browser-use/browser-use

Browser Use is the leading open-source Python framework that enables large language models to autonomously control web browsers. With nearly 80,000 GitHub stars, it has become the de facto standard for AI-powered browser automation, allowing developers to build agents that can navigate websites, fill forms, extract data, and complete multi-step web tasks using natural language instructions. The framework takes a hybrid approach to web understanding, combining DOM parsing with vision-based analysis through Playwright. When an agent receives a task, Browser Use loads the target page, processes the DOM to identify interactive elements like buttons, links, and input fields, and optionally engages a vision model for layout-dependent decisions. The LLM then reasons about what actions to take and executes them in sequence. Browser Use is LLM-agnostic, supporting OpenAI GPT-4, Anthropic Claude, Google Gemini, and local models via Ollama, alongside its own optimized ChatBrowserUse model. The framework includes a CLI for interactive browser control, a template system for quick project scaffolding, and extensible custom actions via Python decorators. For production deployments, Browser Use offers a cloud browser infrastructure with stealth capabilities, CAPTCHA handling, and proxy rotation. Backed by a $17 million seed round led by Felicis with participation from Paul Graham and Y Combinator, Browser Use has achieved an 89.1% success rate on the WebVoyager benchmark across 586 diverse web tasks. Its MIT license and active development — with 8,739 commits and 118 releases — make it a robust foundation for any AI browser automation project.