Show HN: Webact – token-efficient browser control for AI agents (GitHub)
github.comI built webact while experimenting with browser control for LLM agents and kept running into the same problem: token usage.
Most tools wrap Playwright and return huge accessibility trees or DOM dumps, which gets expensive quickly. webact takes a simpler approach — it talks directly to Chrome via the DevTools Protocol and returns a small "page brief" instead.
It’s just a single JS file (~196KB), zero dependencies, and it uses your existing Chrome session.
Happy to hear thoughts. How does it compare to https://github.com/vercel-labs/agent-browser ? Which also provides token efficient way to interact with the browser.