HF Dataset MCP

Instructions

Implementation Reference

• src/tools/get-dataset-info.ts:37-51 (handler)

  The handler function for the 'get_dataset_info' tool that executes the logic.

  async ({ dataset, config }) => {
    const data = await fetchDatasetViewer<InfoResponse>("/info", {
      dataset,
      config,
    });
  
    return {
      content: [
        {
          type: "text" as const,
          text: JSON.stringify(data, null, 2),
        },
      ],
    };
  }

• src/tools/get-dataset-info.ts:5-27 (schema)

  The interface defining the structure of the dataset info response.

  interface InfoResponse {
    dataset_info: {
      description?: string;
      citation?: string;
      homepage?: string;
      license?: string;
      features?: Record<string, unknown>;
      builder_name?: string;
      dataset_name?: string;
      config_name?: string;
      version?: Record<string, unknown>;
      splits?: Record<
        string,
        {
          num_examples: number;
          num_bytes: number;
        }
      >;
      download_size?: number;
      dataset_size?: number;
    };
    partial: boolean;
  }

• src/tools/get-dataset-info.ts:29-53 (registration)

  The function that registers the 'get_dataset_info' tool with the McpServer instance.

  export function registerGetDatasetInfo(server: McpServer) {
    server.tool(
      "get_dataset_info",
      "Get the schema, metadata, and row counts for a dataset configuration",
      {
        dataset: z.string().describe("Dataset ID (e.g., 'stanfordnlp/imdb')"),
        config: z.string().describe("Configuration name (from list_splits)"),
      },
      async ({ dataset, config }) => {
        const data = await fetchDatasetViewer<InfoResponse>("/info", {
          dataset,
          config,
        });
  
        return {
          content: [
            {
              type: "text" as const,
              text: JSON.stringify(data, null, 2),
            },
          ],
        };
      }
    );
  }