AI startup Browser Use has raised $17 million in seed funding to improve how AI agents interact with websites. The round was led by Felicis, with participation from Paul Graham, A Capital, and Nexus Venture Partners. Browser Use, a Y Combinator participant, simplifies website structures into a text-based format, making it easier for AI agents to understand and navigate. This contrasts with traditional vision-based approaches that often fail. The company’s founders, Magnus Müller and Gregor Zunic, developed the technology after combining web scraping with data science. The open-source approach has been a key factor in attracting interest. Felicis sees web AI agents as the next frontier in task automation, bridging the gap between static AI models and the dynamic web. Over 20 companies in the current Y Combinator batch are already using Browser Use.
While a universally accepted definition of an AI “agent” remains elusive, numerous startups are racing to develop agent-like tools capable of automating online tasks. Browser Use, one such company, has garnered significant attention from both developers and investors due to its innovative approach to making websites more accessible to AI agents.
Browser Use has disclosed to TechCrunch that it has successfully closed a substantial $17 million seed funding round. The round was spearheaded by Felicis’ Astasia Myers, with contributions from notable investors including Paul Graham, A Capital, and Nexus Venture Partners. This funding round marks the first public announcement of the company’s financial backing.
A participant in Y Combinator’s Winter 2025 cohort, Browser Use has rapidly gained prominence. The viral success of Chinese startup Butterfly Effect’s Manus tool, which utilizes Browser Use’s technology, significantly boosted the company’s visibility.
Founded in 2023 by Magnus Müller and Gregor Zunic through ETH Zurich’s Student Project House accelerator, Browser Use originated from Müller’s extensive experience in web scraping. The co-founders, both pursuing master’s degrees in data science, conceived the idea of merging web scraping with data science to instruct a browser to execute specific actions.
Within a mere five weeks, Müller and Zunic developed a Browser Use demo that quickly gained traction, leading them to release it as an open-source project.
Browser Use’s core technology transforms website elements and buttons into a simplified, text-based format that AI agents can readily process. This facilitates the agents’ ability to comprehend available options and make independent decisions.
“Many AI agents depend on vision-based systems, attempting to navigate websites using screenshots, which often leads to errors,” Müller explained. “We transform [websites] into a format that agents can easily interpret. This method enables us to execute the same tasks repeatedly and cost-effectively.”
With a growing number of AI firms striving to improve their agents’ interaction with websites, Müller envisions Browser Use becoming a “foundational layer” for this purpose. He noted that over 20 companies within the current Y Combinator Winter batch have already integrated Browser Use into their operations.
“Companies are approaching us, asking how they can optimize their websites for easier navigation by AI agents,” Müller stated. “Certain websites, such as LinkedIn, frequently alter their structure, causing agents to frequently encounter difficulties.”
Astasia Myers of Felicis revealed that the firm has been actively exploring the AI agent sector for several years. Browser Use presented a compelling opportunity to expand their portfolio in this area. Myers highlighted the company’s founding team and its open-source-first strategy as key factors in their investment decision.
“We believe that web-based AI agents represent the next significant advancement in automating human tasks,” Myers commented to TechCrunch. “Web AI agents serve as a dynamic link between static, primarily text-based pre-trained models and the constantly evolving digital environment.”