Datacite

What Datacite does

Datacite exposes DataCite's public DOI registry to MCP-compatible AI clients. Its records cover research datasets, software, samples, and other registered outputs rather than scholarly papers. Search results can include DOI, title, creators, publisher, resource type, publication year, and citation, download, or view counts.

The integration is an independent Pipeworx implementation and is not affiliated with, endorsed by, or published by DataCite.

How it works

The hosted MCP endpoint is https://gateway.pipeworx.io/datacite/mcp. Requests are forwarded to the public DataCite JSON:API at https://api.datacite.org. A full Pipeworx gateway endpoint is also available at https://gateway.pipeworx.io/mcp.

A local npm package, @pipeworx/mcp-datacite, provides a stdio bridge. An MCP client starts the package with npx, and the bridge sends requests to the same hosted endpoint. Network access is required; the local bridge has no offline mode.

The pack endpoint lists the Datacite tools and also includes shared Pipeworx meta-tools. PIPEWORX_META_TOOLS=1 enables those shared tools in the local package, including gateway-wide routing through ask_pipeworx. The connection's initialize response defines the exact tool scope for that connection.

Setup and configuration

Add the hosted endpoint to an MCP client with a server entry such as:

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

For a local stdio connection, use:

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

Anonymous use is available and limited to 50 calls per day. A free Pipeworx account increases the allowance to 200 calls per day. Add the account key as PIPEWORX_API_KEY; the bridge sends it as a Bearer token. PIPEWORX_META_TOOLS=1 is optional and adds the shared gateway tools to the local server's tool list.

Clients without MCP support can call tools over HTTP. For example, POST https://gateway.pipeworx.io/v1/tools/search_dois accepts a JSON request, while GET https://gateway.pipeworx.io/v1/tools/search_dois can be used to inspect a tool.

Tools and capabilities

• search_dois searches DataCite records using free text, resource type, year, publisher, and affiliation filters. Results can be paginated and sorted by relevance, recency, citations, downloads, or views.
• get_doi retrieves a single record by DOI, including titles, creators, affiliations, ORCIDs, abstracts, publisher, dates, related identifiers, subjects, license, sizes, and formats.
• list_repositories finds DataCite-registered repositories, with query filtering and pagination. Examples include Zenodo, Dryad, and Figshare.

These tools support questions such as finding recently published genomics datasets, locating highly downloaded climate data, examining a DOI's licensing details, or identifying places to deposit a dataset.

Limitations and notes

DataCite is intended for datasets, software, samples, and related research outputs; paper discovery belongs with Crossref. The service depends on network access and the Pipeworx gateway. The hosted and local routes expose shared Pipeworx meta-tools alongside the pack tools, and the local package's anonymous call allowance is capped unless a key is configured.

When using an HTTPS proxy with the local bridge, Node's built-in fetch does not use HTTPS_PROXY by default. Newer Node versions can honor that variable when NODE_USE_ENV_PROXY=1 is set.

The project is licensed under MIT.