THE FIELD GUIDES
AI on a Mac
AI on a Mac
Set up your Mac once: Homebrew, Git, Node, Claude Code, and GitHub and Vercel to publish what you build.
Set up your Mac
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"brew install git nodecurl -fsSL https://bun.sh/install | bashnpm install -g @anthropic-ai/claude-codeclaudecp -r ~/.claude ~/.claude-backupcurl -fsSL https://ourlifeos.ai/install.sh | bashclaude and paste this, and Claude does the install for you:Read https://ourlifeos.ai/install and install LifeOS for me.GitHub and Vercel
brew install ghgh auth loginnpm install -g vercelvercel loginAI cloud
AI that runs on the companies' servers. You sign in or use an API key, and you need internet.
The quickest way: chat in your browser
CLI AI coding tools
Every AI coding tool you can run in Terminal. Works on every Apple Silicon Mac (M1 to M5). Do steps 1 and 2 of "Set up your Mac" first. Install, then start it with the second line. Each asks you to sign in or paste an API key the first time.
brew install uvkimi or vibe isn't found after installing, run this and reopen Terminal:uv tool update-shellnpm install -g @anthropic-ai/claude-codeclaudenpm install -g @openai/codexcodexnpm install -g @google/gemini-cligemininpm install -g @github/copilotcopilot/login.curl -fsSL https://x.ai/cli/install.sh | bashgrokuv tool install kimi-clikimi/login and paste your Kimi API key.npm install -g @qwen-code/qwen-code@latestqwen/auth. It needs an API key.uv tool install mistral-vibevibenpm install -g opencode-aiopencode/connect to sign in to a provider.npm install -g @vegamo/deepcode-clideepcodeAI offline
Free AI models that run on your Mac itself. Nothing leaves your computer, and it works without internet.
The quickest way: in your browser
Ollama: free models on your own Mac
brew install --cask ollamaollama run gemma3:4bollama launch claudelaunchctl setenv OLLAMA_ORIGINS "https://chathonu.com"http://localhost:11434Run Claude Code fully offline with a local model
This is how I ran Claude Code on my Mac with no internet, no API key and no subscription in the video. You point Claude Code at a coding model running in Ollama. It's slower than the cloud (a big answer can take a few minutes on-device), but it works in airplane mode.
ollama pull qwen2.5-coder:7bnano ModelfileFROM qwen2.5-coder:7b
# A coding buddy with a Sword-Art-Online gamer personality.
SYSTEM """
You are Kirito, a sharp and upbeat coding assistant.
Explain code like a game: bugs are monsters, debugging is clearing a dungeon.
Always give working, well-commented code and a short, clear explanation.
Stay positive and keep the player moving.
"""
PARAMETER temperature 0.7
PARAMETER num_ctx 8192ollama create kirito -f Modelfileollama run kiritoLater, to see this Modelfile again: ollama show kirito --modelfile. To back it up: ollama show kirito --modelfile > ~/Documents/kirito.Modelfile.
kirito, or qwen2.5-coder:7b if you skipped the Modelfile):export ANTHROPIC_BASE_URL="http://localhost:11434"export ANTHROPIC_MODEL="kirito"~/.zshrc, then reopen Terminal.echo 'export DISABLE_AUTOUPDATER=1' >> ~/.zshrc && source ~/.zshrcclaudeUse your Mac's Ollama from your phone
brew install --cask tailscaletailscale serve --bg 11434If the tailscale command isn't found, open the Tailscale app once first.
Stuck on a step? Send a message and say which one.