There’s a small open-source project that went from zero to nearly 100,000 GitHub stars in eighteen months. Not Ollama. Not LangChain. Something most people haven’t heard of yet: Browser Use.
What it does sounds almost too simple. It lets AI agents interact with websites the way a human would — clicking buttons, filling forms, navigating pages, extracting data. Think Selenium or Playwright, but instead of writing brittle CSS selectors, you describe what you want in natural language and the agent figures out the DOM.
The reason this matters is embarrassingly obvious once you see it: most real-world business processes still live behind web UIs. Internal CRMs. Admin dashboards. Legacy tools that never got an API. Before Browser Use, connecting an AI agent to these systems meant either expensive integration projects or fragile screen-scraping scripts that broke every time someone moved a button.
Browser Use is the bridge. It gives agents access to software that was never designed for programmatic access.
The growth curve here is what caught my attention. Faster than Ollama. Faster than anything I’ve tracked in the AI tooling space. That kind of velocity doesn’t happen because something is a nice-to-have. It happens because it solves a pain point so fundamental that people adopt it before they even realize they’re depending on it.
I think what we’re watching is the early innings of something larger: a shift from API-first integration to agent-first computing. For decades, the way software talked to other software was through carefully designed APIs — structured, documented, versioned. But the vast majority of software in the world doesn’t have a clean API. It has a web interface built for humans. Browser Use doesn’t ask those systems to change. It meets them where they are.
There’s a pattern here worth noticing. Ollama made local LLMs trivial. Open WebUI gave people a ChatGPT-like experience they could own. Continue brought Copilot-style completion to any model. Each of these tools removes a friction point in the chain between “I have an idea” and “the AI is doing the work.” Browser Use removes the last big one: getting the agent out into the world where the actual work happens.
The caveats are real. Dynamic SPAs with heavy JavaScript still trip it up. I wouldn’t trust it with mission-critical financial transactions yet. But six months ago, I would’ve called it a demo toy. Now it’s saving teams dozens of hours per week on internal automation workflows.
The trajectory is the thing to watch. When a tool goes from novelty to infrastructure this fast, the question isn’t whether it will matter. The question is what gets built on top of it.
Sources: Kunal Ganglani’s open-source AI projects roundup, June 2026