The Agentic Leaderboard Tracks High-Momentum AI Newcomers

A new tracking platform called The Agentic Leaderboard has launched to identify and rank open-source AI agents that have debuted within the last 30 days. The tool aims to surface high-traction newcomers in a rapidly expanding field where traditional ranking metrics often overlook emerging projects.

The platform identifies the "New 100" agents by automatically sweeping GitHub every Monday. In a single weekly sweep, the pipeline scanned 1,757 agentic repositories to evaluate their growth and quality. Currently, CopilotKit holds the top spot as the leading newcomer on the leaderboard.

The Momentum Score

To avoid the bias of legacy metrics, the platform utilizes a "Momentum Score" rather than total popularity. According to the author on Hacker News, this system was created because existing "best AI agents" lists frequently rank the same five projects from 2023, leaving developers without a clear view of recent innovation.

The Momentum Score is calculated using a weighted formula: star velocity accounts for 35%, while traction—including forks and watchers—makes up 20%. Development activity contributes another 20%, with the remaining percentage split between README depth and craft (15%) and general relevance (10%).

Industry Implications

This shift toward momentum-based tracking allows developers and researchers to discover emerging frameworks before they reach mainstream adoption. By prioritizing development velocity over long-term community health metrics, the tool helps identify projects that are gaining immediate traction but lack the legacy data, such as extensive CI/CD coverage, that established projects possess.

This systematic approach provides a window into the next generation of industry-standard tools, surfacing projects that show the signals of serious development and early adoption in a crowded open-source ecosystem.

Limitations and Outlook

While the leaderboard identifies high-growth projects, it does not measure actual performance. The Agentic Leaderboard author clarified on Hacker News that the scoring is based on static analysis of repository metadata rather than sandboxed execution. Consequently, the tool indicates that a project has real traction and professional signals, but it cannot verify if the agents complete their intended tasks effectively.

Observers will be watching to see if the projects currently topping the "New 100" maintain their velocity as they transition from newcomers to established frameworks.