Browser Use

Browser Use is the infrastructure layer for AI agents that need to actually click, type, and read their way around the live web. The core is an open-source Python harness that gives any LLM a thin, self-healing toolset for controlling a real browser, paired with a hosted cloud that adds stealth Chromium instances, residential proxies across 195+ countries, CAPTCHA solving, and a managed API for kicking off agent runs at scale.

It's aimed at developers and AI teams building web agents for scraping, QA, monitoring, lead generation, and form-filling workflows where headless HTTP scraping breaks down. The free open-source library is what put the project on the map (one of the most-starred agent repos on GitHub), while the paid cloud handles the messy parts -- fingerprinting, proxy rotation, session persistence, and a 24/7 hosted 'Browser Use Box' that pairs the harness with Claude. Pricing is usage-based on the cloud side with a free tier for testing, and the company markets enterprise plans to Fortune 500 customers.

It's model-agnostic by design: you bring your own LLM (OpenAI, Anthropic, Gemini, local models via Ollama, etc.), and Browser Use exposes the DOM in a structured, token-efficient form so the agent can reason about pages without drowning in raw HTML. There's also a 'Browser Use 1' custom model trained specifically for web tasks.