ggml-org/llama.vscode

What is it?

Key features

• 01Auto-suggest completions as you type, with keys to accept a full suggestion, one line, or one word
• 02Runs FIM-compatible models locally through llama.cpp, with automatic llama.cpp install on Mac and Windows
• 03Llama Agent for agentic coding with internal tools, custom JavaScript tools, and MCP server support
• 04Model and env management, including import and export, for completion, chat, embeddings, and tools
• 05Search and download models from Hugging Face inside the extension

Use cases

• Getting private, offline code completion from a local model while editing in VS Code
• Chatting with a local or external model about code without leaving the editor
• Running an agent in the editor that uses MCP tools already set up in VS Code
• Trying different completion or chat models on consumer hardware with low VRAM settings

Quick start

1. 1

   Install the llama-vscode extension from the VS Code marketplace

2. 2

   Install llama.cpp on macOS with Homebrew

   brew install llama.cpp
3. 3

   Start the llama.cpp server with a default FIM model (suited to 16GB+ VRAM)

   llama serve --fim-qwen-7b-default

Things to know

• !Linux users must install llama.cpp manually from binaries or source.
• !Completion requires FIM-compatible models and enough VRAM or RAM for good speed.

In collections

Topics

FAQ

What is llama.vscode?

VS Code extension for local LLM code completion, AI chat, and an agentic coding assistant backed by llama.cpp. llama.vscode is a Visual Studio Code extension that brings LLM-assisted text completion, chat, and agentic coding into the editor. It is built around llama.cpp, so completions and chat can run against local models on your own machine, with external providers also supported for chat and agent use.

Is llama.vscode free to use? What license does it have?

llama.vscode is open source under the MIT license. Check the license terms before using it in commercial products.

Is llama.vscode still maintained?

Actively maintained. The last commit to llama.vscode was on Nov 5, 2026.

What can I use llama.vscode for?

Getting private, offline code completion from a local model while editing in VS Code Chatting with a local or external model about code without leaving the editor Running an agent in the editor that uses MCP tools already set up in VS Code Trying different completion or chat models on consumer hardware with low VRAM settings

How do I install llama.vscode?

Install the llama-vscode extension from the VS Code marketplace Install llama.cpp on macOS with Homebrew: brew install llama.cpp Start the llama.cpp server with a default FIM model (suited to 16GB+ VRAM): llama serve --fim-qwen-7b-default

Where can I get llama.vscode?

The source code and installation instructions are on GitHub: https://github.com/ggml-org/llama.vscode