VS Code 1.122: A guide to using offline AI with Ollama and local modeling.

Over the past few years, AI has become an indispensable part of the modern programming environment. However, most current AI features rely on cloud services, requiring a constant internet connection and often forcing users to log in to their accounts. This creates significant limitations for businesses with high security requirements or development environments that are completely isolated from the internet.

With the latest Visual Studio Code 1.122 update, Microsoft is gradually addressing this issue. The new release not only upgrades the AI ​​agent experience but also adds the ability to use AI completely offline through locally run models, while improving web testing tools and bug reporting workflows right within VS Code.

For many developers, this may be one of the most notable changes to VS Code this year.

AI in VS Code no longer requires login.

The most notable feature of VS Code 1.122 is Microsoft's expansion of the Bring Your Own Key (BYOK) functionality. Previously, many AI features in VS Code required users to log in through GitHub or related services to activate chatbots and integrated AI tools.

Starting with the latest version, BYOK can function even when users are not logged in. This means that features such as AI Chat, calling tools, MCP server, and AI agent workflows can all operate in environments with limited or no internet connectivity.

This is a particularly important change for businesses with stringent security requirements, internal systems that are not allowed to access the internet, research environments, or projects that operate within an air-gapped environment. Instead of sending data externally, the entire AI processing process can now take place directly on the user's computer.

VS Code now supports offline AI workflows with Ollama.

One of the biggest benefits of this change is the ability to integrate VS Code with local AI models like Ollama.

In fact, many programmers are now using open-source models like Llama, Qwen, Gemma, DeepSeek, or Mistral via Ollama to run AI directly on their personal computers.

Previously, integrating these models into software development workflows often required quite complex configurations. But with VS Code 1.122, Microsoft is significantly simplifying this process.

After setting up BYOK and connecting to the local model, users can utilize AI features directly within VS Code without needing any cloud account.

In other words, you can now build a fully AI-powered programming environment that operates 100% offline.

How to enable AI Local in VS Code

Setting up the new feature is quite simple. After upgrading to VS Code 1.122, users open the Command Palette and find the Manage Language Models option .

Here, you can add the AI ​​model provider you want to use. After successfully configuring at least one BYOK model, the Chat interface in VS Code will automatically appear.

Notably, login prompts will also be removed. This allows the development environment to operate completely independently of external cloud services.

For developers using Ollama or other locally running open-source models, this is the quickest way to turn VS Code into a true offline AI IDE.

Interestingly, Microsoft is continuing to expand the role of AI agents in its software development process.

According to the company's announcement, the agents within VS Code can now activate device emulation via Playwright code.

This allows AI to automatically check for issues such as:

1. The interface is broken on mobile devices.
2. UI elements are displaying at the wrong size.
3. Require error,
4. Unusual behavior on the small screen.

Instead of developers manually testing each device, AI can help detect errors earlier in the development process. This is a rather interesting step in the trend of shifting AI from assisting in code writing to assisting in testing and ensuring software quality.

New error reporting wizard right in VS Code

Another small but very helpful change is that Microsoft has improved the bug reporting process. VS Code 1.122 adds a brand-new Issue Reporting Wizard. Instead of having to manually gather information and submit bug reports, users will be guided step-by-step to create higher-quality reports.

This wizard supports:

1. Add system information
2. Error description
3. A screenshot is attached.
4. Video recording of the error

The goal is to help the VS Code development team reproduce and fix problems faster, while reducing the number of bug reports lacking essential information.

At first glance, this might seem like just a minor update to VS Code. But in reality, fully offline AI support has far greater significance. For the past few years, the AI ​​industry has been almost entirely focused on cloud models. However, the trend towards local AI is growing rapidly. Microsoft's official integration of local AI workflows into VS Code shows that the company sees enormous potential in this trend.

Instead of forcing every programmer to rely on cloud-based AI services, VS Code now allows users to choose the model that best suits their needs, from cloud AI to local AI.

VS Code 1.122 isn't a revolutionary update in terms of interface or programming features. However, the changes below are highly significant for the developer community.

The ability to use AI without logging in, support for local modeling via Ollama, built-in device emulation tools, and a new error reporting system all aim to make VS Code a more flexible development environment.

In particular, for those concerned about privacy, data security, or wanting to build completely offline AI workflows, this could be one of the most noteworthy features Microsoft has just added to VS Code.