Kaggle datasets

Download and search Kaggle datasets from the terminal using the Kaggle API.

Arka includes a kaggle skill for downloading and searching Kaggle datasets. It uses the official Kaggle API via the kaggle CLI (preferred) or the Python kaggle package.

Quick examples

kaggle status
kaggle search titanic
kaggle download heptapod/titanic
kaggle download heptapod/titanic --unzip -o ~/Downloads

Natural language also works:

arka "download kaggle dataset heptapod/titanic"
arka "kaggle search housing prices"
arka "kaggle status"

Or via the CLI subcommand:

arka kaggle download heptapod/titanic --unzip

Commands

Command	Description
kaggle status	Check whether Kaggle credentials are configured
kaggle search <keywords>	Search datasets by keyword
kaggle download <owner/dataset>	Download a dataset to your Downloads folder
kaggle download <owner/dataset> -o DIR	Download to a custom directory
kaggle download <owner/dataset> --unzip	Download and extract archives
kaggle open <owner/dataset>	Open dataset page in browser (no API key)
kaggle download <owner/dataset> --open	Open in browser when credentials are missing

Credentials

Kaggle requires an API token. Create one at kaggle.com/settings (Account → API → Create New Token).

Option A — environment variables (recommended for Arka):

KAGGLE_USERNAME=your_username
KAGGLE_KEY=your_api_key

Add these to ~/.config/arka/.env.

Option B — kaggle.json (standard Kaggle layout):

mkdir -p ~/.kaggle
# unzip the token download into ~/.kaggle/kaggle.json
chmod 600 ~/.kaggle/kaggle.json

Run kaggle status to verify credentials before downloading.

Without API keys

The official Kaggle API requires credentials. If you do not have an API token yet, you can still open a dataset page in your browser and download files manually from the Kaggle website.

kaggle open heptapod/titanic
kaggle open https://www.kaggle.com/datasets/heptapod/titanic

Natural language:

arka "open kaggle dataset heptapod/titanic"
arka "kaggle open heptapod/titanic"

You can also pass --open (or --browser) on download to open the page instead of failing when credentials are missing:

kaggle download heptapod/titanic --open

If you run kaggle download or kaggle search without credentials, Arka prints a short message suggesting kaggle open owner/dataset for manual download.

API approach

The skill prefers the kaggle CLI if it is on PATH:

kaggle datasets download -d owner/dataset -p /output/dir
kaggle datasets list -s "query"

If the CLI is not installed, it falls back to the Python kaggle package:

pip install kaggle

Related skills

Skill	Example
generate_data	generate 100 users as csv
data_ask	ask data in reports folder
competitions	search kaggle competitions

See also: Generate data