GitHub Copilot Copilot Agent Mode Gemini 2.5 Pro and GPT-4.1 MCP server support Copilot output window context See your Copilot usage Save or share your prompts with prompt files IDE Microsoft 365 Agents Toolkit – June
Visual Studio 2022 Release Notes | Microsoft Learn
Excerpt
Learn about the latest features, bug fixes, and support for Visual Studio 2022. Download today.
Share via
We’re excited to announce the availability of Visual Studio 2022 v17.14. This update focuses on stability and security, along with AI improvements for all developers.
Copilot Agent Mode is the next evolution in AI-assisted development and has moved out of preview.
It unlocks a new way to build with GitHub Copilot, going beyond single-turn prompts to handle entire tasks from start to finish.
To get started, open Copilot Chat, click the Ask button, and switch to Agent.
From Idea to Code
Agent Mode understands your intent, creates a plan, pulls in the right context, and takes action. It can fix bugs, edit files, and build your solution—all from a single prompt until your goal is reached.
With MCP integration, Agent Mode can also pull in rich context from your connected systems—like GitHub, CI pipelines, or monitoring dashboards—to take even smarter actions.
Try it out and see what improvements it can offer!
Gemini 2.5 Pro and GPT-4.1
New Models just dropped in Visual Studio Chat.
More power, speed, and flexibility — all within your IDE!
🚨 Gemini 2.5 Pro is now in Visual Studio
Gemini 2.5 Pro, widely regarded as one of the top-performing models available, is now supported in Visual Studio. It joins GPT-4.1, GPT-4o, Claude Sonnet 3.7, Claude Sonnet 3.5, and o3 mini to bring a truly versatile multi-model AI experience to your IDE.
🚨 GPT-4.1 is now the default
GPT-4.1 is now the default model in Visual Studio, offering a faster, more responsive, and noticeably smoother experience across the board.
You can now select the best model for the task, code, and workflow at hand. Try them out and let us know what you think!
MCP server support
Connect Visual Studio to AI agents using the new Model Context Protocol (MCP) — a standardized way to share context, access data, and drive intelligent features.
Visual Studio now supports MCP servers, unlocking smarter and more connected AI development. MCP is an open protocol that standardizes how apps and AI agents share context and take action.
With MCP in Visual Studio, you can do more than just retrieve information from your MCP servers, such as logs, test failures, PRs, or issues. You can also use that information to drive meaningful actions in your code, your IDE, and even your connected systems across your stack.
Set Up Your MCP Server
Add an mcp.json file to your solution, and Visual Studio will detect it automatically. It also recognizes configs from other environments like .vscode/mcp.json.
Use Your MCP Servers
Open the Tools dropdown in the Copilot Chat panel to view connected MCP servers. From there, Copilot can pull in context and take action using your existing systems.
Note: You will need to be in Agent Mode to access and interact with MCP servers.
Bring the full power of your stack into Copilot without leaving Visual Studio!
Copilot output window context
You can now reference the Output Window as context in Copilot Chat.
We know the Output Window is essential for tracking builds, debugging, and understanding what is going on under the hood. Now, Copilot can read that information – so you can ask questions, get help, and take action based on your output logs.
How to Use It
Use #output to reference it directly in chat.
Click the ➕ icon in the chat input to add it as context.
Ask Copilot directly: Check my output logs and help me fix this error.
Or right-click in the Output Window and choose Explain with Copilot to attach it instantly.
Try it out and see how much easier it is to read and act on your logs!
See your Copilot usage
You can now track your Copilot usage directly in Visual Studio.
Want to know how much you’ve used Copilot this month? Visual Studio now includes a Copilot Consumptions panel to easily track your usage!
Whether you’re using Copilot Free or a paid plan, you can check your number of chat requests and code completions right from the IDE.
Access it by clicking the Copilot badge in the top-right corner of your IDE and selecting Copilot Consumptions.
This opens a panel showing your current usage for the month.
Tip: If you’re using premium models, you’ll also see usage multipliers listed in the model picker, so you always know exactly what you’re using.
Save or share your prompts with prompt files
You can now create reusable prompt files in your repository to quickly run or share your frequently used prompts.
Writing effective prompts can be time consuming, but it’s the key to getting the most out of Copilot Ask and Agent mode. We’ve enabled reusable prompt files to allow you to leverage these great prompts again and again. From there, you can run quickly from the chat window or share with your team.
How to Write Prompt Files
Write your prompts the way you normally would in the prompt box, attaching # references to add specific information from your solution like methods, classes, and files. You can even reference other prompt files to include common base instructions.
Then, copy it from the prompt box into a markdown file, and save it with the .prompt.md file extension in the root of your repository in the .github/prompts folder
Tip: You can even get Copilot’s help in refining your prompt to be more specific
How to Run Your Prompts It
Start typing #prompt: to reference all the prompt files from your .github/prompts folder directly in chat or click the ➕ icon in the chat input to add it as context.
Add any additional information in the prompt box or just send the prompt!
Give it a try and speed up your AI workflows!
📣 Take this survey to help make the feature even better.
IDE
Microsoft 365 Agents Toolkit – June
Agents Toolkit 17.14 GA June release updates.
The June release contains several fixes for 17.14 GA release:
Enable Custom Engine Agent launch and preview in Microsoft 365 Copilot
There is a Custom Engine Agent template available in our toolkit which is called Weather Agent. We have updated this template to enable it to be launched and previewed in the Microsoft 365 Copilot chat. To try out, simply select the Copilot launch profile when you debug.
Fixed an error pop up when debug failed with installation of Microsoft 365 Agents playground. Now toolkit will pop up clear message and instructions on how to fix the issue.
Several bugs in templates README file introduced by our new branding changes were fixed.
Updated the toolkit icon in extension manager.
Fixed the issue when launching Agents Playground, the required channel ID value only allows msteams and emulator. Now it allows other values like webchat.
Important
A breaking change has been made to Domain-Specific Language model serialization. Developers using the default serialization mechanism for their models that contain custom property types will need to add TypeConverters for those custom property types. For an example of how to do this, see Customizing File Storage and XML Serialization.
Gemini 2.5 Pro は、利用可能な最もパフォーマンスの高いモデルの 1 つと広く認識され、Visual Studio でサポートされるようになりました。 GPT-4.1、GPT-4o、Claude Sonnet 3.7、Claude Sonnet 3.5、o3 mini に参加して、真に汎用性の高いマルチモデル AI エクスペリエンスを IDE に提供します。
🚨 GPT-4.1 が、デフォルトになりました
GPT-4.1 が Visual Studio のデフォルト モデルになり、より高速で応答性が高く、ボード全体で非常にスムーズなエクスペリエンスが提供されるようになりました。
コメントを残す