What is LocalClicky?

LocalClicky is a Mac menubar app that lets you control your computer entirely with your voice — completely offline. Say "Computer" to start a session, chain commands back to back, and say "goodbye" when you're done. Everything runs on your machine: voice transcription, AI reasoning, vision, and text-to-speech. No API keys, no subscriptions, and no data ever leaves your Mac.

Who it's for

• Privacy-conscious users — want voice control without uploading audio, screenshots, or commands to any cloud server
• Power users and developers — need hands-free Mac control for multitasking, automation, or accessibility workflows
• Offline-first enthusiasts — prefer self-hosted, open-source tools that run entirely on local hardware

Key features

Wake-word-activated sessions

Say "Computer" to start a listening session. The app stays active until you say "goodbye" or another exit phrase. You can chain multiple commands without repeating the wake word — just speak naturally, one after another.

Voice Activity Detection with no fixed timeout

LocalClicky uses WebRTC Voice Activity Detection to automatically stop recording when you stop speaking. No awkward countdowns or cut-off commands. If VAD isn't available, it falls back to a 30-second hard cap.

Vision-powered screen control

The app can see your screen on demand using a local vision model (gemma4:e4b). It takes a screenshot, analyzes what it sees, and can move your cursor or click based on the visual content — all without sending images to any external server.

Multi-round tool calling with memory

LocalClicky runs commands, checks results, and retries or confirms actions automatically. It remembers the last 10 exchanges in a session, so you can refer back to previous commands or results naturally.

What stands out

Your voice, your screen, your commands — nothing leaves your machine.

This isn't just a privacy promise; it's the architectural foundation. Every component — Whisper.cpp for transcription, Ollama for AI reasoning and vision, macOS say for speech output, and PyAutoGUI for cursor control — runs locally. There are no cloud APIs, no API keys to configure, and no subscription fees. The MIT license means you can inspect, modify, and redistribute the code freely.

Worth checking out if…

You want a genuinely private, offline voice assistant for your Mac that can open apps, control Spotify, manage files, run shell commands, inject JavaScript into Chrome, create reminders, and even click things on screen — all by speaking naturally. It's especially compelling if you've avoided cloud voice assistants due to privacy concerns or want to experiment with local AI models like Qwen3 and Gemma4 without leaving your desktop.