Floating menu

Select text on any page to translate, read aloud, explain, summarize, or copy it — with AI results streamed into pinnable inline cards.

The floating menu turns any text on a web page into an AI action. Select a phrase, sentence, or block, and a small menu appears next to your selection with one-click actions. Results stream into a card on the page, anchored to your selection — you never lose your spot.

Overview

The floating menu is an in-page feature, so it needs permission to run on the sites you visit. Once enabled, selecting text on any allowed page pops up the menu. The menu starts as a small sparkle pill and expands to a row of actions on hover.

Because it runs on the page, the floating menu is subject to Cordy's optional site access — chat and other panel features work without it, but the selection menu does not.

Turn it on

Two settings control the menu, both under Settings → General:

  1. In-Page Features — grant Cordy access to run on websites. Click Enable and approve Chrome's prompt. Required for the menu to appear at all.
  2. Text Selection Menu — toggle Show menu on text selection. This stores your preference and takes effect as soon as site access is granted. Turn it off if you'd rather not see the menu when selecting text — the right-click option and command palette still work.

Using it

  1. Select text on any allowed page (selections over ~10,000 characters are trimmed before sending).
  2. The floating menu appears near your selection.
  3. Click an action.
  4. The result streams into an inline card. From there you can copy the result, continue it in chat, or dismiss it.

The bubble offers five actions:

Translate

Translates the selection into your default output language (set in Settings → General → AI Output Language), using your configured Translation Model. If you have local AI enabled, the result card also offers a Local Translate option that runs on-device instead of in the cloud.

Read aloud

Speaks the selection with Cordy's text-to-speech. Playback opens in a small capsule pinned to the edge of the page, with play, pause, resume, and stop controls. You can use a local voice (Kokoro), cloud TTS, or the browser's built-in speech — see Read aloud (TTS).

Explain

Explains the selection in plain language — great for jargon, dense sentences, or unfamiliar code. Uses your Explanation Model.

Summarize

Condenses a long selection into its key points. Uses your Summarization Model.

Copy

Copies the selected text to your clipboard — a quick way to grab clean text with no AI involved.

Inline result cards

Translate, explain, and summarize stream their answer into a card pinned near your selection — not a separate window:

  • The answer streams in as it's generated.
  • Cards are draggable, resizable, and pinnable. Pin a card to keep it open (it survives clicking away, Esc, and new translations) so you can compare several selections side by side.
  • The first time an on-device model is needed, the card shows a "Downloading on-device model…" state with a shortcut to Local AI settings.
  • The card footer has Copy (copies the AI result) and, once finished, Continue in Chat — which forwards your prompt and the generated result into the side-panel chat so you can keep going with full context.

Sending a selection to chat

There are three ways to move selected text into a conversation:

  • Continue in Chat on a result card (carries the AI's output plus your prompt).
  • Right-click → Quote to AI — opens the side panel and inserts the quoted text with its source URL. This works even with the floating menu turned off.
  • Command palette → Quote selection — with text selected, open the palette (Ctrl/Cmd + Shift + P) and choose Quote selection. See Command palette.

Which model runs each action

The floating menu uses your per-feature model routing, so translate, explain, and summarize can each use a different model (for example, a fast, cheap model for translation). Configure these in Settings → AI Config → Scenario Model Routing — see Providers & models.

Privacy

The floating menu only reads the text you select, and only when you click an action. Nothing is sent in the background. Cloud actions go to the provider you configured (and the first one triggers the one-time consent dialog); local actions stay entirely on your device.

Troubleshooting

  • No menu appears when I select text. Site access isn't granted, or the Text Selection Menu toggle is off. Check both under Settings → General. After a Cordy upgrade you may need to re-enable site access once.
  • The menu is missing on a specific site. Some pages (Chrome system pages, the Web Store, and pages that block content scripts) don't allow extensions to run.
  • "Downloading on-device model…" never finishes. The local model download stalled or WebGPU isn't available — see Local & on-device AI, or switch that action to a cloud model.