Cue Use cases Voice coding in Cursor and desktop editors

Voice coding in Cursor and desktop editors

In Cursor or another supported desktop editor on Mac or Windows, select or show the relevant code and describe the change you want. Review and edit the proposed code before you save it or run tests.

  • Agent proposes, you review
  • Mac and Windows
  • Your tests decide
Illustration: at a home desk in the evening, a developer gestures toward his monitor while explaining a code change out loud.
IllustrationGenerated with Lovart for this guide. It is not a screenshot of Cue or of a real result.
Voice coding can help you describe a change while the relevant code is already in front of you. Editor and text-field behavior varies by current Cue version, and you remain responsible for reviewing the implementation.

3 example scenarios

Refactoring a component
You say: "Draft a version of this component that uses a reducer for the form state"
Cue does: Cue uses the code context you provide to prepare a proposed rewrite. Review the diff and behavior before applying it.
Reading an error
You say: "Explain this stack trace and suggest the next checks"
Cue does: Cue prepares an explanation and possible checks from the visible or selected context. Verify the diagnosis against the source and runtime.
Drafting a utility
You say: "Draft typed debounce and throttle helpers for this project"
Cue does: Cue prepares code for you to review, adapt to the repository conventions, and validate with your own tests.

How Cue does it

If you choose Claude, Codex or Gemini as the agent in Cue, that agent keeps its own sign-in, permissions, available models and billing. Choosing it in Cue does not connect or authorize it.

Illustrative exampleWritten for this guide with fictional details. It is not output from a recorded Cue run.

You say to Agent, with the function selected

“Extract the retry logic into its own function and add a max-attempts argument.”

Proposed change to review

- async function fetchReport(id) {-   for (let i = 0; i < 3; i++) {-     try { return await api.get(`/reports/${id}`); }-     catch (err) { await sleep(500 * 2 ** i); }-   }-   throw new Error('Report unavailable');- }+ async function withRetry(task, maxAttempts = 3) {+   for (let attempt = 1; ; attempt++) {+     try { return await task(); }+     catch (err) {+       if (attempt >= maxAttempts) throw err;+       await sleep(500 * 2 ** (attempt - 1));+     }+   }+ }+ const fetchReport = (id) => withRetry(() => api.get(`/reports/${id}`));
Behaviour change to catch
The old function threw “Report unavailable” after the last attempt; the proposal rethrows whatever error the last attempt hit. Decide which one callers expect.
Not done for you
Nothing is saved, applied or committed. Apply the change yourself, then run the tests.

Try Cue

Download for Mac or Windows. Cue is free to start.

Get Cue →

FAQ

Can I use Cue for voice coding in Cursor or VS Code?

Try Cue in Cursor, VS Code, or another desktop editor or text field available on your supported Mac or Windows setup. This page does not promise a dedicated editor integration. Behavior varies, so test your current Cue version with your own editor, permissions, and project.

How do I start a voice coding request?

Use the Cue hotkey configured for your current app version, then describe the change while the relevant code is selected or visible. Check the current product instructions for shortcut details.

Which spoken languages work for voice coding?

Language support can vary by product version and workflow. Test Cue with your own language, technical vocabulary, microphone, and codebase before relying on it.

Should I apply generated code without review?

No. Review the proposed code, adapt it to your repository, and run the same tests and checks you would use for any other code change.