It also supports voice: hold a key, speak an instruction, and it executes against your selected text or screen context. Transcription runs locally via CoreML (Whisper-based), so nothing leaves your machine.
How it works under the hood: - Text capture/replacement via CGEvent keyboard simulation and clipboard save/restore - Menu bar app with Carbon framework hotkey registration - Supports OpenAI, Anthropic, Google, and Ollama — direct API calls with your own keys, no proxy server - API keys stored in Apple Keychain - Voice uses on-device Parakeet (CoreML Whisper) — no cloud transcription
It's completely free and open source. No subscription, no credits, no account required.
https://github.com/michael-elkabetz/echoo
I'd love feedback from anyone who's tried similar tools — what worked, what didn't, and what you'd want from something like this.
If you want full privacy, point it at Ollama and everything stays on your machine - the AI processing, the voice transcription, all of it.
Happy to answer questions about the implementation. The trickiest parts were reliable text replacement across different app frameworks (Electron vs native Cocoa vs web views) and getting voice activation to feel responsive with on-device models.