Browser Use, a startup focused on making websites more accessible for AI agents, has raised $17 million in a seed round led by Felicis’ Astasia Myers, with participation from Paul Graham, A Capital, and Nexus Venture Partners.
The company, part of Y Combinator’s 2025 winter batch, has gained traction due to its innovative approach, which transforms website interfaces into structured text that AI can process more reliably than traditional vision-based systems.
Co-founders Magnus Müller and Gregor Zunic, who met during their master’s studies at ETH Zurich, initially developed the project as a web-scraping experiment. Their demo, built in five weeks, quickly gained popularity, with over 20 Y Combinator-backed companies now using the tool.
Chinese startup Butterfly Effect’s viral Manus tool further propelled Browser Use into the spotlight.
Unlike vision-based automation, which relies on screenshots and often fails when websites change, Browser Use provides a structured text representation of web elements, ensuring faster and more reliable execution.
The startup’s open-source-first approach has driven rapid adoption, amassing 50,000 GitHub stars and over 15,000 developers actively using its tools for tasks like login automation, data extraction, and CRM integrations.
So, what’s the deal with Browser Use, and why should you care? Let’s break it down.
Browser Use, a company you might not have heard of until now, has raised $17 million to make websites easier for AI “agents” to understand.
These agents are like digital assistants that can book flights, scrape data, or even test software, all on their own.
The funding round, which hadn’t been leaked before, came from some heavy hitters in the venture capital world.
Felicis took the lead, but Paul Graham, yeah, the Y Combinator co-founder himself, threw in some cash too, alongside A Capital and Nexus Venture Partners.
Browser Use was dreamed up by Magnus Müller and Gregor Zunic, two data science grad students who met at ETH Zurich in 2024.
Müller had been tinkering with web-scraping tools for years—those are programs that pull info from websites, and Zunic brought some fresh ideas to the table. They cooked up Browser Use last year through ETH Zurich’s Student Project House accelerator, and what started as a weekend experiment exploded into something huge.
In just four days, they built a demo, tossed it onto Hacker News, and bam, the internet went wild for it. They decided to open-source it, meaning anyone can use and tweak the code—and now they’ve got over 50,000 GitHub stars and 15,000 developers playing with it.
Why’s this matter? Well, that $17 million isn’t just a pat on the back—it’s fuel to turn their idea into a game-changer. Investors don’t throw around that kind of money unless they think there’s a payoff, and Browser Use might just be onto something that could reshape how AI interacts with the web.
What Browser Use actually does
Okay, let’s get into the nitty-gritty. You know how we humans click around websites, buttons here, forms there?
AI agents try to do the same, but most of them rely on “vision-based” systems, basically taking screenshots and guessing what’s what.
Problem is, that’s slow, pricey, and when a website changes its layout, like LinkedIn does all the time, those agents get lost. Müller put it bluntly in a TechCrunch interview: “A lot of agents rely on vision-based systems and try to navigate websites through screenshots, and in the process, things break.”
Browser Use flips the script. Instead of fumbling with screenshots, their tool breaks a website down into a “text-like” format that AI can read easily. Think of it like translating a messy webpage into a clear instruction manual for robots.
Buttons, dropdowns, input fields, it all gets organized so AI agents can make smart decisions without tripping over themselves. Müller says this makes tasks “faster and more efficient” and cuts costs since you don’t need fancy visual processing. Plus, it’s reliable, you can run the same task over and over without hiccups.
This isn’t just techie jargon, it’s a big deal. Their blog brags about features like multi-tab management (think of AI juggling browser tabs like a pro), element tracking (pinpointing exactly what got clicked), and even self-correcting powers to fix errors on the fly.
It works with all the hot AI models out there, GPT-4, Claude 3, Llama 2, you name it. Already, companies are using it for stuff like logging into sites, pulling data, or automating customer service tasks.
Timing’s everything, right? Browser Use is riding a wave of hype around AI agents. These aren’t your old-school chatbots, they’re smarter, more independent, and thanks to advances in large language models (LLMs), they’re popping up everywhere.
The internet’s a goldmine of data, but it’s a mess, unstructured, chaotic, and built for humans, not machines. Browser Use wants to bridge that gap, making the web as easy for AI to navigate as it is for us.
Their blog makes a bold claim: “Within a few years, automated workflows will outnumber human interactions on the web.” Sounds wild, but there’s some meat to it.
A 2023 report from Gartner predicted that by 2027, 30% of enterprise software would involve autonomous agents, up from almost nothing in 2022. Browser Use’s 50,000 GitHub stars and 15,000+ developers show the demand’s already there.
Heck, over 20 companies in their Y Combinator batch are using it, and a Chinese startup called Butterfly Effect leaned on Browser Use for its viral Manus tool, sending awareness through the roof.
Müller’s seeing the shift firsthand. He told TechCrunch, “There are companies coming to us and saying, ‘What can we do to make it easier for agents to navigate our website?’”
That’s a flip, businesses are now designing for AI, not just humans. It’s a hint that Browser Use could become a backbone for this new web, a “fundamental layer,” as Müller calls it.
The $17 million isn’t chump change for a seed round. For context, PitchBook data shows the average seed round in 2024 was around $3 million. Browser Use’s haul is a standout, signaling investors think it’s got legs.
Open-sourcing the tech was a smart move too, it’s built a community fast, and as Müller and Zunic wrote on their blog, “If you have an idea, just build it and launch it.” That hustle caught eyes.
With this cash, Browser Use is hiring. They’re after “exceptional engineers” to push their tech further, according to their site.
The goal? Keep growing that infrastructure so AI agents can take over more of the web. Müller’s betting on a future where automation rules online, and they’ve got the momentum, 50,000 GitHub stars don’t lie.
But it’s not all smooth sailing. Websites like LinkedIn that keep tweaking their design could still throw curveballs. And there’s the bigger picture: if AI agents start crawling the web like humans, what about privacy? Could they scoop up data in ways we’re not ready for?
There’s no hard data yet on how this might play out, but it’s a question folks are starting to whisper about on X and tech forums. Plus, competition’s heating up, other startups are chasing the same AI agent dream.
As Myers put it, this could be “the next frontier.” Whether it’s a gold rush or a bubble, Browser Use is in the driver’s seat, for now.
