Zenodo

What the Zenodo MCP server does

The Zenodo MCP server connects MCP clients to Zenodo, CERN’s open research data repository. It supports read operations across research records and curated communities, covering datasets, software, presentations, papers, and other items published with DOIs.

Its record search accepts a query and optional filters such as record type, community, sort order, page number, and page size. Once a record is identified, separate tools provide its details and the files attached to it. Community tools support finding communities and listing the records collected in a specific community.

The server is suited to agents that need to locate public Zenodo material, inspect metadata, identify downloadable files, or explore a curated collection without implementing the Zenodo API directly.

How it works

The Zenodo MCP server reads from https://zenodo.org/api/. The available pack tools are:

• search for record queries and pagination
• record for retrieving one record by Zenodo ID
• record_files for listing files belonging to a record
• communities for finding communities
• community_records for listing records in a community

The hosted option uses the Pipeworx gateway. Its scoped MCP URL lists the Zenodo tools together with shared gateway tools, including routing and discovery functions. The gateway can also expose all catalog sources through its full endpoint. These gateway additions are not part of the local package.

Without an MCP client, individual tools can be called through the gateway’s HTTP API.

Setup and configuration

For a hosted connection, add the Pipeworx MCP URL to an MCP client configuration:

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

The Zenodo MCP server is also available as a local npm package. Run it with npx -y @pipeworx/mcp-zenodo, or configure the command and package as a stdio MCP server. This local mode does not require a Pipeworx account or gateway round trip and exposes only the five Zenodo tools.

No environment variables or API keys are documented as required.

Tools and capabilities

Use search to find records by query and narrow results with supported optional parameters. Use record after obtaining a Zenodo ID when an agent needs the single record response. Use record_files to inspect the files available from that record.

The communities tool supports optional query and pagination parameters. community_records takes a community ID and pagination parameters, allowing an agent to inspect the items grouped under a collection.

For natural-language requests, the hosted gateway can route questions through ask_pipeworx, which selects a tool and fills its arguments. This routing is available through the pack endpoint and the full Pipeworx gateway, but not through the standalone local process.

Limitations and notes

The Zenodo MCP server provides the read-oriented tools listed in the README; no write, upload, or publication operation is documented. The hosted endpoint includes shared Pipeworx schemas, so its tool list is larger than the five Zenodo-specific tools. The local stdio process avoids those shared tools and does not provide cross-catalog routing.

The gateway and local package use the same Zenodo source and pack tools, but they differ in transport and available gateway functions. The hosted configuration uses a remote URL, while the standalone configuration communicates over standard input and output.