Browser Extension

The Komand browser extension brings your AI agents into any web page. It provides a sidebar prompt interface with page-aware context injection — your agent can see and act on the content you’re browsing.

Features

A slide-out panel for conversing with your agent while browsing any website. The panel stays open as you navigate between pages, maintaining conversation continuity.

Page Context

The agent can see and reference the current page content:

Page text — the agent reads the visible page content
Selected text — highlight text and ask the agent about it
Page metadata — title, URL, meta description
Structured data — tables, lists, and other structured content

Quick Actions

Right-click context menu for common operations:

Summarise selection — highlight text and get a summary
Add to CRM — extract contact details from a page
Save to memory — store page content for later reference
Ask about this — open the sidebar with the selected text as context

Auto-Fill

The agent can fill forms and interact with page elements:

Form completion — agent fills out repetitive forms based on your instructions
Data entry — paste structured data from the agent’s response into input fields

Supported Browsers

Browser	Engine	Status
Google Chrome	Manifest V3	In development
Microsoft Edge	Manifest V3	In development
Firefox	WebExtension	Planned
Safari	Web Extension	Planned

Use Cases

Research

Ask your agent to summarise or analyse the page you’re reading:

[Browsing a competitor's pricing page]

You:   "How does this compare to our pricing?"
Agent: "Their Pro plan at $59/mo is 20% more than ours ($49/mo)
        but includes 3 features we don't offer: SSO, audit logs,
        and priority support. Their Enterprise tier starts at $199/mo
        vs our $149/mo."

CRM Enrichment

Visit a LinkedIn profile or company page and tell your agent to capture the information:

[Browsing a LinkedIn profile]

You:   "Add this person as a contact"
Agent: "Added Bob Smith, VP of Engineering at Acme Corp.
        Email: bob@acme.com. I've linked him to the existing
        Acme Corp deal in your pipeline."

Form Filling

Let the agent handle repetitive data entry:

[On a vendor registration form]

You:   "Fill this out with our company details"
Agent: "Done. I've filled in:
        Company: Acme Corp
        ABN: 12 345 678 901
        Address: 123 Main St, Brisbane QLD 4000
        Contact: Jane Smith (jane@acme.com)
        Please review and submit."

Competitive Analysis

Browse competitor sites and ask for structured comparisons:

[Browsing a product page]

You:   "Add this product to our comparison sheet"
Agent: "Added 'Widget Pro X' at $299/mo to your comparison sheet.
        It's 20% more expensive than the current leader.
        Key differences: faster API limits, no SSO, limited integrations."

Architecture

The browser extension communicates with your Komand instance via the same REST API used by the dashboard and CLI:

Browser Extension
  ├── Content Script (page context extraction)
  ├── Sidebar UI (React)
  └── Background Worker (API client)
         │
         ▼
    Komand REST API (/api/webchat/message)

Content script runs in the page context, extracting text and metadata
Sidebar UI renders the chat interface as a React application in a shadow DOM
Background worker handles API communication with the Komand server
Messages include page context as metadata, so the agent knows what you’re looking at

Security

The extension follows browser security best practices:

Minimal permissions — only requests access to the active tab when invoked
No persistent background — Manifest V3 service worker activates on demand
Credential isolation — API token stored in chrome.storage.session (cleared on browser close)
Content Security Policy — strict CSP prevents script injection
No data collection — page content is sent only to your Komand instance, never to third parties