LlamaChat Local for VS Code

Run a local GGUF chat model in VS Code and use it from GitHub Copilot Chat and Agent mode.

This is not a FIM (fill-in-the-middle) Tab-complete model. Copilot Agent needs a chat model with tool calling, such as Qwen3.8-27B Q8.

Install

1. In VS Code, open Extensions (⇧⌘X) and search for LlamaChat Local, or run ext install llamachat.llamachat-local.
2. Install llama.cpp if needed: Command Palette → LlamaChat: Install llama.cpp (Homebrew on Mac), or brew install llama.cpp.
3. Command Palette → LlamaChat: Choose GGUF Model, then LlamaChat: Start Local Server.
4. Open Copilot Chat and pick LlamaChat Local.
5. For Agent mode, enable chat.agentHost.byokModels.enabled.

Host vs connect

• Host (default): this extension starts llama-server and loads your GGUF.
• Connect: set llamachat.mode to connect and keep the LlamaChat Mac/iPhone app open on port 8765.

Commands

• LlamaChat: Manage Local Model
• LlamaChat: Choose GGUF Model
• LlamaChat: Start Local Server
• LlamaChat: Stop Local Server
• LlamaChat: Install llama.cpp