click it in your browser. your agent knows exactly which one.
npm i -g @randyyu/elementfree and open source · macOS · chrome, arc, brave, edge and any chromium browser
turn your docs into an assistant your users actually trust.
anything on the page lights up with corners as you move.
one thing or ten. each pick lands in your prompt as a chip.
your agent already knows which button, where it lives, and what it looks like.
everything you'd dig out of devtools, attached to the pick.
// el_7 selector button[data-testid="checkout"] role button "Place order" component CheckoutButton file src/cart/Checkout.tsx:42 styles background #e5484d border-radius 9px height 40px shot el_7.png 320×40
each tab belongs to one session. picks go to the agent that owns the tab.
stuck on "which one?" it brings chrome forward and asks you to point.
yes. MIT licensed, no account, no server. it runs on your machine between chrome and your agent.
no. element reads a page only when you pick something or your agent asks you to.
any chromium browser on macOS: chrome, arc, brave, edge, opera, vivaldi, dia and the rest. not safari or firefox, and not linux or windows yet.