element: select anything for your agent

Click any element in your browser and hand it to your AI coding agent.

Stop describing the button. Click it. element lets you click any element in your real browser and hand it straight to the AI coding agent on your computer. It works with any agent that can run a shell command. HOW IT WORKS • Hold ⌥ (Option) and hover: anything on the page lights up. Click what you mean. • It lands in your agent: as an [element #7] chip in your prompt, or attached to your next message, depending on the agent. • Say "make this bigger" and your agent already knows which element, where it lives in your code, and what it looks like. WHAT YOUR AGENT GETS • CSS selector (prefers ids and data-testid), role and accessible name, text • Trimmed HTML, ancestors, size and the styles that matter • A screenshot cropped to the element • The React, Vue or Svelte component and source file, in development builds SAVE & SHARE • Hold ⌃⌥ and click: element saves a live snapshot of that element and copies a link. Paste it to a teammate, an issue or another agent. • On Pro and Max, make links private to the emails or domains you choose. ALSO • Several agents at once: each tab belongs to one session, switched in the ⌥⇧S module. • Your agent can open pages for you. • Light and dark, and a highlight that reads on any site. SETUP 1. Install the extension. 2. Paste the setup prompt from elmts.dev/install into your agent, or in a terminal: npm i -g @elmts/cli && el install PRIVACY Picking runs on your computer: what you click goes from the extension to your agent and nowhere else. Only signing in and save & share go online. No analytics scripts and no tracking cookies. If you arrive from one of our ad or creator links, we count the visit, install, first pick, sign-up and upgrade against that link's code, as daily totals (opt out with el install --no-ref). Full policy: elmts.dev/privacy.html