← All 33 supported Mac AI tools

How to pause LM Studio on Mac

LM Studio can keep a local OpenAI-compatible server running, hold large quantized weights in unified RAM, and connect to MCP servers via ~/.lmstudio/mcp.json. Here is how to stop it cleanly.

Local LLM · macOS App · Verified 2026-11-19 · macOS 13+ (Apple silicon)

What AI Stopper detects for LM Studio

  • macOS app bundle: LM Studio.app in /Applications or ~/Applications
  • Snapshotted MCP config: ~/.lmstudio/mcp.json

How to stop LM Studio manually on macOS

  1. In LM Studio, eject loaded models (or stop the Developer local server) so weights release immediately.
  2. Quit LM Studio.app with Command + Q and verify the menu bar icon is gone.
  3. Check for any leftover server or helper processes in Activity Monitor or Terminal.

Check for running processes in Terminal (read-only):

ps -eo pid,ppid,comm,args | grep -i "LM Studio.app" | grep -v grep

Check config / MCP entries:

ls -la "$HOME/.lmstudio/mcp.json" 2>/dev/null
Why manual quitting misses edge cases: Even when the main LM Studio window is closed, its background daemon or local server can keep a 7B–70B model resident in unified memory, pushing macOS into swap during builds or video calls.

What happens when you pause LM Studio with AI Stopper

  • Terminates LM Studio.app and child inference helpers in one click.
  • Snapshots ~/.lmstudio/mcp.json locally during pause and restores it cleanly when you resume.
  • Pauses LM Studio together with Ollama, llama.cpp, Jan, Msty, and DiffusionBee for instant RAM recovery.

Related guides & workflows