Use Local AI in Visual Studio with LM Studio | Ollama-Compatible Models
Published 15/09/2026
Description
Run local AI models directly in Visual Studio with LM Studio and an Ollama-compatible endpoint. Contents: 0:00 - Introduction 0:12 - Description 3:05 - Upgrade or install 7:34 - Conclusion In this video, I test the newer bring-your-own-model capabilities in Visual Studio Insiders. Alongside GitHub Copilot, Microsoft Foundry, Anthropic, and OpenAI providers, Visual Studio can now connect to local Ollama-compatible AI endpoints. That means you can use local LLMs for coding assistance without relying entirely on paid cloud subscriptions—and without moving your workflow to VS Code. I use LM Studio as the local AI server and connect it to Visual Studio through its Ollama-compatible endpoint. I then add and test local models, including OrNith 1.5 and Qwen 3.8 27B Q6, directly from the Visual Studio Copilot Chat interface. What you will learn How to install and update Visual Studio Insiders How to open and configure GitHub Copilot Chat How to add Microsoft Foundry, Anthropic, OpenAI, and Ollama-compatible model providers How to connect Visual Studio to LM Studio How to add custom local model IDs How context window and maximum-token settings affect local models How tool calling and vision support are configured Why the current Visual Studio integration expects a localhost endpoint Why a free GitHub login may still be required to use the chat interface How to test whether Visual Studio is using your selected local LLM When local AI can be a useful alternative to cloud-based coding assistants Local AI is becoming more capable, more private, and more affordable. It may not yet offer the same flexibility or performance as every cloud model, but it gives developers a real choice: use local models, cloud models, or mix both depending on performance, privacy, and budget requirements. If you are interested in Visual Studio, C#, .NET, local LLMs, LM Studio, self-hosted AI, AI workstations, Windows development, Docker, Linux, net