Dictate to your agents, audio stays on your computer
Describe a long task instead of typing it. Every agent’s composer has a mic button; whisper.cpp turns your speech into text on your machine and inserts it at the cursor.
Fully local
Recording and transcription happen on your computer; nothing is sent to a server. whisper.cpp is built for each platform: Metal-accelerated on macOS, and a build with no extra runtime to install on Windows and Linux; x64 machines need a CPU with AVX2. The packaged app ships with the small model. Otherwise it uses whisper-cli from your PATH, and if no model is found the app offers to download the base model.
Using it
Click the mic once and speak, then click again when you’re done. The text lands at the cursor, so you can edit before sending. The language follows the app’s interface language (Turkish or English).
Questions
Is my audio sent anywhere?
No. whisper.cpp transcribes it on your machine.
Do I need the internet?
The packaged app ships with the model. If there’s no model, you need to download it once.
Which languages does it understand?
It uses the app’s interface language: Turkish or English.
Related features
Bring your agents to one desk.
Start on the free plan. Download it, add a project folder and open your first agent.