Installation

The software was tested on MacOS (Sequoia) and UNIX-based systems. Windows wasn’t actively tested as a release candidate.

Requirements

• macOS, Linux, or Windows
• minimum 4GB RAM for interactive browsing mode (requires local inference via Ollama with qwen2.5:3b)

macOS / Linux

# Core (API + local models)
curl -fsSL https://swizzlm.com/install.sh | bash

# Full (core + browser automation + Ollama)
curl -fsSL https://swizzlm.com/install.sh | bash -s -- --full

Windows

via PowerShell

# Core
irm https://swizzlm.com/install.ps1 | iex

# Full
powershell -c "& { irm https://swizzlm.com/install.ps1 } -Full"

via Winget

winget install SwizzLM.Swizz

The binary is installed to ~/.swizz-llm/bin/swizz (macOS/Linux) or %LOCALAPPDATA%\swizz-llm\bin\swizz.exe (Windows) and automatically added to your PATH.

Ollama (optional)

Ollama is used for running local models and interactive browsing. The app will auto-start Ollama if the binary is in your PATH. The --full install script installs Ollama and pulls the required model automatically.

LM Studio (optional)

LM Studio is a desktop app for running local models. You can download and install it to support more options for local inference.