Voice transcription that stays in sync with your plans
Record audio snippets while you plan, capture terminal commands, and configure transcription defaults per project. Everything routes through the same Monaco editor and terminal sessions used elsewhere in the app.
Capture task intent in context
Start recording from the task description panel. Audio is streamed through the useVoiceTranscription hook and saved straight into the Monaco editor so you can refine the text before generating plans. Highlight the transcript and use the text improvement popover if you need to tighten wording before moving on.
- Real-time feedback while recording
- Automatic retries with helpful error messages
- Language, temperature, and model defaults stored per project
Dictate terminal commands safely
The plan terminal modal exposes the same transcription controls. Dictated text is appended to your active PTY session using backpressure-aware writes so long commands land exactly once.
- Start and stop recording without leaving the terminal
- Chunked writes prevent partial commands
- Compatibility with claude, cursor, codex, and gemini sessions
Fine-grained configuration
Adjust the transcription model, temperature, and language for each project. Settings are persisted via the project task settings API so team members share the same defaults.
- Use OpenAI GPT-4o Transcribe or GPT-4o Mini Transcribe
- Apply separate defaults for task descriptions and terminal usage
- Project-level overrides layered on top of server defaults
Robust error handling
The transcription client surfaces clear messages for authentication, network, and provider errors. Logs are persisted so you can review what was sent before retrying.
- Friendly guidance for microphone or auth issues
- Structured error messages mapped from server responses
- Local audit trail alongside plan drafts and terminal logs
Unlock Hands-Free Development
From voice to code, seamlessly capture ideas and execute commands. This is how voice input should work - natural, integrated, powerful.