Installation

System Requirements

macOS only — RCLI requires Apple Silicon (M1 or later) and macOS 13+. The engine uses Metal GPU for inference and CoreAudio for voice I/O.

Platform

macOS 13 Ventura or later

Chip

Apple Silicon (M1, M2, M3, or later)

Storage

~1GB for default models, more for additional models

RAM

8GB minimum, 16GB recommended

RCLI will not run on Intel Macs. The voice pipeline is optimized specifically for Apple Silicon’s Metal GPU and unified memory architecture.

Installation Methods

Homebrew (Recommended)

Install RCLI via the official Homebrew tap:

Tap the RCLI repository

brew tap RunanywhereAI/rcli https://github.com/RunanywhereAI/RCLI.git

Install RCLI

brew install rcli

This installs the rcli binary and required shared libraries to /opt/homebrew/bin/.

Download AI models

rcli setup

Downloads ~1GB of default models:

LLM: LFM2 1.2B Tool Q4_K_M (731 MB)
STT: Zipformer streaming (50 MB) + Whisper base.en (140 MB)
TTS: Piper Lessac medium (60 MB)
VAD: Silero VAD v5 (0.6 MB)
Embeddings: Snowflake Arctic Embed S (34 MB)

Verify installation

rcli info

This displays engine info, installed models, and system hardware details.

One-Liner Install

Install Homebrew (if needed), RCLI, and models in a single command:

curl -fsSL https://raw.githubusercontent.com/RunanywhereAI/RCLI/main/install.sh | bash

What does the install script do?

The install.sh script:

Checks for macOS and Apple Silicon (exits if not detected)
Installs Homebrew if not already present
Taps the RunanywhereAI/rcli repository
Installs the rcli formula
Runs rcli setup to download AI models (~1GB)

The script is idempotent and safe to re-run.

View the install script source

The install script is open source and can be inspected before running:

curl -fsSL https://raw.githubusercontent.com/RunanywhereAI/RCLI/main/install.sh

Key validation steps:

Platform check: [[ "$(uname -s)" == "Darwin" ]]
Architecture check: [[ "$(uname -m)" == "arm64" ]]
Auto-detects latest release from GitHub API

Model Setup

RCLI requires AI models for STT, LLM, TTS, VAD, and embeddings. The rcli setup command downloads the default model set (~1GB) to ~/Library/RCLI/models/.

Default Models

Model	Provider	Size	Purpose
LFM2 1.2B Tool Q4_K_M	Liquid AI	731 MB	Language model with tool calling
Zipformer	k2-fsa	50 MB	Streaming STT (live microphone)
Whisper base.en	OpenAI	140 MB	Offline STT (batch transcription)
Piper Lessac	Rhasspy	60 MB	Text-to-speech voice
Silero VAD v5	Silero	0.6 MB	Voice activity detection
Snowflake Arctic Embed S	Snowflake	34 MB	Text embeddings for RAG

Manual Model Download

If rcli setup fails or you prefer manual installation, run the download script directly:

bash /opt/homebrew/Cellar/rcli/*/share/scripts/download_models.sh

Or download models manually from the RCLI models panel:

rcli models

Model Storage Location

All models are stored in:

~/Library/RCLI/models/
├── lfm2-1.2b-tool-q4_k_m.gguf
├── whisper-base.en/
├── zipformer/
├── piper-voice/
├── silero_vad.onnx
└── espeak-ng-data/

Active model selection is persisted in ~/Library/RCLI/config.

Post-Installation

Test Microphone Access

RCLI requires microphone permissions. On first run, macOS will prompt for access. Test your microphone setup:

rcli mic-test

This displays real-time audio levels and VAD (voice activity detection) status.

Run Your First Command

Launch the interactive TUI:

rcli

Press SPACE to start recording, speak a command like “What’s the time?”, then release SPACE. RCLI will transcribe, process, and respond. For continuous voice mode:

rcli listen

View Available Actions

RCLI ships with 43 macOS actions. View the full list:

rcli actions

Example actions:

open_app — Open applications by name
create_note — Create Apple Notes
play_on_spotify — Control Spotify playback
set_volume — Adjust system volume
search_web — Open web searches in Safari

Upgrading

Upgrade RCLI Binary

Update RCLI to the latest version via Homebrew:

brew update
brew upgrade rcli

Upgrade Models

RCLI supports 20+ models across LLM, STT, and TTS. Upgrade to larger or more accurate models:

rcli upgrade-llm
# Interactive picker for Qwen3 0.6B/2B/4B, LFM2 2.6B, etc.

Or use the unified model manager:

rcli models

This opens an interactive TUI for browsing, downloading, and switching models across all modalities (LLM, STT, TTS).

Clean Up Unused Models

Remove models you’re not using to free disk space:

rcli cleanup

This shows all installed models and allows you to delete unused ones.

Troubleshooting

Installation fails with 'Command Line Tools' error

If brew install rcli fails on macOS 26+ (beta), the install script includes a fallback that downloads and installs the binary manually:

curl -fsSL https://raw.githubusercontent.com/RunanywhereAI/RCLI/main/install.sh | bash

The script detects Homebrew installation failures and falls back to direct tarball extraction.

Microphone permission denied

RCLI requires microphone access to function. If denied:

Open System Settings → Privacy & Security → Microphone
Enable access for Terminal (or your terminal app)
Restart your terminal and re-run rcli

Test microphone access with rcli mic-test.

Models fail to download

If rcli setup fails due to network issues:

Check your internet connection
Retry: rcli setup
Manual download: bash scripts/download_models.sh (from the RCLI source directory)

Models are downloaded from:

HuggingFace (LLM, embeddings)
GitHub Releases (STT, TTS, VAD via k2-fsa/sherpa-onnx)

rcli not found in PATH

After Homebrew installation, ensure /opt/homebrew/bin is in your PATH:

echo 'eval "$(/opt/homebrew/bin/brew shellenv)"' >> ~/.zshrc
source ~/.zshrc

Verify with which rcli.

Metal GPU not detected

RCLI requires Metal GPU for inference. If you see errors about Metal:

Ensure you’re on Apple Silicon (check with uname -m — should output arm64)
Update macOS to the latest version (minimum macOS 13)
Run rcli info to see detected hardware

Intel Macs are not supported.

Next Steps

Quickstart

Get your first voice command running in 5 minutes

Commands Overview

Learn all RCLI commands and options

Model Management

Explore 20+ supported AI models

macOS Actions

Discover all 43 macOS actions

Get Started

Core Features

Commands

Models

Actions

Advanced

Development

Installation

System Requirements

Platform

Chip

Storage

RAM

Installation Methods

Homebrew (Recommended)

One-Liner Install

Model Setup

Default Models

Manual Model Download

Model Storage Location

Post-Installation

Test Microphone Access

Run Your First Command

View Available Actions

Upgrading

Upgrade RCLI Binary

Upgrade Models

Clean Up Unused Models

Troubleshooting

Next Steps

Quickstart

Commands Overview

Model Management

macOS Actions

Build docs developers (and LLMs) love

Get Started

Core Features

Commands

Models

Actions

Advanced

Development

​System Requirements

Platform

Chip

Storage

RAM

​Installation Methods

​Homebrew (Recommended)

​One-Liner Install

​Model Setup

​Default Models

​Manual Model Download

​Model Storage Location

​Post-Installation

​Test Microphone Access

​Run Your First Command

​View Available Actions

​Upgrading

​Upgrade RCLI Binary

​Upgrade Models

​Clean Up Unused Models

​Troubleshooting

​Next Steps

Quickstart

Commands Overview

Model Management

macOS Actions

Build docs developers (and LLMs) love

System Requirements

Installation Methods

Homebrew (Recommended)

One-Liner Install

Model Setup

Default Models

Manual Model Download

Model Storage Location

Post-Installation

Test Microphone Access

Run Your First Command

View Available Actions

Upgrading

Upgrade RCLI Binary

Upgrade Models

Clean Up Unused Models

Troubleshooting

Next Steps