Agent Mode
Use AI to transform, translate, summarize, or rewrite text with your voice.
Agent Mode (agentic dictation) lets you interact with AI using your voice. Select text in any app, speak a command, and get an AI-generated response without leaving your current workflow.
Reply to this saying the proposal looks good and we can start next week.
Hi Sarah, the proposal looks good to us. We are happy to move forward and can start next week. Let me know what you need from our side to kick things off.
How it works
- Select text in any application (optional, you can also use Agent Mode without a selection).
- Activate Agent Mode using the agent hotkey.
- Speak your request, for example "translate this to French" or "summarize this in two sentences".
- AI processes your request. The agent window appears showing the streaming response, along with the text you had selected.
- Paste the result. Press Enter to paste the response into your app.
Recording modes
Like dictation, Agent Mode offers two recording styles:
Hold the agent hotkey to speak, release to process.
Best for quick commands where you know exactly what to say.
| Action | What happens |
|---|---|
| Select text (optional) | Captures context for the agent |
| Hold agent hotkey | Recording starts |
| Speak your command | Audio is captured |
| Release hotkey | Command is processed, response streams in |
Default hotkey: Right ⌥ (macOS) / Ctrl + Alt (Windows)
Press once to start, press again to process.
Better for longer or more complex instructions.
| Action | What happens |
|---|---|
| Select text (optional) | Captures context for the agent |
| Press agent hotkey | Recording starts |
| Speak your command | Audio is captured |
| Press hotkey again or Enter | Command is processed, response streams in |
| Press Esc | Cancels the command |
Default hotkey: Right ⌥ + Space (macOS) / Ctrl + Alt + Shift (Windows)
The agent window
When Agent Mode processes your request, a floating window appears at the bottom-center of your screen. This window:
- Shows the AI response as it streams in, with full markdown formatting
- Shows the text you had selected as context
- Stays on top of other windows without stealing focus from the app you're working in
- Can be dragged to a different position by its header
Keyboard shortcuts in the agent window
| Key | Action |
|---|---|
| Enter | Paste the response into your app |
| ⌘ + C (macOS) / Ctrl + C (Windows) | Copy the response to the clipboard |
| Hold agent hotkey (while window is open) | Speak a follow-up to refine the response |
| Esc | Close the window |
Example use cases
| Voice command | What it does |
|---|---|
| "Translate this to Spanish" | Translates selected text to Spanish |
| "Make this more formal" | Rewrites the selection in a formal tone |
| "Summarize this" | Creates a concise summary |
| "Fix the grammar" | Corrects grammar and spelling |
| "Explain this code" | Provides an explanation of selected code |
| "Write a reply to this email" | Drafts a response based on the email content |
| "Convert to bullet points" | Restructures text as a bulleted list |
No selection needed
You can use Agent Mode without selecting text first. Activate it and ask any question.
Refining responses
After receiving an initial response, you can continue the conversation. While the agent window is open, hold the agent hotkey again and speak a follow-up. The agent remembers the context of your previous exchange.
For example:
- Select a paragraph, then say "Summarize this"
- Read the summary, hold the agent hotkey again, then say "Make it shorter"
- The agent refines the summary further
Hands-free voice trigger
You can also reach the agent mid-dictation by saying "Hey Glaido" followed by your command. See Voice Activation.
Configuring agent hotkeys
You can customize agent hotkeys in Settings > Hotkeys. See Hotkeys for details on configuring keyboard shortcuts.