VS Code: Use Local AI Models & Offline LLMs Now

Visual Studio Code now offers a new feature that allows users to integrate their own AI models for development. This means you can use models and services beyond GitHub Copilot. Specifically, Visual Studio Code 1.122 introduced "Use BYOK [Bring Your Own Key] without a GitHub sign-in." This feature enables the use of chat, tools, and Model Context Protocol servers in restricted environments where GitHub sign-in isn't possible. Crucially, it supports fully offline workflows with local models like Ollama. This means you can now use locally hosted LLMs for chat and tools directly within VS Code. However, using a local LLM for inline and next-edit suggestions still requires additional tooling. To use this "BYOK" mode, you need a way to host your model, as VS Code doesn't have its own hosting mechanism. Products like LM Studio can help manage LLMs on your own hardware. The model host doesn't even need to be on the same system as VS Code. When choosing a model, consider its size and suitability for coding. Models like Gemma4, Qwen3.5 9B, and Codestral 22B v.0.1 can fit comfortably into 8GB VRAM. To set this up, ensure VS Code's AI features are enabled. You can then manage third-party.