Agent mode is now generally available with MCP tools support in Visual Studio
Copilot agent mode is on by default in Visual Studio. Agent mode helps you accomplish end-to-end development tasks by planning, taking action, and iterating until your goal is complete. Unlike traditional chat prompts, agent mode can reason across multiple steps, edit your code across files, fix bugs, and even respond to errors till it achieves your goal. And it does this all from a single prompt.
To get started, open the Copilot Chat panel in Visual Studio, click the Ask button, and switch to Agent.
Connect Visual Studio to your stack with MCP server support (Preview)
Visual Studio now supports Model Context Protocol (MCP) servers (Preview), enabling smarter and more connected AI development. MCP is an open protocol that allows Copilot to access context and data from your development environment, including logs, test results, pull requests, and issues. Copilot can use this data to take meaningful action in your code, your IDE, and across your stack.
To get started, add an mcp.json file to your solution. Visual Studio will detect it automatically, including configurations from environments like .vscode/mcp.json.
You can view your connected MCP servers in the Copilot Chat Tools dropdown. Realize that you must be in agent mode to use MCP integrations.
New models available in Visual Studio: Gemini 2.5 Pro and GPT-4.1
Visual Studio now supports Gemini 2.5 Pro and GPT-4.1, giving you more flexibility and power when using GitHub Copilot.
GPT-4.1 is now the default model in Visual Studio, providing faster, smoother responses across the board.
Track your GitHub Copilot usage directly in Visual Studio
You can now track how much you’re using Copilot each month with a new “Consumption” panel in Visual Studio. Whether you’re on a free or paid plan, you can view the number of chat requests and code completions you’ve used so far directly in the IDE.
Click the Copilot badge in the upper-right corner of Visual Studio and select Copilot Consumption to access your usage dashboard.
If you’re using premium models, you’ll also see usage multipliers in the model picker for added transparency.
Copilot can now use your Output Window as context
The “Output Window” is critical for debugging, tracking builds, and understanding what’s happening in your project, and now Copilot can use that information too.
With this update, you can reference the “Output Window” using `#output` in chat, click the plus icon in chat input to add output context, and right-click in the “Output Window” and choose Explain with Copilot.
Improvements to the code completion experience
- Next edit suggestions (NES) are now in Visual Studio. NES analyzes your recent code edits, predicts what you’re likely to do next, and suggests changes anywhere in your file, guiding you to the next likely edit location with jump and accept actions.
- Generate doc comments for C++ and C# with AI. Just start typing a comment in the format used by your project (such as `///`), and Copilot will complete the rest based on the function signature and content.
To see what else is new, check out the Visual Studio release notes.
Join the discussion within GitHub Community.