Browser extension
The browser extension puts a Polylane agent next to whatever you have open. It has two surfaces: an in-page overlay on supported provider dashboards, and a side panel that works on any tab.
Install and sign in
The extension ships for Google Chrome and installs from the Chrome Web Store. It uses your existing session from console.polylane.com: sign in there once and the extension picks it up. If you are signed out, the overlay shows a banner; press Enter to open the sign-in page in a new tab.
The overlay
Press Cmd+. (macOS) or Ctrl+. (Windows and Linux) to toggle the overlay on a supported page. Type a question and the agent answers in place, streaming next to the dashboard you are looking at. When the panel is open and you have not started a thread, the extension suggests questions about the current page; press 1, 2, or 3 to send one.
The overlay runs on the Polylane console and on these provider dashboards: Cloudflare, the Amazon Web Services console, Fly.io, Render, Vercel, GitHub, Slack, Sentry, Datadog, Honeycomb, and Axiom.
The side panel
Press Cmd+B (macOS) or Ctrl+B (Windows and Linux), or click the Polylane icon in the toolbar, to open the side panel. It works on any tab. The side panel lists the threads you started from the extension across your workspaces, lets you continue or delete them, and can load a thread into the current tab. Messages sent from the side panel can attach the active tab as context.
Keyboard shortcuts
| Keys | Action |
|---|---|
Cmd+. | Toggle the overlay on the current page |
Cmd+Shift+. | Start a new thread |
Cmd+' (hold) | Push-to-talk: hold to speak, release to send |
Cmd+B | Open the side panel (rebindable at chrome://extensions/shortcuts) |
1 2 3 | Send the matching suggested question |
↑ ↓ | Scroll through the latest response |
Cmd+Shift+O | Open the current thread in the console |
Esc | Close the panel, stop voice, or cancel in-flight work |
On Windows and Linux, use Ctrl in place of Cmd.
What the agent sees
When you send a message, the extension captures a bundle from the active tab and uploads it as a source document on the message:
- the page URL, title, and viewport
- the element under your cursor and the focused element
- your text selection, the visible text, and the interactive elements on screen
It also captures a screenshot of the visible tab, resized to at most 1280 pixels on the long edge, and attaches it alongside the page context.
Password fields and credit-card fields never send their contents: the agent receives the length of the value, not the value. Any element inside a subtree marked data-polylane-private="true" or data-sensitive is skipped entirely. You can turn off page context collection and screenshot capture independently in the extension's options page.
Browser actions
On supported dashboards, the agent can act on the page during a thread: it points at elements, clicks, types, and navigates. It asks for your consent before taking control, the page glows while the agent is acting, and pressing Esc or the stop button cancels in-flight actions immediately. You can disable browser actions in the extension's options page.
Voice
Hold Cmd+' to talk to the agent instead of typing, and release to send. Each voice session starts a fresh thread.
Connecting accounts from a dashboard
When you open a provider dashboard for an account that is not yet connected to a Polylane workspace, the extension offers to connect it. Press Enter and it walks you through the provider's flow: creating an API token, approving an OAuth install, or deploying the CloudFormation stack for Amazon Web Services. When you create a token on the provider's page, the extension detects it, validates it, and finishes the connection without a manual paste. See Clouds and Integrations for what each connection unlocks.