Citation Network MCP

Search the literature, trace who cites whom across generations, map research fronts and intellectual bases, and pull the counts, metrics and bibliography you need, from a conversation. It is an MCP server, so any MCP client can use it: Claude Desktop, Claude Code, Cursor.

What you get

Citation networks Coupling, co-citation, multi-generation lineages and the network within any paper set, with SPC/SPLC/SPNP main paths, key routes, RPYS and a historiograph. More	Audited, not just built Citer sets verified across search strategies, reference lists checked against Crossref, and the sentences behind each citation. More
Two data sources Scopus by default. Add source="openalex" to run the same analyses without a Scopus subscription. More	Bibliometrics Where a topic is published and in which quartile, journals above a percentile cut-off, counts per year, themes over time, BibTeX. More
Full-text search Find Elsevier papers that use a term in their body, and see how often and where each one uses it. More	Honest about access One call tells you which tools your Scopus access supports, and why the rest fail. More
Main-path skill Walks Claude from a query to a drawn main path map. Ships with the Claude Code plugin. More	Tested 510 test functions, CI on Linux, macOS and Windows, and a live check of every tool against the real APIs. More

How it compares with the other Scopus MCP servers and with bibliometrics packages (bibliometrix, pybliometrics, VOSviewer, Pajek): comparison.

Ask things like

Map the research front around these six papers on organizing visions.

Trace two generations of work citing Swanson & Ramiller (1997) and show me the main path.

Build the citation network of every Basket of Eight paper citing the organizing-vision papers, and show how the main path cites its predecessors.

How has publishing on "digital transformation" grown since 2010?

Get SJR and CiteScore for the Basket of Eight, and BibTeX for the papers we just found.

More, with the tools each one triggers: prompt examples.

How it works

You ask in plain words and the assistant picks the tools. The server calls Scopus with your own key, and Crossref, Semantic Scholar and open-access sources where a tool needs them. Results come back as a summary in the reply, and the full data goes to files that Pajek, VOSviewer, Gephi or Zotero can open. Responses are cached, and long jobs run in the background.

Tools

	Family	Tools	Guide
	Search and records	search_scopus · search_all · search_fulltext · get_abstract_details · resolve_identifier · search_authors · get_author_profile · get_fulltext	search
	Citations	get_references · get_citing_papers	search
	Networks	bibliographic_coupling · co_citation · citation_lineage · citation_network · historiograph · rpys · research_fronts	networks
	Audit	resolve_citers · citation_context · path_transmission · coding_agreement · index_coverage · check_retractions	audit
	Bibliometrics	publication_counts · topic_landscape · thematic_evolution · get_journal_metrics · find_journals · get_bibtex · import_records	bibliometrics
	Diagnostics and jobs	diagnose_connection · get_quota_status · get_server_info · job_status · job_result	access

Every parameter of every tool: tool reference.

Install

You need an API key from the Elsevier Developer Portal (register with your institutional email). OpenAlex needs no key.

Claude Desktop, one click:

1. Download scopus-plus-mcp-<version>.mcpb from the latest release.
2. Open it (or drag it into Settings → Extensions), click Install, and paste your API key when asked. Claude stores it securely.

Claude Code, two commands:

claude plugin marketplace add michalhron/scopus-plus-mcp
claude plugin install scopus-plus-mcp@michalhron

Then make your key available, either in your shell (export SCOPUS_API_KEY=...) or, better, in the OS secret store.

Any other MCP client (needs uv):

{
  "mcpServers": {
    "scopus-plus": {
      "command": "uvx",
      "args": ["scopus-plus-mcp"],
      "env": { "SCOPUS_API_KEY": "YOUR_KEY" }
    }
  }
}

Then ask your assistant to run diagnose_connection. It checks your key and tells you which tools your Scopus access supports.

The Claude Code plugin also installs the main-path skill, which walks Claude through a main path analysis from a query to a drawn map. For Claude Desktop, zip its folder and upload it as a skill.

Documentation

Feature guides:

• Search, records and full text: finding papers and authors, identifiers, importing exports, full-text search and retrieval.
• Citation networks and main paths: coupling, co-citation, lineages, main path analysis, historiograph, RPYS, research fronts.
• Audit: verified citer sets, reference-list completeness, citation contexts, the transmission audit, retractions.
• Bibliometrics and bibliography: counts per year, topic landscape, journal cut-offs and metrics, thematic evolution, BibTeX.

Reference and setup:

• Tool reference: every tool and parameter, generated from the code.
• Data sources: Scopus vs OpenAlex, when to use which, measured coverage.
• Configuration: all settings, and keeping keys in the OS secret store.
• Access and troubleshooting: off-campus access, tokens, proxies, reading diagnose_connection.
• Comparison: this project vs the other Scopus MCP servers, and vs bibliometrics packages.
• Prompt examples, development, changelog, roadmap.

Origins

Formerly citation-network-mcp (and before that michalhron/scopus-mcp). This project began as a fork of qwe4559999/scopus-mcp by thinktraveller and qwe4559999, which provides Scopus search, abstracts, author profiles and citing papers. Everything since version 0.2 was developed here by Michal Hron. MIT licensed; see LICENSE.

Scopus and ScienceDirect are trademarks of Elsevier B.V. This project is independent and not affiliated with or endorsed by Elsevier; it uses their public APIs with your own key and subscription.