New Open-Source Voice AI Tool Runs Entirely On-Device for Apple Silicon Macs With No Cloud Required

A powerful new open-source voice AI tool called RCLI now lets Apple Silicon Mac users run a full speech-to-text, LLM, and text-to-speech pipeline entirely on-device with no cloud or API keys required, supporting 38 voice-controlled actions and ultra-fast local document querying at just 4ms latency.

• RCLI is an open-source, on-device voice AI tool for macOS that runs a full speech-to-text, LLM, and text-to-speech pipeline natively on Apple Silicon with no cloud or API keys required.
• It supports 38 macOS voice-controlled actions, local document querying via hybrid RAG retrieval at ~4ms latency, and a terminal UI for model management — all powered by a proprietary GPU inference engine called MetalRT built for Apple Silicon.
• Installation is available via a single curl command or Homebrew, requires macOS 13+ on Apple Silicon, with M3 or later needed for MetalRT while M1/M2 devices automatically fall back to the open-source llama.cpp engine.