Software Engineering :: Pair Programming :: AI Assistant :: GitHub Copilot :: About :: VS Code Agent Mode Just Changed Everything

readme.md

Software Engineering :: Pair Programming :: AI Assistant :: GitHub Copilot :: About :: VS Code Agent Mode Just Changed Everything

⪼ Made with 💜 by Polyglot.

• ..

VS Code Agent Mode Just Changed Everything

Agent Mode

Summary

This video is a practical tutorial demonstrating the latest AI-powered features in Visual Studio Code, including Agent Mode, next-edit suggestions, Model Context Protocol (MCP) servers, and bring-your-own-key support for models. The host walks through setting up these new tools, using agent mode to install Tailwind CSS, configuring an MCP server to enable the agent to interact with a PostgreSQL database, and finally having the agent build a complete web application (a URL list app) from a Product Requirements Document (PRD). The tutorial highlights both the power and current limitations of agentic AI development inside VS Code, ultimately encouraging viewers to explore these new tools for their own projects.

Highlights

Getting Started with Agent Mode

• Agent mode now available in VS Code alongside ask and edit modes
• Users can enable agent mode through settings even if it hasn't fully rolled out
• Agent mode automates more complex developer tasks, simulating a real developer

Using Tools and Fetching Context

• Agent, ask, and edit modes differ mainly in their level of autonomy and file/system permissions
• AI models have training cutoffs and may lack up-to-date context, making it crucial to provide links to current docs
• The fetch tool lets the agent read web documentation and follow instructions precisely

Building an App with AI Agent

• The agent can follow a PRD to build an actual application from scratch
• Copilot can be primed with project best practices and custom instructions
• The database schema can be injected into the agent’s context, but only if the agent can “see” it

Model Context Protocol (MCP) Servers

• MCP servers bridge the gap between VS Code and local services (like Postgres)
• Users install an MCP server (as an npm package, Docker container, or Python package)
• The server lets the agent query the live database schema and operate with full context

End-to-End Demo: App Generation

• The agent receives the PRD and proceeds to build the URL list app
• Real-time code edits, database queries, and UI changes are performed by the agent
• The app supports features like adding, reordering, and persisting links

Next-Edit Suggestions & Bring Your Own Key

• VS Code now suggests follow-up edits when you change code, speeding up iterative development
• Users can connect their own API keys for external or local LLMs (e.g., Gemini, Ollama)

Final Takeaways

• The new AI and agentic features make VS Code a more powerful, autonomous development environment
• Developers are encouraged to experiment with agent mode and MCP for their own workflows
• The host closes by inviting viewers to imagine what they will build next with these new tools

---