Citation Network MCP

What Citation Network MCP MCP server does

Citation Network MCP MCP server supports literature discovery, citation tracing, network analysis, and bibliometrics through an MCP interface. The repository describes six tool families: search and records, citations, networks, audit, bibliometrics, and diagnostics. It can search Scopus, use OpenAlex as an alternative source for supported analyses, and call Crossref, Semantic Scholar, and open-access sources where individual tools require them.

The network functions cover bibliographic coupling, co-citation, citation lineages, citation networks, historiographs, RPYS, and research fronts. These functions support questions such as identifying the research front around a set of papers, following citing work across multiple generations, or finding the main path through a citation network. The server also provides publication counts, topic landscapes, topic profiles, thematic evolution, journal metrics, journal discovery, and BibTeX output.

As a tool in the 🔬 Research category, Citation Network MCP MCP server is aimed at developers and researchers building conversational workflows for literature reviews, bibliometric studies, and citation analysis.

How it works

An MCP client receives a natural-language request and selects the relevant server tools. The server uses a supplied Scopus API key for Scopus operations. OpenAlex does not require a key and can be selected with source="openalex" for analyses supported by that source. Crossref, Semantic Scholar, and open-access sources may be used by tools that need additional records or validation.

Results are returned as a conversational summary, while complete outputs are written to files. The documented formats include JSON, CSV, GraphML, Pajek, PNG, HTML, and BibTeX. Exported files can be opened by Pajek, VOSviewer, Gephi, or Zotero. Responses are cached, and longer operations can run as background jobs with separate status and result tools.

Setup and configuration

Scopus access requires an API key from the Elsevier Developer Portal. The documented environment variable is SCOPUS_API_KEY. OpenAlex use does not require an API key.

Claude Desktop can be installed from the release .mcpb package and prompts for the key during installation. Claude Code can install the repository’s plugin through its marketplace commands, then read the key from the shell or an operating-system secret store. Other MCP clients can run the server with uvx; the repository notes that uv is required for this route.

Tools and capabilities

Citation Network MCP MCP server exposes tools in these groups:

• Search and records: Scopus and cross-source search, full-text search, abstract details, identifier resolution, author search, author profiles, and full-text retrieval.
• Citations: reference lists and papers citing a selected work.
• Networks: coupling, co-citation, lineages, citation networks, historiographs, RPYS, and research fronts.
• Audit: citer resolution, citation contexts, path transmission, coding agreement, index coverage, and retraction checks.
• Bibliometrics: publication counts, topic analysis, thematic evolution, journal metrics, journal discovery, BibTeX, and record import.
• Diagnostics and jobs: connection diagnosis, quota status, server information, job status, and job results.

The repository also describes a Claude Code skill that guides a main-path workflow from a query to a drawn main-path map.

Limitations and notes

Scopus operations depend on the permissions and quota associated with the configured API key. The server includes a diagnostic tool that reports which tools the current Scopus access supports and why unsupported operations fail. OpenAlex is an alternative for supported analyses, but the documentation does not state that every Scopus tool is available through OpenAlex.

Citation Network MCP MCP server can combine data from multiple external sources, so results may depend on source coverage and identifier resolution. The repository emphasizes auditing citer sets, checking references against Crossref, and inspecting citation sentences, but these checks are tool-dependent rather than a guarantee that every result has identical coverage.

The project requires Python 3.10 or newer and is released under the MIT license.