Figshare

What Figshare MCP server does

Figshare MCP server gives AI agents access to public research outputs hosted on Figshare. Its coverage includes datasets, papers, posters, presentations, curated collections, and institution listings. Agents can search for relevant records, retrieve a specific article or collection, and inspect the files attached to an article.

The server is intended for read-oriented research workflows. For example, an agent can search for climate datasets, retrieve metadata for a matching article, and list the files available for download or further inspection.

How it works

The tools use the Figshare API at https://api.figshare.com/v2/ as their data source. You can connect through the hosted Pipeworx endpoint or run the npm package as a local MCP process. The hosted endpoint uses streamable HTTP, while the standalone package communicates through standard input and output.

The hosted pack endpoint lists the Figshare tools together with shared Pipeworx meta-tools. Those additional tools can route questions across the wider Pipeworx catalog, so the number shown by tools/list is higher than the six Figshare-specific tools. The standalone process exposes only the Figshare pack and does not add those gateway tools.

Setup and configuration

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

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

The Figshare MCP server can also run locally without a Pipeworx account or gateway round trip:

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

The package can be started directly with npx -y @pipeworx/mcp-figshare. A full Pipeworx gateway is also available at https://gateway.pipeworx.io/mcp when an agent needs the catalog's broader tool set.

Tools and capabilities

The Figshare-specific tools are:

• articles: Search articles with optional query, item type, page, page size, and ordering parameters.
• article: Retrieve one article, optionally specifying a version.
• article_files: List files associated with an article, optionally by version.
• collections: Search curated collections with pagination options.
• collection: Retrieve one collection by ID.
• institutions: List institutions.

The hosted gateway also accepts HTTP tool calls. For example, an agent or script can send a POST request to /v1/tools/articles with a JSON query instead of using an MCP client.

Limitations and notes

Article and collection lookups require the relevant IDs, while article version arguments are optional.

A scoped gateway connection always includes shared Pipeworx meta-tools. There is no documented option to remove them from that endpoint. If a smaller tool schema is important, use the standalone stdio package. The gateway's exact tool scope can change, so its initialize response is the authoritative source for the connected session.