Write by voice
Turn what you say into a draft in a supported text field.
Speak your request — Cue takes the context from your screen and works with you on the result
Free to start · Mac & Windows · GPT-6 Astra, Claude Fable 5.1 and Gemini 3.8
May 2026 Gemma 4 benchmark: 44% lower polish latency → Featured in Google DeepMind's Gemmaverse ↗
Recorded demo; interface and available actions vary by version, connections and permissions
Speak, ask, act or capture
Turn what you say into a draft in a supported text field.
Share a passage and a question. Work through the answer with Cue.
Describe your goal and constraints. Review a draft or the next step.
Record with permission, then review the transcript and key points.
Dictation types your words. For questions and tasks, use Agent. Choose what to share and review the result before using it.
Bring your materials, find a useful next step
Illustrative workflows built from example material
Scenario illustration · example content
Hi Alex — we are ready to move forward with the annual licence. $48,000 a year works, and we would like to start next Monday.
One question: if we commit to a two-year term, is there any flexibility on price?
Best, Maria
Hi Maria — thank you, that is good news.
For a two-year term we can offer 10% off the annual price, so $43,200 a year. No other terms change.
Let me know if that works and I will send the updated agreement.
Scenario illustration · example content
Scenario illustration · example content
7.1 This Agreement begins on the Effective Date and continues for twenty-four (24) months.
7.2 Unless either party gives written notice at least ninety (90) days before the end of the term, the Agreement renews for a further twelve (12) months.
9.3 Each party’s total liability under this Agreement is limited to the fees paid in the twelve (12) months before the claim.
Scenario illustration · example content
Scenario illustration · example content
A 45-second segment from Google’s Connect China event, with English subtitles
The Gemma team used Cue as their example: dictation latency down 44%, feature engagement up 30%, marginal inference cost down to zero, and a demo on a laptop with no internet, running Gemma 4 E4B.
Both figures come from Google DeepMind’s case study.
Choose the context, explain the task, review before acting
Choose Agent mode and check its hotkey in Cue settings. Use the controls shown in your installed version.
Select only context you have permission to share. State the task, preferred format, and limits; don't assume other apps or past tasks are included.
Ask for a draft first. Review facts and wording, then give approval before any supported action that sends, publishes, or changes something.
New to Cue? Follow the voice typing guide, try a first Voice Agent task, or learn voice workflows with Claude Code and Codex.
Choose by the task, the context you share and the controls you need
Official documentation reviewed . These are documented workflows, not a side-by-side performance test. Features depend on the app, version, permissions and plan.
On a narrow screen, scroll the table sideways to read the requirements and sources.
| Product and surface | Documented workflow | Requirements and sources |
|---|---|---|
| Siri on Mac | Voice requests for supported app actions and configured shortcuts. | Compatible Mac, Siri settings and supported actions. |
| ChatGPT desktop | Desktop Voice can direct tasks; supported app integrations and Computer Use provide context or actions. | Voice: eligible paid plans, rollout and workspace settings. Computer Use: enabled plugin, supported region and permissions. |
| Cursor Agent | Dictate a coding prompt; use file, terminal and browser tools. | Check tool availability, review the transcribed prompt and configure action approvals. |
| Cue | Dictate into a supported, focused field, or switch to Agent mode for a specific task. | Configure shortcuts and permissions, provide relevant context and review results. |
Cue is an AI assistant for everyday work on Mac and Windows. Use Dictation to enter text in a supported field, or Agent to ask about material you provide and work on an explicit task. Meeting capture is a separate workflow. Check permissions and review results.
Siri supports app actions and configured shortcuts. ChatGPT desktop offers voice, app integrations and Computer Use subject to plan, rollout and permissions. Cue combines Dictation and Agent modes for supported desktop workflows. Compare the specific task, context and controls you need; review results before taking action.
Cue is free to start. Cue Plus is $19.99/month. Compare plans.
Read Cue's Privacy Policy for details about data and requests.
Cue is available for Mac and Windows. Current releases are 0.7.0 for macOS and 0.5.62 for Windows.
Both Dictation and Agent can use GPT-6 Astra, Claude Fable 5.1 or Gemini 3.8 in Cue. Claude Fable 5.1 requires Cue Plus; see the pricing page and the in-app list for plan details.
Built by
You talk to your computer all day. To Slack, to ChatGPT, to yourself. Why not just tell it what to do?
We got tired of clicking through the same 47 steps every day. So we built Cue — and let AI agents build most of it. One founder, six AI teammates, zero patience for busywork.
Free to start. Ready when you are.