// the find
lmstudio-ai/lms
LM Studio CLI
lms is the official command-line companion to LM Studio, letting you start/stop the local inference server, load and unload models, stream logs, and chat from the terminal instead of the desktop GUI. It's for people already running LM Studio who want scriptable or headless control over it.
Covers real operational needs, not just a thin wrapper: server start/stop, model load/unload with --json output for scripting, log streaming, runtime/device management, and even a plugin dev workflow with separate Deno and Node plugin process handling. It's built on lmstudio.js, so the CLI shares the same client/protocol code as the SDK rather than reimplementing it. The chat subcommand has an actual terminal UI built with React components (ChatInput, ChatMessagesList, slash commands) instead of a bare readline loop.
It cannot be built or run standalone — contributing requires cloning the entire lmstudio.js monorepo recursively, which is a lot of friction for what looks like a simple CLI repo. The README documents maybe 10% of the actual surface area (nothing on daemon, link, runtime, or dev subcommands beyond pointing at --help), so you're exploring blind. Test coverage is thin relative to the number of subcommands — a handful of .test.ts files for dozens of commands. And fundamentally this is a remote control for a closed-source desktop app, so its ceiling is whatever LM Studio's proprietary backend decides to expose.