Chrome Connector

Kendr, in the browser you are already signed in to.

Kendr adds a side panel to Chrome: a chat that works in the tab in front of you. It reads pages you share, fills forms, opens and groups tabs, attaches screenshots you paste, and can learn a workflow by watching you do it once. Sign-in is your existing kendr.org session — nothing to register. Hosted runs can also click and type through the linked extension; Kendr Desktop remains an optional local path. Both routes enforce the same approval, tab-scope, and protected-field boundaries.

Download connector Download connector Download page

Side panel Your Chrome profile Local bridge Approval gated Version 0.10.1 Current release

What the connector is

Kendr Desktop already includes its own built-in browser worker for ordinary web tasks. The Chrome Connector is the separate path for tasks that need your real Chrome session, such as a logged-in SaaS dashboard, internal console, or documentation site behind an account.

Extension

Runs inside Chrome

Lists tabs, captures visible state, and executes approved browser actions in pages you permit.

Native host

Stays on your computer

Connects the extension to Kendr Desktop through Chrome Native Messaging without exposing the local API token to extension JavaScript.

Desktop connector

Appears in Kendr tools

Kendr can request tab snapshots and browser actions through the same local connector approval flow as other tools.

Built-in browser

Still the default

Use the built-in browser for normal browsing automation. Use Chrome only when the existing signed-in profile matters.

Install from kendr.org

Until the Chrome Web Store listing is live, the extension is loaded unpacked. Two steps get it working; the third is only for driving tabs.

1
Load the extension
Download kendr-chrome-connector-0.10.1.zip and extract it. Open chrome://extensions, enable Developer mode, choose Load unpacked, and select the extracted extension folder.
1
Load the extension
Download the current Kendr for Chrome package and extract it. Open chrome://extensions, enable Developer mode, choose Load unpacked, and select the extracted extension folder.
2
Sign in at kendr.org
Click the Kendr icon and press Sign in at kendr.org. The panel authenticates with the session cookie the site already holds, so there is no token to paste and no host to register. Signing out on the site signs the panel out too.

Chrome itself

Kendr does not redistribute Google Chrome. Install Chrome from google.com/chrome, then install the Kendr extension from kendr.org.

Optional: acting in tabs

A signed-in panel can read a page you share with it and answer about it. Clicking, typing, navigating and grouping tabs happen through Kendr Desktop over Chrome Native Messaging, which keeps that work — and the local API token — on your own machine. Set it up only if you want it.

1
Install Kendr Desktop
Download and open the current Desktop build from kendr.org/download.html.
2
Register the native host on Windows
Run the installer from the extracted bundle. It already knows the extension’s ID — that ID is fixed for every install, so there is nothing to copy.
3
Open Kendr Desktop
Signed in to the same kendr.org account. There is nothing to press: the panel polls for the native host and picks Desktop up within about half a minute of it starting. Name this Chrome profile in the panel’s ⋮ menu (for example “Work”) and Kendr shows that name whenever it acts in this browser.

powershell -ExecutionPolicy Bypass -File .\native-host\install-windows.ps1

Using the side panel

Click the Kendr icon in Chrome’s toolbar to open the panel. It has no account of its own — it calls kendr.org as the person already signed in there, and promotes itself to Kendr Desktop when that is running so it can act in tabs as well as read them. The ⋮ menu shows which of the two is serving the conversation, and what this browser has shared.

Two acting modes

Ask before acting pauses for your approval before anything changes a page. Act without asking lets Kendr work straight through. Either way it refuses to type into password, card, or token fields — those stay yours.

Attach and paste

The + menu attaches an image, PDF, or document, and Ctrl+V attaches a screenshot straight from your clipboard. Files travel to Kendr Desktop in pieces, so large captures work without giving Chrome an API token.

Teach Kendr a workflow

Press the teach icon, narrate what you are doing, and walk through the task once. Kendr records both the narration and the steps, then writes a workflow description into the composer for you to read and edit before anything is repeated. Credential fields are recorded as omitted, never as values.

You always know which Chrome

Chrome tells an extension nothing about the profile it runs in, so you name it once in the ⋮ menu (“Work”, “Personal”). Kendr shows that name whenever it acts, so an agent is never quietly driving an account you did not mean.

Connector capabilities

Tool	Purpose	Approval expectation
list_tabs	See open tab IDs, titles, URLs, and active state.	Read-only.
active_tab	Find the currently active Chrome tab.	Read-only.
snapshot	Read a bounded page snapshot with element references for later actions.	Read-only.
capture_screenshot	Capture the scoped tab for visual inspection; protected or opaque fields make the capture fail closed.	Read-only.
list_tab_groups	See the tab groups in the window.	Read-only.
list_downloads	See recent downloads and their state.	Read-only.
navigate / open_tab	Move a tab to a URL, or open a new one.	Approval required.
click	Click a referenced element using a real browser input event.	Approval required.
type	Type into a referenced element using real input. Password, card, and token fields are refused.	Approval required.
press_key	Send a key press such as Enter or Escape.	Approval required.
group_tabs / close_tabs	Gather a task’s pages into a named group, or close tabs.	Approval required.
upload_file	Place a local file into a page’s upload field.	Approval required.
wait_for_download	Wait for a download the page started to finish.	Approval required.

Privacy and safety boundary

The extension does not export Chrome cookies, local storage, session storage, or saved passwords. Its kendr.org calls are authenticated by the HttpOnly session cookie Chrome attaches, which extension JavaScript cannot read.
The extension never receives the Kendr Desktop loopback bearer token; the native host keeps that credential local.
Page text reaches a kendr.org run only for origins you have explicitly shared from the ⋮ menu. Browsing a site is not sharing it.
Snapshots are bounded and meant for task context, not full-page archival scraping.
Actions that change a tab, type text, click controls, or navigate should pass through Kendr's approval flow.
Use the connector only for Chrome profiles and sites you trust Kendr Desktop to operate.

You will see a debugging bar

When Kendr clicks or types, Chrome briefly shows a yellow “Kendr is debugging this browser” bar. That is expected and it is the point: it means Chrome itself generated the click or keystroke, so pages that reject simulated input behave normally. Kendr attaches for the action and detaches immediately, so the bar appears and disappears rather than staying put. Nothing is recorded while it is shown.

What it cannot do

Chrome blocks every extension from acting on its own pages, so Kendr cannot touch chrome://settings, chrome://extensions, or the Chrome Web Store. It also cannot act on a site until you grant that site, and it works only in the Chrome profile where you installed it.

Chrome references

The connector follows Chrome's extension platform boundaries. Chrome documents the Native Messaging manifest and host registration model in the Native Messaging guide, and documents tab metadata and permissions in the Tabs API reference.