alibaba/page-agent v1.7.1: In-Page GUI Agent for Web Automation — 16K Stars, No Screenshots Required
GitHub·high signal
Alibaba's page-agent hit 16K stars (+610 today) with v1.7.1 released April 3. Unlike browser-use, Playwright, or Selenium which control from outside, page-agent lives INSIDE the webpage via a single script tag. It uses text-based DOM manipulation — no screenshots, no OCR, no multimodal model needed — reducing LLM token cost by 10-100x compared to vision-based approaches. Ships with MCP Server support (beta) and Chrome extension for multi-page workflows.