Desktop tools FAQ

All of them are free and open source. To install them, see Install a desktop tool.

WhisperKey

How do I use it?

Press Ctrl+Space from wherever you are to start recording; speak, then press it again to stop. The text is typed where your cursor was: an editor, a terminal, a form, a chat box.

Is my audio uploaded anywhere?

No. The recording is saved to a local file and transcribed by a local process, so there’s no server to upload it to. Once setup has finished, the app makes no outbound connections; the only internet use is downloading the model the first time, which takes about forty seconds.

Does it understand Spanish?

As it ships, no: it comes with the Whisper small.en model and only understands English. The model and the language are settings, so they can be changed.

Can I change the hotkey?

Yes, in Settings. If the combination you pick is already held by another application, the one you had keeps working and the app tells you why, instead of leaving you with no hotkey.

What if the transcription fails?

Every recording is written to disk before transcription starts. If the transcriber times out or the text lands in the wrong window, the audio is still there and History retries it with one click.

Does it need a graphics card?

No. It runs on the CPU, faster than real time on an ordinary laptop: a thirty-second clip takes a few seconds.

Lookout

What is it, and what camera do I need?

Lookout lets your AI assistant see what’s on your desk — a circuit board, a printed page, a device with no API — through a camera plugged into your computer. Any UVC webcam works, which is almost any webcam. No drivers and no ffmpeg needed.

How do I ask my assistant to use it?

The installer drops a skill into your agent’s skills folder. After that you can ask, for example, “look at my desk and check the red wire”. There’s no MCP server or API key to set up. From the command line, lookout look --out /tmp/view.jpg takes a picture and lookout focus focuses the lens.

Does it record without me knowing?

No. A frame exists only because something asked for one: there’s no background capture and no memory of earlier frames. The API listens only on 127.0.0.1 and the app makes no outbound connections. And unplugging the camera is still the strongest guarantee.

What can it see, and what can’t it?

• Reliable: whether a cable is plugged in and where, whether an LED is lit and what colour, whether a component is present or missing, roughly what is where on a board, large printed text, the general state of a workspace.
• Not reliable: resistor colour bands, tiny SMD markings, small print, anything behind something else, anything at a steep angle, anything in shadow.

Does closing the window stop it?

It lives in the system tray and closes to the tray. The window is optional: it’s a viewfinder for aiming the camera.

Audio Engine

What does it do?

It’s an AI audio engine that runs on your machine: speech to text, text to speech and voice cloning from a short reference clip, powered by audio.cpp. Once installed it exposes an OpenAI-compatible API on 127.0.0.1 — on port 8080 in the examples on its page — for your own apps, and an MCP server for your language models.

What happens the first time I open it?

The engine ships inside the app, but the models, about 390 MB, download themselves the first time. After that it works offline.

Can I use it without writing code?

Yes. The app has a panel for each job: Speak (type or paste text, pick a voice and speed, and get a voiceover you can play and save), Transcribe (record from your mic or drop in an audio file and get the text back) and Podcast (paste notes, split them into a back-and-forth, assign two voices and render a whole show).

What about Windows?

There’s an installer. It’s cross-compiled from Linux, so SmartScreen will warn you (More info → Run anyway). If you’d rather build it yourself on Windows, there’s a script for that on the Audio Engine page.

Vibe Builder

Does it need internet?

Yes. The desktop app is a thin native window around the hosted builder at builder.vibedout.app; the engine that writes and publishes the code runs in the cloud. It needs internet because what it builds ends up at a public address, and a public address can’t live on your computer.

Do I need an account?

Yes: you sign in once with your vibedout account and the app remembers you.

Where does what I build end up?

You describe the app in plain language, watch it being generated with a live preview and, when it works, publish it with one click: it goes live at its own <name>.vibedout.app address, a real URL you can send to anyone.

Is it different from builder.vibedout.app?

It’s the same builder, in its own window, without the browser chrome or the risk of losing the session among forty tabs.

DIY Platform

What is it?

A desktop workbench for building software: terminals, git, run scripts and a knowledge graph of your own code, with AI agents working in the same panes instead of in a separate chat. You can hand a terminal to an agent mid-task and take it back.

What runs on my machine and what runs in the cloud?

Most of it runs on your machine: the editor, git, terminal sessions and your project’s files. AI agent calls, and anything you choose to publish, talk to a server.

Do I need an AI key?

No. Workers AI is included by default, with no key, no account and no cost to try it. If you want a different model, you can use your own Claude, GPT or Gemini key from a project’s AI settings.

What about Windows?

The installer is built on a real Windows machine, not cross-compiled from Linux. It’s still unsigned, so SmartScreen will warn you the first time.