Browser Automation.
An agent with eyes, hands, and a steering wheel inside Chromium.
Most agents call APIs and call it a day. J@rv1s drives a real browser. It reads what's on screen the way you would, decides what to click, fills the form, waits for the spinner, and verifies the result before claiming success. When the page changes, the planner replans instead of crashing.
Every browser action passes through the planner, the allow-list gate, and lands as an auditable receipt.
Reads the DOM and the rendered pixels. Selectors are a fallback, not the source of truth.
When a button moves or a modal appears, J@rv1s notices and adjusts the plan instead of dying on a stale selector.
Watch it work in a visible window, or let it run in the background. Same agent either way.
Every action is logged with screenshots. Undo or re-run any step from the receipt.
Plain English: "renew my library books and email me the new due dates."
J@rv1s drafts the steps and shows you before touching the browser.
Chromium opens. Login, navigation, form fill, confirmation — all logged.
What it did, what it skipped, and the new due dates pinned to your inbox.
No. J@rv1s pauses and hands the browser back to you when it hits a human-verification challenge.
Sessions live in your local profile. J@rv1s never sees or transmits your passwords.