Microsoft has rolled out a new update for its popular code editor, Visual Studio Code, bringing some handy new features and security improvements. The update, version 1.104, was released on September 11 and is available for Windows, Linux, and Mac. It focuses on making AI model choices easier and safer for developers.

Automatic Model Selection Makes AI Coding Smarter

One of the biggest new features is the ability for VS Code to automatically pick the best AI model for chat interactions. When developers set the model picker to “Auto,” VS Code will choose the optimal model to keep things running smoothly and avoid hitting rate limits. This feature will soon be available to all GitHub Copilot users inside VS Code.

The automatic model selection can pick from several models, including Claude Sonnet 4, GPT-5, GPT-5 mini, GPT-4.1, and Gemini Pro 2.5. If an organization has disabled access to some models, the picker will respect those settings. This makes working with AI more seamless and less manual.

The update also introduces the final version of the LanguageModelChatProviders API. This API allows extensions to add their own models, whether hosted locally or in the cloud. Users can then choose these models easily from the chat interface, adding more flexibility to AI-powered coding.

Enhanced Security for Autonomous Agents

Security is a big focus in this release. When using agent mode, VS Code’s AI agents can automatically edit files in a workspace. While this is useful, it can also be risky if the agent makes unintended or malicious changes. To address this, the update now requires agents to ask for user confirmation before editing certain files, reducing the chance of accidental damage.

The release also improves security around terminal commands. The terminal auto-approve setting, which controls whether commands are run automatically, can now be turned on or off more easily. This gives developers better control over what actions are taken automatically, boosting safety while maintaining usability.

In addition, the update builds on earlier features like the autostart configuration for MCP servers. Now, if a workspace has an AGENTS.md file, VS Code will automatically recognize it as context for chat requests. This is especially helpful for teams managing multiple AI agents. When running tasks that request user input, the agent now detects this and prompts the developer in chat, keeping things interactive and clear.

Other Improvements to Boost Developer Experience

Several smaller updates make working in VS Code more efficient. The list of changed files has been improved to better support developers working in agent mode, making it easier to see what has changed. The editor now also allows developers to customize how quickly inline suggestions appear with a new setting, helping prevent suggestions from interrupting typing.

Mathematical equations in chat responses are now generally available and turned on by default, making it easier to understand complex math within AI chats. The code search feature has been upgraded with a newer embeddings model, providing more relevant results when searching through large codebases.

Terminal support has been strengthened too. Tools like terminalSelection and terminalLastCommand, which used to be extensions, are now part of the core features. This change aims to improve overall reliability and performance.

Overall, the latest VS Code update brings smarter AI features, better security, and smoother workflows. These improvements help developers work more confidently and efficiently, especially when working with AI-powered tools.

Inspired by

• https://www.infoworld.com/article/4056026/vs-code-1-104-emphasizes-ai-model-selection-agent-security.html

Sources

• gmpg.org
• visualstudio.com
• visualstudio.com
• wikipedia.org
• podigee.com