Agent Usage on the Hugging Face Hub (huggingface.co)

Hugging Face has launched a new dataset that provides insights into the usage of AI agents on their Hub, signifying a step forward in understanding how various tools are utilized within the AI/ML community. The dataset tracks requests made through the huggingface_hub library, capturing the share of traffic associated with named AI harnesses like Claude Code, Codex, and Cursor. It highlights shifts in usage patterns month-over-month and day-over-day, allowing developers to see the impact of newly launched tools and their contributions to overall Hub traffic. This initiative not only facilitates a clearer view of which agents are dominating the space but also encourages developers to register their harnesses to ensure visibility in these analytics. By exposing how requests correlate with distinct users, the dataset assists in identifying heavily automated tools versus those with a broader user base. The ability to analyze Agent performance with precision, coupled with automatic traffic detection based on User-Agent tokens, paves the way for more effective optimization and development in AI applications, underscoring Hugging Face's role as a vital platform in the evolution of AI technologies.