The Agenticle Leaderboard launched The New 100, a weekly ranking of open-source AI agents that measures what is being built right now, not what was popular months ago. The board refreshes every Monday with a 30-day rolling window, and as of September 7, 2026, it draws from 1,647 agentic repositories created in the last 30 days.
Momentum Score, Not Just Star Count
The ranking uses a Momentum Score rather than raw star count. The top newcomer is OpenBot by CopilotKit, a coding and software engineering agent that accumulated 4,407 stars in 22 days for a Momentum Score of 92.4. Star count alone does not capture velocity: a project with 7,100 stars over 24 days (ranked third) scores lower than one with 4,407 stars over 22 days because the scoring weights recency and growth rate.
The second-highest scorer is Nanako0129, another coding agent, which hit 2,400 stars in just 10 days for a score of 89.0. The youngest project in the top 10 is 2akouwu at 8 days old with 1,000 stars and a score of 80.9. At the other end of the age spectrum, Leonxlnx reached 3,100 stars in 29 days for a score of 78.4.
Coding Agents Dominate the New Class
Coding and software engineering agents make up the largest category on the board. Of the top 15 newcomers, nine fall into this category. Agent frameworks, general AI agents, chatbots, research and reasoning tools, browser and computer use agents, multi-modal agents, workflow orchestration, and autonomous agents round out the rest.
The diversity is notable. Human-Agent-Society, an autonomous agent project, reached 681 stars in 8 days with a score of 75.2. Player-YN, a browser and computer use agent, hit 1,600 stars in 10 days. totec448-spec, a workflow orchestration tool, scored 79.9 with 1,300 stars in 16 days. The board captures a snapshot of where builder attention is flowing, not just which projects are biggest.
Automatic Discovery, No Submission Required
The pipeline sweeps GitHub every Monday for agentic projects created in the last 30 days. No submission is required for inclusion. Projects that are missing can be submitted directly, but the default path is automatic discovery. This means the board reflects what the GitHub ecosystem is producing, not what developers remember to promote.
The 30-day window means the board is entirely new every month. Projects age off as they pass the 30-day mark, and new launches take their place. For teams tracking the agent ecosystem, this creates a reliable weekly signal of what is gaining traction immediately after launch, without the noise of older projects that have had time to accumulate stars through existing community size rather than genuine new interest.
What the Numbers Suggest
The concentration of coding agents at the top reflects the current demand for tools that assist with software development tasks. But the presence of research and reasoning agents, workflow orchestration tools, and multi-modal agents in the top 30 suggests the ecosystem is broadening beyond code generation. Projects like halofyai (research and reasoning, 424 stars in 16 days) and HanyuanWang (multi-modal, 907 stars in 16 days) indicate builder interest in agents that handle non-coding tasks.
The age distribution is also telling. Several projects scored above 60 with fewer than 500 stars, which means they gained momentum quickly relative to their size. useagenthq (283 stars in 9 days, score 63.9) and Apeireth (100 stars in 3 days, score 61.0) are examples. The Momentum Score rewards velocity, so a small project that catches fire fast can outrank a larger one that grew slowly.
For developers evaluating which new agent projects to watch, the board provides a filtered view of a very large space. The 1,647 repositories created in 30 days is a lot of signal to sift through manually. The ranking does the sifting, and it resets monthly, so the signal stays fresh.