News
Custom Agents Lead Microsoft AI Roadmap for Visual Studio
Microsoft detailed its November roadmap for AI in Visual Studio, emphasizing continued progress toward fully integrated, agentic experiences within the IDE.
In a Nov. 5 announcement, product manager Rhea Patel outlines active development areas across new agents, Agent Mode and Chat improvements, Model Context Protocol (MCP) expansion, and model management. Microsoft notes these items represent ongoing work rather than guaranteed deliveries for the month.
With new agents, Microsoft is refining how developers discover and switch between agents and modes while enabling support for more complex workflows. In progress are user-created custom agents, a Test Agent, and a Debugger agent, along with an investigation into running multiple Visual Studio agents concurrently. These efforts are intended to streamline multi-agent collaboration across coding, debugging, and testing tasks.
For Agent Mode and Chat, the roadmap highlights several user-driven improvements, including slash commands for invoking prompts and managing chats, enhanced memories in chat, and new capabilities for dynamic tool calling and summarizing thread history to maintain context within longer interactions. Microsoft's goal is to make AI-driven conversations more efficient and persistent across development sessions.
The Planning section lists updates that make planning read-only in chat and improve the user experience with inline previews. These enhancements are designed to make agents' intent clearer before execution, allowing developers to preview upcoming actions directly within chat threads.
Advancements to the Model Context Protocol aim to bring the entire development stack into Visual Studio with enterprise-grade security and governance. Ongoing work includes MCP sampling window UX improvements, token and performance optimizations for MCP servers, governance enablement within Visual Studio, and a unified MCP UX experience. These updates move Visual Studio closer to full MCP specification compliance while improving scalability and policy control.
In the models area, Microsoft continues to expand model options for chat. Additions include automatic model selection, integration of GPT-5 Codex, improved handling of models nearing deprecation, and different system prompts per model. These features are designed to ensure that developers can leverage the most effective model for each workflow without losing continuity or customization.
Patel concludes the post by reiterating Microsoft's goal of making Copilot available at every stage of development, from searching and error correction to unit testing and committing code. Feedback is invited through linked Developer Community tickets.
About the Author
David Ramel is an editor and writer at Converge 360.