Examples

This page provides real-world examples of using Kaggle-MCP with Claude, demonstrating how to interact with Kaggle through natural language conversations.

Authentication

User: I'd like to use Kaggle for my data science work. Can you help me set up my credentials?

Claude: I'd be happy to help you set up your Kaggle credentials. First, you'll need to get your API key from Kaggle:

1. Go to your Kaggle account settings at https://www.kaggle.com/settings/account
2. Scroll down to the API section
3. Click "Create New API Token" to download a kaggle.json file

Once you have that file, you can authenticate using the authenticate() tool by providing your username and key from the downloaded file:

authenticate("your_username", "your_api_key")

Alternatively, you can move the kaggle.json file to ~/.kaggle/kaggle.json and set the correct permissions with:
chmod 600 ~/.kaggle/kaggle.json

Let me know when you have your API credentials, and I'll help you authenticate.

Exploring Competitions

User: Show me the top Kaggle competitions happening right now.

Claude: Let me list the currently active Kaggle competitions for you:

competitions_list(sort_by="prize", category="featured")

Here are the top competitions sorted by prize amount. Is there any specific competition you're interested in learning more about?

User: Tell me more about the Titanic competition.

Claude: I'll get the details of the Titanic competition for you:

competition_details("titanic")

This is Kaggle's legendary Titanic Machine Learning competition, where the goal is to predict which passengers survived the Titanic shipwreck. It's a great beginner-friendly competition.

Would you like me to:
1. Show you the competition files
2. Download the data files
3. Show you the current leaderboard
4. Something else?

Working with Datasets

User: Can you help me find datasets related to climate change?

Claude: I'd be happy to help you find climate change datasets on Kaggle:

datasets_list(search="climate change", sort_by="votes", size="all")

These are the top-voted datasets related to climate change. Would you like to:
1. Get more details about any of these datasets
2. Narrow the search (e.g., by file type, license, etc.)
3. Download one of these datasets
4. Something else?

User: The global temperature dataset looks interesting. Can you download it for me?

Claude: I'll download the global temperature dataset for you:

dataset_download_files("berkeleyearth/climate-change-earth-surface-temperature-data", path="/path/to/download")

The dataset has been downloaded to the specified path. Would you like me to:
1. List the files in the dataset
2. Help you analyze the dataset
3. Look for notebooks that use this dataset
4. Something else?

Exploring and Using Kernels

User: Find me some good notebooks for sentiment analysis.

Claude: Let me search for popular sentiment analysis notebooks on Kaggle:

kernels_list(search="sentiment analysis", sort_by="votes", kernel_type="notebook", language="python")

These are the top-voted Python notebooks for sentiment analysis. Would you like to:
1. Get more details about any specific notebook
2. Download one of these notebooks
3. Refine the search
4. Something else?

User: The first one looks useful. Can you download it?

Claude: I'll download the notebook for you:

kernel_pull("username/sentiment-analysis-notebook", path="/path/to/download")

The notebook has been downloaded. Would you like me to:
1. List the files in the notebook
2. Explain the approach used in the notebook
3. Help you adapt this notebook for your own dataset
4. Something else?

Working with Models

User: Find me pre-trained models for image classification.

Claude: Let me search for popular image classification models on Kaggle:

models_list(search="image classification", sort_by="votes")

These are the top-voted image classification models. Would you like to:
1. Get more details about any specific model
2. Download a model
3. Refine the search
4. Something else?

User: Can you show me the details of the first model?

Claude: I'll get the details of the image classification model you're interested in:

model_get("username/image-classification-model")

Here are the full details of the model, including its description, framework, and usage information. Would you like to:
1. Download this model
2. See other versions of this model
3. Look for notebooks that use this model
4. Something else?

Data Analysis Workflow

User: I want to participate in the Titanic competition. Help me get started.

Claude: I'll help you get started with the Titanic competition. Let's break this down into steps:

1. First, let's get the competition details:
   competition_details("titanic")

2. Now, let's download the competition files:
   competition_download_files("titanic", path="/path/to/titanic")

3. Let's list the files to see what we're working with:
   competition_list_files("titanic")

Great! Now that we have the data, we can start analyzing it. We typically have train.csv and test.csv files.

Would you like me to:
1. Help you explore the data
2. Show you some strategies for feature engineering
3. Help build a simple model
4. Find successful notebooks for inspiration
5. Something else?

Advanced Use Cases

User: I want to create and upload a new dataset of weather data.

Claude: I'll help you create and upload a new weather dataset to Kaggle. Here's the process:

1. First, let's initialize the dataset metadata:
   dataset_initialize_metadata("/path/to/weather_data")

2. Now, let's prepare the metadata with a proper description and license:
   dataset_update_metadata("your-username/weather-dataset", '''
   {
     "title": "Weather Dataset 2023",
     "subtitle": "Comprehensive weather data from multiple sources",
     "description": "This dataset contains daily weather measurements including temperature, precipitation, and wind speed from various global stations.",
     "licenses": [{"name": "CC0-1.0"}],
     "keywords": ["weather", "climate", "temperature", "precipitation"]
   }
   ''')

3. Finally, let's create the dataset:
   dataset_create_new("Weather Dataset 2023", "/path/to/weather_data", license_name="CC0-1.0", description="Comprehensive weather data from multiple sources", is_private=False)

The dataset has been created! Would you like to:
1. Update any metadata
2. Check the status of the upload
3. Create a kernel to analyze this dataset
4. Something else?

These examples show the flexibility and power of using Kaggle-MCP with Claude. You can have natural conversations while leveraging the full capabilities of the Kaggle platform.