hf-mcp

The hf-mcp skill acts as a bridge between your coding assistant and the Hugging Face ecosystem, enabling deep integration with the Hub's vast library of models, datasets, and infrastructure. This skill is designed for ML engineers, data scientists, and researchers who need to streamline their development workflows by interacting directly with Hub resources through the Model Context Protocol (MCP).

• Advanced discovery of machine learning models and datasets using granular search filters (task, library, tag, and trending metrics).
• Seamless retrieval of repository details, including README documentation and metadata, to assist in model evaluation and selection.
• Integration with Hugging Face compute infrastructure, allowing agents to execute training jobs or data processing tasks directly from the IDE or terminal.
• Dynamic invocation of Gradio Spaces, effectively turning interactive web-based demos into reusable API-like tools for transcription, image generation, and more.
• Academic research support through integrated paper searching, linking, and metadata extraction for tracking state-of-the-art developments.

Usage notes and practical tips:

• Always utilize sorting parameters like 'trendingScore' or 'downloads' when querying for models to ensure you are working with battle-tested or current popular solutions.
• For jobs requiring access to private repositories, ensure you include the 'HF_TOKEN' secret in your environment configuration.
• Use the 'dynamic_space' operation 'discover' to identify which Gradio Spaces are available as programmatic tools within your current session.
• When searching for dataset specificities, leverage tag-based filtering (e.g., language, task category) to optimize your data retrieval performance.
• This skill requires an active connection to an authorized HF MCP server; ensure your environment is set up according to the official Hugging Face MCP documentation before invocation.