Bringing agents onto the world wide web — Paul Klein IV, Browserbase

Browserbase argues that better supporting software can make browser agents more useful with existing models. Its talk describes reusable skills and memory so agents do not rediscover a site on every run, alongside infrastructure that renders pages consistently.

These address distinct sources of repeated work and uncertainty. Remembering how a site works can spare an agent another exploration, while consistent rendering avoids unexpected switches between mobile and desktop layouts. The proposed payoff is more repeatable automation of existing web workflows. The argument places substantial engineering work in the system around the model, including how it preserves context and presents the browser.

“When OpenClaw shipped, people started buying Mac minis to run it from home, SSHing in and clearing captchas off a residential IP.”

“His argument is that browser agents are no longer held back by the models. The capability is already there and the engineering around it is missing, which makes the overhang something any team can close rather than wait on a lab for.”

“The most reliable browser agents in production are multimodal and write code alongside clicking, often intercepting network requests and replaying them rather than driving pixels.”

“The payoff is not in San Francisco. It is the logistics company in Singapore, the bank in South Africa, and the lumber factory in Mexico, all running on PHP forms with people clicking buttons every day.”