HCompany just shipped a Chrome extension that turns the browser into a workspace you can teach. HoloTab, released alongside their new Holo3 computer-use model, watches you perform a task once — clicking, typing, narrating — and captures it as a reusable routine. After that, the AI agent replays it on demand or on a schedule, navigating interfaces and making decisions on its own.
The pitch is straightforward: no API keys, no code, no setup. You describe what you want or just show it. HCompany CEO Charles Kantz framed it bluntly: “Building something powerful is one thing; making it accessible and easy to use is another.” They’re betting the average knowledge worker doesn’t want to configure a RPA bot but does want to stop cross-referencing competitor pricing across twenty tabs to update a spreadsheet. The routines feature is the sharp end of that bet. Hit record, do the monotonous thing once, and HoloTab builds a goal-aware script that handles the repetition forever.
What’s different here is the model underneath. Holo3 is HCompany’s most advanced computer-use AI, built to understand interfaces visually and plan actions the way a person would — parsing buttons, fields, and layouts rather than brittle CSS selectors. That vision-layer approach means it should degrade more gracefully when a site redesigns a checkout flow or moves a button. The extension is free, available now on the Chrome Web Store, and explicitly aimed at people who don’t have an engineering team to wire together agent frameworks.
Skepticism is healthy in computer-use AI right now. Demos are slick; real-world reliability on messy, ad-laden web pages is another thing entirely. But the zero-friction distribution — it’s a browser extension, not a developer beta — is a genuine move toward making autonomous agents a utility rather than a research project. The question isn’t whether people will try it. It’s whether HoloTab will still work when a site throws up a cookie banner, a CAPTCHA, and an unskippable promo modal all at once.
