Claude Code For Free with LM Studio

Описание: Run Claude Code locally on your Mac Mini! 🚀
In this video, we’re bridging the Claude Code CLI with LM Studio to create a powerful, private, and local coding environment optimized for Apple Silicon. This is perfect for Mac Mini users who want to save on API costs and keep their data on-device.

If you enjoy this deep dive into AI dev tools, like and subscribe for future posts on Claude Code and local LLM workflows!

🔗 Resources & Downloads
LM Studio: https://lmstudio.ai
Claude Code Documentation: https://code.claude.com

🤖 Recommended Models (Search for "MLX" versions)
Llama-3.3-8b-4bit
Qwen2.5-coder-7b
Note: Set Context Length to 32,768 (32k) in LM Studio settings.

💻 Setup Commands
1. Install Claude Code
Via Homebrew: brew install --cask claude-code
Via Official Script: Visit https://code.claude.com/docs/en/setup

2. Environment Configuration
Add these lines to your ~/.zshrc (or ~/.bashrc):

export ANTHROPIC_BASE_URL="http://localhost:1234/v1"
export ANTHROPIC_AUTH_TOKEN="lmstudio"

3. Bypass Onboarding
Run this in your terminal to skip the initial setup prompts:

Type '{"hasCompletedOnboarding": true}' into ~/.claude.json before running

⚡ The "Claude-Local" Shortcut

Paste this function into the bottom of your ~/.zshrc file. This allows you to just type claude-local and it will automatically detect whichever model you have loaded in LM Studio!

function claude-local() {
1. Automatically grab the current running model name from LM Studio
local MODEL_NAME
MODEL_NAME=$(lms ps | awk '$1 != "IDENTIFIER" && $1 != "" && $1 !~ /^-+$/ {print $1; exit}')

if [[ -z "$MODEL_NAME" ]]; then
echo "❌ Error: No model detected in LM Studio."
echo "Ensure your model is loaded and 'lms ps' shows it as RUNNING."
return 1
fi

echo "🤖 Connecting Claude to: $MODEL_NAME"

Launch Claude with the detected model
Add --dangerously-skip-permissions if you want to skip file prompts
claude --model "$MODEL_NAME" "$@"
}

Don't forget to run "source ~/.zshrc" after saving!