The VS Code 1.107 release introduces a unified chat view that centralizes the management of all your agents, including background, cloud, and local sessions. Users can easily view and manage multiple agents simultaneously, with the ability to expand a side panel for more workspace. This feature allows you to run multiple tasks concurrently, such as planning a project and creating unit tests, without waiting for one to finish before starting another. Additionally, you can delegate tasks to different types of agents: cloud agents operate on remote infrastructure and can interact with repositories, while background agents run locally or in isolated git work trees to avoid conflicts.
Another significant addition is the language models editor, designed to simplify the management of various AI models. Through the models picker, users can view detailed information about each model, including its name, context size, and capabilities. The interface allows you to hide or show specific models easily and filter through them to find what you need quickly. You can also add new models from installed providers, streamlining the process of working with multiple AI models within VS Code.
The inline chat experience has been enhanced to better align with other chat features in VS Code and to optimize it for quick, single-line code changes. A notable new feature is rename suggestions for TypeScript, which predict when a symbol rename should occur rather than just suggesting text edits. This feature highlights potential renames and allows users to apply them efficiently using a keyboard shortcut. Currently, this functionality is available experimentally for TypeScript, with plans to extend support to other programming languages in the future.
Improvements have also been made to the chat view’s appearance, including a new title control at the top that displays the chat’s title and provides easy navigation back to previous views. This behavior can be customized through settings, enhancing the overall user experience when interacting with chat sessions. Additionally, the release introduces experimental support for reusable cloud skills, which are markdown-based instructions that can be packaged and reused across different conversations and agents. This feature enables users to create expertise once and apply it multiple times, demonstrated with an example of testing local web applications using Playwright and Python.
Finally, the VS Code website has been upgraded with a fast client-side search feature, making it easier and quicker to navigate documentation. The release notes contain even more features and improvements, encouraging users to explore and share their favorite updates. Overall, this release focuses on enhancing productivity, collaboration, and the integration of AI-driven tools within the VS Code environment.