Crossref

What crossref MCP server does

The crossref MCP server connects an MCP client to Crossref, a DOI registration agency with metadata for scholarly works such as journal articles, books, chapters, datasets, and proceedings. It is part of the 🔬 Research category and is intended for agents that need structured publication information rather than full-text documents.

The service can search for works, retrieve a publication by DOI, and inspect a journal’s latest publications. Crossref records may include titles, authors, ORCIDs when supplied by publishers, abstracts where available, journal and publisher details, publication dates, subjects, references, citation counts, funder information, open-access links, and relationship data.

How it works

The crossref MCP server exposes three pack-specific tools over MCP:

• search_works searches academic papers, books, and datasets. Searches can target a title, sort by citations or publication date, and filter by date range, type, author, or journal. A journal filter accepts a journal name or ISSN.
• get_work retrieves metadata for a publication using its DOI. DOI inputs should be passed in bare form, such as 10.1038/nature12373, rather than with a URL or doi: prefix.
• get_journal returns the five most recent works associated with a journal ISSN, including titles, authors, DOIs, and publication dates.

The hosted endpoint is available at https://gateway.pipeworx.io/crossref/mcp. The gateway endpoint also lists shared Pipeworx meta-tools, including ask_pipeworx, discover_tools, and related functions. A local package exposes only the Crossref tools over standard input and output. The same data source can also be called through the gateway’s HTTP tool endpoint.

Setup and configuration

For a hosted MCP connection, add the Crossref endpoint to an MCP client configuration:

{
  "mcpServers": {
    "crossref": {
      "url": "https://gateway.pipeworx.io/crossref/mcp"
    }
  }
}

The hosted route does not require an account for initial calls. Crossref itself is public and free to access, and the gateway identifies requests for its polite-pool handling.

To run locally without a gateway account, use the npm package with an MCP client that supports stdio:

{
  "mcpServers": {
    "crossref": {
      "command": "npx",
      "args": ["-y", "@pipeworx/mcp-crossref"]
    }
  }
}

The package can also be started directly with npx -y @pipeworx/mcp-crossref.

Tools and capabilities

The crossref MCP server fits DOI-to-metadata workflows, searches for a particular paper by title, and queries recent content from a known journal. It can help an agent find an author’s works, compare citation counts, restrict results to a publication period, or distinguish publication types such as journal articles and book chapters.

Crossref is often useful as a canonical DOI metadata source. Citation and reference coverage depends on publisher submissions, so an apparently empty reference list may not represent the complete published list.

Limitations and notes

ORCID data is not universal, particularly for older or smaller-publisher records. A work can also have multiple DOIs, such as a preprint and an accepted version. Check relationship fields for links between versions and for retraction notices before relying on a record in a citation workflow.

Crossref accepts normalized bare DOIs. Reference lists and citing-paper information vary by publisher, and exhaustive citation analysis may require another research source for comparison.