Browser Use

Real browser automation with two modes: agent-browser for direct control via CLI commands, and browser-use for autonomous tasks where an LLM figures out the steps. You get Playwright under the hood with element refs that make clicking and scraping straightforward. Good for multi-step flows like form filling, authenticated sessions, or structured scraping from dynamic pages. The autonomous mode is interesting for open-ended tasks like pricing comparisons across sites, though you'll burn tokens. Google and Bing block headless browsers with CAPTCHAs, so stick to smaller sites or use DuckDuckGo. Refs reset on navigation, so snapshot after every page load.