Local Agents
A Godot 4.5 plugin that runs a local LLM inside a game through a bundled llama.cpp backend, giving an in-game agent node both a text reply and an action a creature or NPC can act on.
Local Agents is a Godot game engine plugin that runs a local large language model inside a game, using llama.cpp as the backend so it runs entirely on the player's machine with no cloud service, API key or internet connection needed. A node loads a .gguf model and produces both a text reply and an action a creature or NPC can act on, so the model can write dialogue live during play and decide what a character does next.
It has three parts: a .gguf model, such as the default quantized Qwen3-4B-Instruct build, a native C++ GDExtension wrapping llama.cpp for text generation, whisper.cpp for transcription and Piper for speech, and the Godot-side plugin itself, which includes the agent node, a memory graph resource and an editor panel for downloading models.
Neither the native extension nor the model is committed to the repository, so a fresh clone needs a downloaded or self-built extension binary plus a model before it will run.