Show HN: Browser Harness – simplest way to give an AI control of a real browser (github.com)

A new tool named Browser Harness has been unveiled, designed to grant large language models (LLMs) full control over real web browsers via a streamlined, self-healing architecture directly built on the Chrome DevTools Protocol (CDP). This innovative harness allows LLMs to autonomously identify and write necessary functions, effectively allowing them to perform complex browser tasks without pre-defined frameworks. The agent can create code mid-task—such as uploading files or interacting with web applications—by editing core helper scripts, which enhances flexibility and reduces the barriers for AI-driven automation. The significance of Browser Harness lies in its potential to transform how AI interacts with web environments, making it easier for developers to implement AI-driven automation without extensive manual coding. With a free tier offering multiple concurrent browser sessions and the capability to handle tasks like CAPTCHA solving, it opens the door for broader access to AI tools across various industries. By enabling LLMs to generate "skills" for specific tasks automatically, users can benefit from continuous learning and adaptation, significantly improving productivity in tasks such as data entry, web scraping, and online communications while minimizing the need for human oversight.