huggingface_hub

huggingface_hub is the official command-line interface and Python client for the Hugging Face Hub, a platform for sharing open-source machine learning models, datasets and demos. The project ships both interfaces in one package: the `hf` CLI for terminal use and the `huggingface_hub` Python library. Installation is available through a standalone installer script for macOS, Linux and Windows, or via pip/uv for the Python package, which also installs the CLI. Optional dependency groups (for example the MCP module) keep the default install minimal. Core capabilities described in the README: - Download individual files or entire repositories, with files stored in a local cache folder. - Upload single files or whole folders to model, dataset or Space repositories. - Create and manage repositories on the Hub. - Run inference against deployed models. - Run Jobs on Hugging Face infrastructure. - Search for models, datasets and Spaces. - Share model cards to document models. - Engage with the community through pull requests and comments. The CLI is explicitly designed for both humans and coding agents: command output adapts when invoked by an agent, and a generated skill reference can be installed for tools such as Claude Code, Codex, Cursor and OpenCode. Authentication uses Hub tokens, with a non-interactive token option for automated environments. The README also notes integrations with other open-source ML libraries, offering free model and dataset hosting, file versioning for large files via the Xet storage backend, in-browser widgets, CDN-accelerated downloads, and usage statistics. Contribution guidance is provided for feature requests and bug reports.