Kernels API

Kaggle-MCP provides tools to search, download, and manage Kaggle kernels/notebooks directly through Claude.

Available Tools

kernels_list

List available Kaggle kernels with extensive filtering options.

kernels_list(search="", user="", language="all", kernel_type="all", output_type="all", sort_by="hotness", page=1, page_size=20)

Parameters:

• search: Term(s) to search for
• user: Display kernels by a specific user
• language: Display kernels in a specific language (all, python, r, sqlite, julia)
• kernel_type: Display kernels of a specific type (all, script, notebook)
• output_type: Display kernels with a specific output type (all, visualization, data)
• sort_by: Sort kernels by (hotness, votes, updated, created)
• page: Page number for results paging
• page_size: Number of items per page

Returns: JSON string with kernel details

kernel_list_files

List files in a specific kernel.

kernel_list_files(kernel)

Parameters:

• kernel: Kernel identifier in format <owner>/<kernel-name>

Returns: JSON string with file details

kernel_output

Download the output of a Kaggle kernel.

kernel_output(kernel, path="")

Parameters:

• kernel: Kernel identifier in format <owner>/<kernel-name>
• path: Folder where output will be downloaded (defaults to a temp directory)

Returns: Success message or error details

kernel_pull

Pull/download code from a kernel.

kernel_pull(kernel, path="", metadata=False)

Parameters:

• kernel: Kernel identifier in format <owner>/<kernel-name>
• path: Folder where kernel will be downloaded (defaults to a temp directory)
• metadata: Whether to generate kernel metadata file

Returns: Success message or error details

kernel_status

Get the status of a kernel.

kernel_status(kernel)

Parameters:

• kernel: Kernel identifier in format <owner>/<kernel-name>

Returns: JSON string with kernel status details

kernel_initialize_metadata

Initialize kernel metadata file for later upload.

kernel_initialize_metadata(path=".", kernel_type="notebook", language="python")

Parameters:

• path: Directory where metadata file will be created
• kernel_type: Type of kernel (notebook or script)
• language: Language of kernel (python, r, or rmarkdown)

Returns: Success message or error details

kernel_push

Push a new version of a kernel or create a new kernel.

kernel_push(folder_path)

Parameters:

• folder_path: Path to the folder containing kernel files and metadata

Returns: Success message or error details

Examples

Here are some examples of how to use the kernel tools with Claude:

# Search for Python notebooks about machine learning
kernels_list(search="machine learning", language="python", kernel_type="notebook")

# Download a specific kernel
kernel_pull("username/kernel-name", "/path/to/download")

# Check the status of a kernel
kernel_status("username/kernel-name")

# Create a new kernel
kernel_initialize_metadata("/path/to/project", kernel_type="notebook", language="python")
kernel_push("/path/to/project")