There may not be an agreed definition of “agent” for AI yet, but many startups want to create “agent” tools to automate various tasks online. The use of browsers, one such company, has attracted a lot of interest from developers and investors thanks to solutions that make websites more “readable” for AI agents.
The browser usage told TechCrunch that it raised a significant $17 million seed funding led by Felicis’ Astasia Myers through participation from the capital, Paul Graham and Nexus Venture Partners. The company’s funds have not been reported previously.
The use of the browser, part of Y Combinator’s winter 2025 batch, has become infamous in recent months. The Chinese startup butterfly effect has raised awareness of new heights with the use of the Brower in the Virus Manustool.
Magnus Müller and Gregor Zunic founded the use of the browser through ETH Zurich’s Student Project House Accelerator last year. Muller has been working on web scraping tools for years, meeting Zunic in 2024, and the pair received a master’s degree in data science. Together, they came up with the idea of combining web scraping and data science to encourage browsers to perform tasks.
Müller and Zunic have built a browser that uses the demo in five weeks. They then opened sourced it.
Using a browser essentially breaks down the buttons and elements of a website into a more digestible “text-like” format aimed at agents. This helps agents understand the various options and autonomously make decisions.
“Many agents rely on vision-based systems to try to navigate their website via screenshots. [the] Process, things break,” Muller said. [websites] Something that agents can understand. This approach means that you can perform the same task over and over at a cheaper cost. ”
With an increasing number of AI companies wanting to enable agents to interact with their websites more elegantly, Müller believes that using browsers could be a “basic layer” that caters to this need. He added that over 20 companies in the current Y Combinator Winter Batch use their browsers to use for their own requirements.
“There are companies that come to us and say, ‘What can we do to help agents navigate our website?'” Mueller said. “For example, there are sites like LinkedIn. This changes the way websites work, so agents often fail with such sites.”
According to Myers, Felicis has been actively considering the AI agent space for the past few years, and the use of the browser felt like a good opportunity to grow the company’s portfolio there. She said the company’s founding team and its open source-first approach sealed the contract.
“I think Web AI agents are the next frontier that will really help you with end-to-end automation of human tasks,” Myers told TechCrunch. “[W]The EB AI agent is this dynamic bridge between static pretraining models that focus mostly on text in a constantly changing digital landscape. ”
Source link
