AI agents are becoming a major focus for startups and investors, even though the market still lacks a single agreed-upon definition of what an agent is. Browser Use is one of the companies trying to make that category more practical by helping software navigate websites with fewer breakdowns.
The company told TechCrunch it has raised a $17 million seed funding round. The round was led by Felicis’ Astasia Myers, with participation from Paul Graham, A Capital, and Nexus Venture Partners.
What Browser Use Is Building
Browser Use focuses on a specific problem: websites are designed for humans, not for autonomous software. Buttons, menus, and changing page layouts can be difficult for AI agents to interpret reliably, especially when the agent is trying to complete the same task more than once.
The company’s tool breaks down website buttons and elements into a more digestible, text-like format. That gives agents a clearer representation of what is available on a page, so they can understand options and make decisions autonomously.
This is different from relying only on screenshots. According to Magnus Müller, many agents use vision-based systems to move through websites, and that process can fail. Browser Use is built around converting websites into something agents can understand more directly.
For developers, the promise is not simply that an agent can see a page. It is that the agent can interpret the structure of the page in a way that supports repeated work. Müller said this can allow the same tasks to run again and again at a cheaper cost.
How The Startup Gained Attention
Browser Use is part of Y Combinator’s 2025 winter batch, and it has drawn attention from both developers and investors. The company’s visibility rose further after Chinese startup Butterfly Effect used Browser Use in its viral Manus tool.
The startup was founded by Magnus Müller and Gregor Žunič last year through ETH Zurich’s Student Project House accelerator. Müller had been working on web-scraping tools for years, and he met Gregor Žunič in 2024 while they were both pursuing master’s degrees in data science.
The idea came from combining web scraping with data science to prompt a browser to perform a task. Müller and Žunič built a demo in five weeks, and the project quickly gained traction. They later open sourced it.
That open-source-first approach appears to be part of the company’s appeal. Felicis’ Astasia Myers said the team and the open-source strategy helped convince the firm that Browser Use was the right opportunity in the AI agents space.
Why Website Navigation Matters For AI Agents
The web is not static. Websites can change how pages work, how buttons behave, or how workflows are arranged. Müller pointed to LinkedIn as an example of a site that changes how it works all the time, which can cause agents to fail.
That creates a clear need for infrastructure between AI models and the live web. Browser Use is positioning itself as a fundamental layer for companies that want web AI agents to interact with websites more gracefully.
The source article describes growing demand from companies that want to make their own websites easier for agents to navigate. Müller said companies have approached Browser Use asking what they can do to support agent navigation on their sites.
Within the current Y Combinator winter batch, Müller said more than 20 companies used Browser Use for their own requirements. That detail matters because it shows the tool is not only a concept for future agent products; it is already being used by other startups building around similar needs.
What Investors See In The Market
Felicis has been watching the AI agents space for several years, according to Myers. Her view is that web AI agents are becoming a frontier for end-to-end automation of human tasks.
The investment thesis, based on the source article, is that agents need a bridge between text-focused models and the changing digital landscape of websites. Browser Use aims to supply part of that bridge by making web pages more readable to agentic systems.
The company’s $17 million seed round gives it more room to pursue that role. It also shows that investors are willing to back developer tools that sit beneath consumer-facing agent products, not only the agents themselves.
For Browser Use, the central bet is straightforward: if more companies build AI agents that must operate across websites, those agents will need reliable ways to understand pages. A tool that turns the messy structure of the web into something agents can process could become important infrastructure for that shift.