At its core LocaLM is a model loader plus a plugin engine: GGUF and HuggingFace models (which ones) through an OpenAI-compatible server, with chat always on and everything else - the coder agent, image/music/video generation, Knowledge (RAG), durable memory, voice, scheduled jobs, and MCP - a plugin you install when you want it. See Offline first for exactly what does and does not touch the network.

Quick start

# Install (Linux/macOS) - or on Windows, clone and double-click setup.bat
curl -fsSL https://raw.githubusercontent.com/Matlan1/localm/master/install.sh | bash

localm pull owner/repo:model-Q4_K_M.gguf   # Download a GGUF from HuggingFace
localm gui                                  # Chat + coder + plugins in your browser
localm run mymodel --prompt "Explain RDNA2" # Offline terminal chat or single prompt
localm serve mymodel                        # OpenAI-compatible API server

The one-click install puts localm in ~/localm, not on your PATH: run ~/localm/localm.sh <args> (on Windows, localm.bat <args> from the clone), or answer y to “Make ‘localm’ runnable from any…