kaggle-kernels

List all kernels

$ kaggle [[k|kernels]] list

$ kaggle [[k|kernels]] files [kernel_name]

$ kaggle [[k|kernels]] init [[-p|--path]] [path/to/directory]

$ kaggle [[k|kernels]] push [[-p|--path]] [path/to/directory]

$ kaggle [[k|kernels]] pull [kernel_name] [[-p|--path]] [path/to/directory]

$ kaggle [[k|kernels]] output [kernel_name]

$ kaggle [[k|kernels]] status [kernel_name]

kaggle kernels subcommand [options] [arguments]

list
    List kernels with optional filters for competitions (-c), user, page, sort-by (hotness, dateRun, dateCreated), kernel-type (script/notebook)

-m, --mine
    Filter to only your own kernels

-p, --public
    Filter to only public kernels

--page <page>
    Page number for paginated results

--user <username>
    Filter by specific user

--sort-by <sort>
    Sort list by hotness/dateRun/dateCreated

--kernel-type <type>
    Filter by script or notebook

pull <kernel>
    Download kernel source and data to local dir

push [-p <path>] [-q]
    Upload and execute local notebook/kernel

-c <competition>
    Push to specific competition

--wait
    Wait for kernel to complete before returning

--run-all
    Run all cells during push

-m <message>
    Commit message for push

status <kernel>
    Check kernel execution status

--path <path>
    Local path for status check

output <kernel>
    Download kernel outputs

-h, --help
    Show help for subcommand

-v, --version
    Show Kaggle CLI version (global)

The kaggle kernels command is part of the official Kaggle CLI tool, enabling users to interact with Kaggle Kernels—cloud-based Jupyter/IPython notebooks hosted on Kaggle's platform for data science competitions, datasets, and models.

It supports key operations like listing kernels with filters (e.g., user's own, public, by competition), pulling kernel source code and data to local directories, pushing local notebooks to Kaggle for execution and sharing, monitoring kernel execution status, and retrieving outputs.

Essential for developers and data scientists, it streamlines workflows by integrating with local IDEs like VS Code or JupyterLab, allowing version control via Git, automated submissions, and collaboration without relying solely on the web UI. Usage requires a Kaggle account, API credentials (kaggle.json), and the Python-based CLI installed via pip.

For example, list your kernels with kaggle kernels list -m, update a notebook with kaggle kernels push, or track progress via kaggle kernels status mykernel. Handles large datasets efficiently but depends on internet and Kaggle's compute limits.

Requires kaggle.json API key in ~/.kaggle; internet required; subject to Kaggle compute quotas and privacy rules; large kernels may timeout.

pip install kaggle; verify with kaggle --version.

Download API key from Kaggle profile; kaggle config set -g api-key ~/.kaggle/kaggle.json; set permissions: chmod 600 ~/.kaggle/kaggle.json.

Introduced in Kaggle API v1.0+ around 2018 as Kaggle expanded notebook features; evolved with Google acquisition (2017+), adding GPU/TPU support and push/status enhancements by 2022.

kaggle(1), pip(1), jupyter(1), git(1)