The assistant beside the page, not in another tab
An extension for Chrome, Edge and Firefox opens the assistant in a side panel. It reads the page you are viewing, answers from it and, if you allow it, acts on it: click, fill a field, pick from a list. Every action goes through an approval, unless you decide otherwise.
Read the page, then act on it
The assistant sees what you see, and acts only within the limits you give it.
Read the current page
The address, the title, the readable text and the elements you can interact with — plus your selection, if you have one.
Search and get around
Search a term in the page with its context, list your open tabs, switch between them and scroll to the right place.
Act with your approval
Click an element, fill a field, pick from a dropdown, press a key, open an address — each of these is shown for approval before it runs.
See the page
With a model that can see images, the assistant can capture the visible area to understand a layout that text alone does not convey.
Everything else in the platform
The same panel gives access to your models, assistants, knowledge bases, prompts and connectors — the page you are on simply becomes one more source.
Sensitive fields stay yours
Passwords, card numbers and one-time codes are never read or filled by the assistant. That refusal lives in the extension itself, not in an instruction given to the model.
Installed by the person, governed by your administrators
Unlike the Office add-ins, the extension installs from the official stores — but access stays under your control.
- 01
Chrome, Edge and Firefox
One package published on the Chrome Web Store, which Edge installs from the same listing, and a Firefox build on AMO. There is no Safari version.
- 02
Connected to your instance, and only yours
On first open, the person enters your organization's server address. The extension checks that the address really answers as a HiloIntelligence instance before connecting.
- 03
No app registration to create
Sign-in uses your platform's own session: no Entra registration and no Microsoft 365 connector required for this extension.
- 04
Access stays yours
Your administrators decide which groups may use the extension. They can also force-install it fleet-wide by browser policy on Chrome, Edge and Firefox — otherwise the person installs it themselves.
What to know before turning it on
An extension that reads pages and acts on them deserves to be described without excess enthusiasm.
- The page text is sent to your instance, truncated at around 10,000 characters, with at most a hundred interactive elements.
- The screenshot covers only the visible part of the page, not the whole page.
- An “act without asking” mode exists, but it is off by default and shows an explicit warning: a booby-trapped web page can try to hijack an assistant. Keep it off for banking, email and shopping.
- There is no allowlist or blocklist of sites. On the connect screen the box granting access to all pages is ticked by default; you can untick it, then grant or withdraw access site by site in your browser's extension settings, at any time.
- No drag-and-drop, hover, right-click or file upload; the usable keys are limited.
- One conversation per panel, not one per tab or per site.
- Requires Chrome 116 or newer, or Firefox 128 or newer.
Install the extension
From your browser's official store. On first open, enter your organization's server address.
Edge installs from the same listing.
Firefox 128 or newer.
And beyond the browser
The extension is only one of the ways to keep the platform at hand.
iOS
App Store publication pending.
Add it to your browser
Create your workspace, then install the extension from your browser's store and connect it to your instance.