HF Dataset MCP

Instructions

Implementation Reference

• src/tools/get-statistics.ts:24-39 (handler)

  The handler function that executes the 'get_statistics' tool logic by calling fetchDatasetViewer.

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

• src/tools/get-statistics.ts:19-23 (schema)

  Input schema definition for the 'get_statistics' tool using Zod.

  {
    dataset: z.string().describe("Dataset ID (e.g., 'stanfordnlp/imdb')"),
    config: z.string().describe("Configuration name"),
    split: z.string().describe("Split name (train, test, validation)"),
  },

• src/tools/get-statistics.ts:15-40 (registration)

  Registration function that registers the 'get_statistics' tool with the McpServer.

  export function registerGetStatistics(server: McpServer) {
    server.tool(
      "get_statistics",
      "Get descriptive statistics for each column in a dataset split",
      {
        dataset: z.string().describe("Dataset ID (e.g., 'stanfordnlp/imdb')"),
        config: z.string().describe("Configuration name"),
        split: z.string().describe("Split name (train, test, validation)"),
      },
      async ({ dataset, config, split }) => {
        const data = await fetchDatasetViewer<StatisticsResponse>("/statistics", {
          dataset,
          config,
          split,
        });
  
        return {
          content: [
            {
              type: "text" as const,
              text: JSON.stringify(data, null, 2),
            },
          ],
        };
      }
    );