wtf is llama.vscode?

llama.vscode is an extension for VS Code that adds AI-assisted code completion, a chat interface, and an agentic coding assistant, all running on your own machine without sending data to any cloud service. It connects to a locally running AI server called llama.cpp, which runs open-source language models on your own hardware. As you type in VS Code, the extension automatically suggests completions. You can accept a full suggestion with Tab, accept just the first line with Shift+Tab, or accept the next word with a keyboard shortcut. There is also a manual toggle to show or hide suggestions. The extension tracks context from open and recently edited files to give better suggestions, and it is designed to stay responsive even on consumer-grade hardware. Beyond code completion, the extension includes a chat panel and an agent mode called Llama Agent. The agent can read files, browse web pages, and use tools from installed MCP servers, which are a standard for connecting AI assistants to external data sources. You can also search for and download models directly from Hugging Face through the extension interface. Installing the extension requires two things: installing llama.vscode from the VS Code marketplace, and separately setting up the llama.cpp server. On Mac and Windows, the extension can install llama.cpp automatically through a menu option. The README provides configuration guidance based on how much graphics memory your computer has. A companion plugin for Vim and Neovim called llama.vim is also available from the same organization.