Skyvern's Vision-First Browser Agent Crosses 21.5K Stars

Skyvern climbed to roughly 21,500 GitHub stars this week, making it one of the fastest-growing browser automation agents on the platform, per its GitHub repo.

The project already claims more than 10 million executed workflows and sub-300ms response times. It skips the DOM selectors most browser agents rely on. Instead, it reads a page the way a person does, using natural language instructions plus computer vision to click, scroll, and fill in forms.

That matters because DOM selectors are the thing that breaks every time a site ships a redesign. Anyone who's maintained a Playwright suite knows the drill: a class name changes, the suite goes red, someone spends an afternoon rewriting locators. A vision-based agent is betting it can skip that entirely by working off pixels instead of markup.

Skyvern isn't the only one making that bet. Stagehand takes a hybrid approach, still running on Playwright under the hood while adding natural-language steps on top. Skyvern goes further and cuts selectors out of the picture completely.

One caveat: 21,500 stars is a single-source growth signal from GitHub. It's not a benchmark against Stagehand or Playwright on task success rate, cost, or latency at scale. Star counts measure attention, not whether the thing works better in production.

Pixels don't need refactoring the way markup does, so the long-term bet favors Skyvern's approach. What the star count can't tell you is whether it actually beats Stagehand on task success or cost once real teams put it into production. That's the number worth watching next, not the GitHub trend chart.