September 25, 2026
Learn MCP in VS Code through a ready-made example: Playwright in a VS Code Learn lesson
The VS Code Learn lesson from Sep 16 introduces MCP through Playwright. It covers AI agents with tools, permissions, sandboxing, and connecting external data through MCP.
Visual Studio Code
@code
🚀 From the VS Code Learn library: extending agents! In this end-to-end lesson, we will cover the key parts of AI agents: tools, toolsets, permissions, and sandboxing. We will also look at MCP (Model Context Protocol) and use it to connect agents to external data sources. Watch ▶️ https://youtu.be/SN3JTq3Q9-M
· 8.2K views
Previously, MCP in VS Code was configured manually through a configuration file. Now the lesson walks through Extensions: search for `@mcp playwright`, install the server, confirm trust, then select Agent in Chat, enable Playwright tools, and give it a browser task.
The file remains. For a project, VS Code reads `.vscode/mcp.json`, while a portable `.mcp.json` can be placed in the project root. Without the interface, the `code --add-mcp` command adds a server using a JSON configuration.
In VS Code 1.112, the sandbox for local stdio MCP servers requests access to folders and domains when needed.
Source
