What is Mcp Frontend Starter

mcp-frontend-starter is a comprehensive toolkit and starter designed for frontend developers to integrate MCP servers with GitHub Copilot’s Agent in VS Code, enhancing their coding workflow.

Use cases

Use cases include enhancing productivity in coding by leveraging MCP servers for AI-assisted development, exploring new features in VS Code Insiders, and collaborating on projects with shared configurations.

How to use

To use mcp-frontend-starter, ensure you have VS Code Insiders (version 1.99 or later) with GitHub Copilot pre-release installed. Configure your MCP servers in the .vscode/mcp.json file, sharing settings with collaborators.

Key features

Key features include easy configuration of MCP servers, secure storage of sensitive inputs, and direct integration with GitHub Copilot’s Agent mode for an enhanced coding experience.

Where to use

mcp-frontend-starter is ideal for frontend development environments where GitHub Copilot is utilized, particularly in collaborative coding projects.

🚀 MCP Frontend Starter

A comprehensive toolkit, playground, and starter for integrating MCP^1 servers with GitHub Copilot’s Agent in VS Code. Designed specifically for frontend developers to achieve maximum “vibing”^2 in their workflow.

✨ Supercharge GitHub Copilot with MCP Server Tools

MCP support has just landed in VS Code Insiders (v1.99) as a preview feature, ready to enhance GitHub Copilot’s Agent mode (see open issues). This integration lets you leverage any MCP server directly in your agentic coding workflow.

⚠️ This is an early preview feature that may evolve. Use it for exploration and fun! Found a bug or have ideas? File issues here.

🏁 Getting Started with MCP

Prerequisite: VS Code Insiders (version 1.99 or later) with GitHub Copilot pre-release installed. (👋 Welcome to our early adopter audience!)

1. A tour of MCP Servers configurations

Already set up, but feel to edit and add:

1. Servers in .vscode/mcp.json are shared as configurations with all repo collaborators:
   • "servers": {} follows Claude’s structure for claude_desktop_config.json
   • "inputs": [] lets you define custom placeholders for configurations, avoiding hardcoded secrets

   // Example .vscode/mcp.json
   {
       // 💡 Inputs will be prompted on first server start,
       //    then stored securely by VS Code.
       "inputs": [
           {
           "type": "promptString",
           "id": "perplexity-key",
           "password": true, // Encrypted at-rest
           "description": "Perplexity API Key"
           }
       ],
       //
       "servers": {
           // https://github.com/ppl-ai/modelcontextprotocol/
           "Perplexity": {
               "command": "docker",
               "args": [
                   "run",
                   "-i",
                   "--rm",
                   "-e",
                   "PERPLEXITY_API_KEY",
                   "mcp/perplexity-ask"
               ],
               "env": {
                   // Look ma, no hardcoded secrets!
                   "PERPLEXITY_API_KEY": "${input:perplexity-key}"
               }
           }
       }
   }
   

2. Alternatively, configure "mcp": {…} in your VS Code settings for a personalized MCP server list which follows you with with Settings Sync.
3. Enable MCP discovery ("chat.mcp.discovery.enabled": true) to use your existing entries from Claude Desktop and other MCP clients. VS Code will show a trust dialog before enabling those servers:
4. Run MCP: List Servers to see all available servers, manage them, and check logs in the Output panel

2. Unleash Copilot Agent Mode with MCP Tools

1. Open Copilot view in VS Code and switch to Agent mode
2. Click the 🛠️ tools button to see enabled MCP tools (Playwright, Memory, and Perplexity for this repo)
   • Toggle servers/tools on or off as needed
3. Try these prompts that will use tools:
   • “Research with perplexity what the best MCP servers for frontend development are and create a splashy listing page”
   • “Use the playwright tool to check the landing page and fix any ayout bugs”
   • “Change the design to a dark-only mode, and do a visual review with playwright”

That’s it! You’re ready to go. But this is just the beginning—ANY DEVELOPER TOOL CAN BECOME AN MCP SERVER FOR GITHUB COPILOT. 🤯 The possibilities are endless.

🔍 Next Steps

Try these to expand your MCP capabilities:

• Add more MCP servers from the official repository
• Customize which tools are active based on your current task
• Fine-tune tool usage with .github/copilot-instructions.md
• Create your own MCP server in TypeScript, Python, or any language you prefer

🛠️ Development Setup

Easiest option: Use DevContainer in VS Code for a ready-to-go environment with all dependencies.

Local setup:

npm install
npm run dev