Cx Browser — the Chrome and Edge extension

What the Cx Browser extension is

Cx Browser is a small extension for Chrome and Edge that lets the assistant read and operate the websites you have open, with your own sign-ins. Once it is connected, you can ask things like "open the supplier portal and find the delivery note for pump P-101", "read the page I have open and create an issue from it" or "fill in this form with the asset data" right in the chat, and the assistant works in your browser step by step.

The extension does nothing on its own: it only acts when the assistant, in a chat you are running, asks it to. It works with the web app's chat (the browser tools are not available in the mobile app or in the Office add-ins).

It is a companion to the web app, like the Word and Excel add-ins: you keep using Cx Assistant as usual, and the browser becomes one more place it can act for you.

How to get the extension

The extension is being published to the Chrome Web Store (it works in Edge too). Until it is listed there, it is available on request: click your avatar at the bottom of the sidebar, choose Support, pick Technical issue and ask for the Cx Browser extension — we send you the package with the install steps for your browser.

There is nothing to install on the server side and nothing to configure in your project: the extension signs in with your own Cx Assistant account.

Signing in and connecting

1. Click the Cx Browser icon in the browser's toolbar (pin it from the extensions menu if it is hidden).
2. Sign in with the email and password you use in Cx Assistant. If you have two-factor authentication on, the popup asks for the code next: from your authenticator app or the email code (a switch at the top picks the method when both are on), with Resend code for the email and Use a backup code for an authenticator.
3. Give this browser a name (for example "Office laptop") so you recognise it in the chat; it is saved as soon as you leave the field.
4. Click Connect.

When the extension is connected, its icon shows a small ON badge and the popup says Connected. In the web app, a "Browser connected" chip appears above the message box of your chats: the assistant can now use your browser in that chat.

The extension stays signed in and reconnects by itself when you reopen the browser. Click Disconnect to pause it without signing out, or Sign out to remove your account from the extension. Revoking that session from the Sessions list on the Security tab of Settings disconnects the extension as well.

Allowed sites — which websites the assistant may use

Out of the box the extension is allowed on no website at all. You decide where it may work, in the popup:

• Allowed sites: type a site, for example app.procore.com or portal.supplier.com, and press Add. Each site becomes a chip; the × on a chip removes it. The site's subdomains are included.
• Allow every site: one switch for all websites. Use it only if you really want the assistant to be able to read and act anywhere you browse.

The moment you add a site (or turn the switch on), the browser asks you to grant the extension access to exactly that site — accept the prompt, or the site stays blocked even though it is on the list. A site still waiting for the browser's permission shows an amber dot on its chip, and a line under the list offers Grant access to ask again. There is nothing to save: the list is stored as you change it.

When the assistant needs a site that is not allowed, it stops and tells you which one; add it, accept the browser's prompt and ask again. The list lives in your browser only: nobody else can change it, and the assistant cannot add sites by itself.

Some pages are never available, whatever the list says: the browser's own settings pages, the web store, and other non-web pages.

What the assistant can do in your browser

With the extension connected, the assistant works in your browser one step at a time, in the chat you are using:

• See which tabs you have open and open a web address (or go back, forward, reload).
• Read the page you are on: its text, or the list of things it can act on (links, buttons, fields, menus).
• Click an element, type into a field (and press Enter when you ask for it), scroll down to what is below the fold.
• Take a screenshot of the page when the content is visual — a chart, a drawing, a map.

It uses your own sessions: if you are signed in to a portal, the assistant is too. It can chain steps across apps — read a supplier portal, then create the issue or the RFI in your project.

Safeguards that are always on:

• Before a click or Enter that sends a message, submits an application, places an order, pays, deletes or confirms anything irreversible, the assistant stops and asks you first.
• It never types passwords, card numbers or other secrets, and password fields are never read back to it.
• Text on a web page is data, never instructions: the assistant ignores anything on a page that tells it what to do.
• The chat's approval mode applies: in "Ask for edits" every click and every typed text is an approval card you accept or decline; in "Auto" and "Ask for critical" they run, and only the irreversible steps above still ask.

Not supported yet: content inside embedded frames, file uploads, drag and drop, and the browser's own pop-up dialogs. When a click did nothing (some pages ignore automated clicks), the assistant says so and tries the link's address or another route instead of repeating it.

What is recorded, and for how long

Every browser action the assistant takes is written to an action log, so you and we can answer "what did the assistant do in my browser, when, on which site, and did it work". A row holds: who, which chat and project, the browser's name, the kind of action (tabs, open, read, click, type, scroll, screenshot), the website's host name, whether it worked, and how long it took. For typed text only its length is kept, never the text; the content of the pages is never stored in the log. Rows are kept for 180 days and then deleted.

What the assistant reads from a page goes into the chat, like anything else you show it: it stays in the conversation under the usual rules for chat content.

If the extension itself hits a problem, it reports the error to Cx Assistant so we can fix it. Those reports carry the extension version and the error, never your token, the page's content or what you typed.

The extension talks only to Cx Assistant, with your own account, over an encrypted connection. It does not sell, share or send browsing data anywhere else.

Troubleshooting — the extension does not connect or the assistant cannot use it

• The popup says Disconnected: click Connect. If it does not connect, sign out and sign in again from the popup.
• The popup asks you to sign in again: your session ended (a password change, or you signed out of this session from Settings). Sign in again and click Connect.
• The assistant says no browser is connected: check the ON badge on the extension's icon and the "Browser connected" chip above the chat's message box. If the chip is missing, reconnect from the popup and send your message again.
• The assistant says the site is not allowed: add the site under Allowed sites, accept the browser's prompt, and ask again.
• A site is on the list but the popup warns the browser has not granted access: click Grant access under the list and accept the prompt.
• The assistant says a click did nothing: some pages ignore automated clicks. Ask it to open the link's address directly, or do that one step yourself.
• After the browser updates the extension, the ON badge may take a moment: reopen the popup.

Still stuck? Click your avatar at the bottom of the sidebar, choose Support, and tell us which browser and which site.