Cursor
The hands-free workflow with spoken updates, custom voices, and dictation.
CURSOR GUIDE →Cross-platform text-to-speech for Cursor and compatible MCP coding hosts, plus an optional hands-free Cursor workflow.
$ npx -y talktocursor
Your assistant speaks progress, completions, and errors through TalkToCursor Cloud, ElevenLabs, or local Voicebox with the voice of your choice.
Dictate your instructions and have them automatically submitted. Use built-in on-device speech-to-text, Cloud dictation, Wispr Flow, FluidVoice, or Handy for seamless hands-free control.
Instant Feedback says “let me take a look” the moment you send a prompt. Get progress updates on every task, only long ones, or none, and Auto-Listen starts after the final reply when enabled.
Use managed Cloud voices with no provider key, customize ElevenLabs, or keep synthesis private on your computer.
Pick a casual or professional speaking style, give finished tasks, quick replies, and questions their own sounds, choose Spoken Response Detail, and manage hands-free features.
A verified 8.7 MB local model detects when you finish speaking. Optional wake phrases start a turn; saying “send it” submits immediately.
Save up to 10 paid named profiles per account, assign them by project and coding host, and keep automatic conflict-safe sync with 30-version/90-day history and rollback. Free accounts can pair devices and sync preferences.
With a paid profile, add up to 50 explicit word or phrase replacements that sync with that profile. Spoken text, generated audio, credentials, and local paths are never synced or stored in Cloud.
In the Cloud dashboard, set hard per-device billing-period budgets, review up to 12 periods by day or device, export CSV, and get in-app warnings at 80% and 95% of the included allowance, with optional email alerts.
TalkToCursor is one package and one set of voice settings. TTS works across compatible hosts, while Smart Turn, wake phrases, spoken “send it,” automatic submission, and the Background Helper are specifically for Cursor on macOS.
The hands-free workflow with spoken updates, custom voices, and dictation.
CURSOR GUIDE →Complement native voice dictation with spoken agent updates through MCP.
CLAUDE CODE GUIDE →Add custom or local TTS and spoken milestones alongside native Codex Voice.
CODEX GUIDE →Complement native voice transcription with spoken agent output through MCP.
ANTIGRAVITY GUIDE →Install TalkToCursor for my current coding assistant by following: https://github.com/MindSyncTech/talk-to-cursor/blob/main/INSTALL.md Configure the MCP and persistent voice-feedback instructions. Ask before enabling paid Cloud services or installing the macOS Background Helper.
Add the MCP server to your Cursor configuration file at ~/.cursor/mcp.json:
{
"mcpServers": {
"talktocursor": {
"command": "npx",
"args": ["-y", "--prefer-online", "talktocursor@latest"]
}
}
}
Start the packaged settings server, then open http://localhost:3847 in your browser.
npx -y --prefer-online --package talktocursor@latest talktocursor-settings
Select a text-to-speech provider in Settings:
Fully quit and reopen Cursor. Open a chat and type "Say hello using the speak tool" - you should hear your assistant's voice. Then, just enable dictation or a transcription tool to start talking to your code.
Use the built-in voice input (on-device, or Cloud dictation on the paid plan), or install Wispr Flow or the free local Handy app. Then:
The Background Helper runs without a setup terminal. The packaged talktocursor-auto-submit command remains a manual fallback.
Choose the speech setup that fits your workflow, then manage voices and hands-free controls from one local settings page.
// config.json { "provider": "cloud", "cloud": { "apiUrl": "https://cloud.talktocursor.com", "voiceId": "Kore", "model": "gemini-2.5-flash-tts" }, "autoListen": true, "autoSubmit": { "enabled": true, "silenceDelay": 3.0 } }
Install TalkToCursor in minutes and experience hands-free coding with voice feedback.