Install

Download the current platform artifact from the download page. Release links are resolved from the live manifest, so use the link shown for your platform instead of relying on an old filename.

Supported platforms

The Download page is the source of truth for which installers are currently published. A platform may show “coming soon” while its release artifact is being prepared.

Install the desktop app

1. Open the Download page and choose an available installer for your operating system.
2. Run the downloaded installer and finish the setup wizard.
3. Launch Agent Wispr from your application launcher, Start menu, or installed app list. The app stays available from its floating widget or system tray/menu bar after launch.

Windows

Run the .exe installer and allow Windows to complete the installation. On first launch, allow microphone access when Windows prompts you. If you need to dictate into a terminal or editor running as Administrator, launch Agent Wispr with the same privilege level; Windows blocks lower-privilege apps from sending input to elevated windows.

macOS

Open the .dmg installer and move Agent Wispr to Applications. On first launch, allow microphone access. In System Settings → Privacy & Security, enable Agent Wispr under Microphone and Input Monitoring when prompted. Quit and reopen the app after changing either permission.

Linux

Make the downloaded AppImage executable, then launch it from your file manager or terminal. Allow microphone access if your desktop environment asks. If the app cannot see your microphone or inject text into the focused window, verify that your desktop session exposes the input device and that the application is running in the same graphical session as the target window.

Configure Agent Wispr

After the app opens:

1. Open Settings → Microphone, choose your input device, and run the local microphone test.
2. Open Settings → Model and download a local model. Smaller models use less memory and respond faster; larger models generally improve accuracy but need more disk space and processing time.
3. Open Settings → General and choose a recording mode. Toggle mode starts and stops recording with the hotkey; push-to-talk records while the hotkey is held.
4. Test a short phrase in a text editor or browser field. The transcript is inserted into whichever window is focused when recording stops.

First launch

1. Open Agent Wispr and allow microphone access when your operating system asks.
2. Select a local transcription model in Settings → Model.
3. Choose push-to-talk or toggle recording in Settings → General.
4. Test a short phrase in a disposable text field before using a terminal or elevated application.

The first model download requires an internet connection. After the model is installed, ordinary transcription runs locally. Optional AI Cleanup is a separate online feature and is covered in the AI Cleanup guide.

Permissions

• On macOS, allow microphone and input access in System Settings → Privacy & Security, then restart Agent Wispr.
• On Windows, run Wispr with the same privilege level as an elevated target application when text injection is blocked.
• On Linux, verify that the desktop session exposes an input device and permits the application to send text to the focused window.

If setup does not work

• If the microphone is silent, return to Settings → Microphone, select the device explicitly, and run the test again.
• If a transcript is not inserted, focus a simple text field first, then check the target app's privilege level and accessibility/input permissions.
• If transcription is slow, choose a smaller model or select CUDA in the model settings when an NVIDIA GPU is available.
• If the app appears to be running already, check the system tray or menu bar; the floating widget may be hidden there.