export
• Readiness probe at GET / (checks x-apify-container-server-readiness-probe header)
• PPE (Pay-Per-Event) charging for each tool invocation
• ERROR_RETRY with exponential backoff (3 retries, 2s base delay)
• Crossref API at https://api.crossref.org/works

Tools

Tool	Description	PPE
searchWorks	Search academic works by query, author, DOI, ISSN, date range	$0.03
getWorkDetails	Get full metadata for a work by DOI (authors, abstract, funding)	$0.02
getReferences	Get bibliography/reference list for a work by DOI	$0.04
exportBibTeX	Export works as BibTeX, RIS, or JSON citation format	$0.01

API Operations

searchWorks

Search for academic works by keyword, author, or filters.

{
  "operation": "searchWorks",
  "parameters": {
    "query": "machine learning",
    "author": "Hinton",
    "type": "journal-article",
    "from_date": "2020-01-01",
    "max_results": 20
  }
}

Returns: Array of works with title, authors, DOI, publication date, journal, abstract, and citation counts.

getWorkDetails

Get full metadata for a specific work by DOI.

{
  "operation": "getWorkDetails",
  "parameters": {
    "doi": "10.1038/nature12373",
    "include_references": true
  }
}

Returns: Complete metadata including authors with ORCIDs, abstract, funding info, license, references, and more.

getReferences

Get the reference list (bibliography) for a work.

{
  "operation": "getReferences",
  "parameters": {
    "doi": "10.1038/nature12373",
    "max_references": 50
  }
}

Returns: Array of references with DOI, title, author, year, journal, pages.

exportBibTeX

Export one or more works as citation strings.

{
  "operation": "exportBibTeX",
  "parameters": {
    "dois": ["10.1038/nature12373", "10.1126/science.1259856"],
    "format": "bibtex"
  }
}

Returns: Formatted citation string(s) in BibTeX, RIS, or JSON format.

Input Schema

The actor accepts an INPUT JSON with operation and parameters:

{
  "operation": "searchWorks",
  "parameters": {
    "query": "deep learning",
    "max_results": 10
  }
}

PPE Pricing

This actor uses Apify's Pay-Per-Event pricing model. Each tool invocation incurs a small charge to cover Crossref API costs:

Event	Charge
searchWorks	$0.03
getWorkDetails	$0.02
getReferences	$0.04
exportBibTeX	$0.01

Error Handling & Retry

The actor implements automatic retry with exponential backoff:

• 3 retries on API failure
• Base delay: 2 seconds, doubling each retry (2s → 4s → 8s)
• Rate limit handling: Detects 429 responses, waits for Retry-After header

Data Sources

• Crossref API: The official REST API for Crossref metadata
• Coverage: 120M+ works from 90,000+ publishers
• Types: Journal articles, books, conference proceedings, datasets, preprints, etc.

MCP Protocol

This actor implements the Model Context Protocol (MCP) for AI agent integration:

• initialize — Returns server capabilities
• tools/list — Returns available tools and schemas
• tools/call — Executes a tool by name with parameters
• handleRequest — Legacy operation handler (operation + parameters)

Deployed Actor

The actor runs at: crossref-mcp.apify.actor

Rate Limits

The Crossref API has rate limits:

• Polite pool: 50 requests/second (with email)
• Default: 25 requests/second
• Unauthenticated: Heavily rate limited

The actor uses the polite pool (mailto=apify@example.com) and implements retry on 429 responses.

Notes

• All dates are in ISO format (YYYY-MM-DD)
• DOIs are cleaned automatically (handles https://doi.org/ prefixes)
• Abstracts are stripped of HTML tags
• BibTeX output escapes LaTeX special characters
• References are capped at 100 per request to avoid timeouts