Share the model on your machine with friends: one binary in front of llama.cpp, vLLM, Ollama or LM Studio; one invite code; they chat from a browser. Self-hosted, end-to-end encrypted, no accounts.