Dataset Viewer

You can run and interact with the Dataset Viewer MCP Server to search, filter, inspect, and download Hugging Face datasets. This server provides a programmatic interface to access dataset content, statistics, and configurations, making it easier to build tools and dashboards that analyze datasets hosted on the Hugging Face Hub.

How to use

To use this MCP server, you run it locally and connect your MCP client to its stdio endpoint. You can validate that a dataset exists, fetch detailed information, retrieve paginated rows, get statistics, search text inside a dataset, filter results with SQL-like conditions, and download datasets in Parquet format. Use a dedicated dataset identifier (for example stanfordnlp/imdb) and provide an auth token if you work with private datasets. The server exposes all listed tools through a consistent set of commands you call from your MCP client.

How to install

Prerequisites: You need Python 3.12 or higher and the uvx command line tool to manage the MCP runtime.

1. Install the dataset viewer MCP locally by cloning the project and setting up the environment.

git clone https://github.com/privetin/dataset-viewer.git
cd dataset-viewer

# Create virtual environment
uv venv

# Activate virtual environment
# On Unix:
source .venv/bin/activate
# On Windows:
.venv\Scripts\activate

# Install in development mode
uv add -e .

Configuration

Environment variable for private datasets: set HUGGINGFACE_TOKEN with your Hugging Face API token.

Claude Desktop integration: add the listed config to your Claude Desktop config file to enable MCP workflow integration.

{
  "mcpServers": {
    "dataset-viewer": {
      "command": "uv",
      "args": [
        "--directory",
        "parent_to_repo/dataset-viewer",
        "run",
        "dataset-viewer"
      ]
    }
  }
}

Notes

This server does not expose an HTTP endpoint by default in the configuration examples. Use the stdio runtime configuration to run the MCP server locally and connect via your MCP client.

Available tools

validate

Check if a dataset exists and is accessible. Parameters: dataset (e.g., 'stanfordnlp/imdb'), auth_token (optional) for private datasets.

get_info

Get detailed information about a dataset. Parameters: dataset, auth_token (optional) for private datasets.

get_rows

Get paginated contents of a dataset. Parameters: dataset, config, split, page (optional), auth_token (optional) for private datasets.

get_first_rows

Get first rows from a dataset split. Parameters: dataset, config, split, auth_token (optional) for private datasets.

get_statistics

Get statistics about a dataset split. Parameters: dataset, config, split, auth_token (optional) for private datasets.

search_dataset

Search for text within a dataset. Parameters: dataset, config, split, query, auth_token (optional) for private datasets.

filter

Filter rows using SQL-like conditions. Parameters: dataset, config, split, where, orderby (optional), page (optional), auth_token (optional) for private datasets.

get_parquet

Download entire dataset in Parquet format. Parameters: dataset, auth_token (optional) for private datasets.