Rubber Duck Thursdays - Building with Agent Mode and MCP
A big announcements week — this episode unpacks Copilot Agent Mode going stable in VS Code and the launch of the official GitHub MCP server, then puts both to work building a real app from scratch.
The GitHub Changelog segment covers the VS Code Copilot March release (agent mode GA, MCP support, next edit suggestions), the GitHub MCP server launch, Copilot Pro Plus tier, Copilot Code Review GA, security campaigns GA with automated issue creation, GitHub Issues and Projects improvements (sub-issues, issue types, 50K item limit), Helm support for Dependabot version updates, and more.
Chris then explains Model Context Protocol from first principles — how MCP servers act as tool packages that give large language models capabilities beyond their built-in feature set, using a clear diagram of the user → agent → LLM → tool flow. He demos the GitHub MCP server by creating a private repository, generating a detailed GitHub issue from a screenshot using vision with GPT-4o, and then implementing the issue requirements using agent mode with Claude 3.7 Sonnet.
The app being built is a PEST analysis trend radar — a strategic planning visualization with concentric rings representing likelihood and color-coded dots for preparedness. Chris configures three MCP servers (GitHub, Playwright, and Anthropic’s sequential thinking reference implementation), showing how they compose together to give the agent richer capabilities. The session highlights the importance of reviewing agent actions, providing clear requirements, and iterating on prompts when the output doesn’t match expectations.
Related Content
Rubber Duck Thursdays - Creating a brickbreaker walkthrough
GitHub
Chris live-codes a GitHub-themed brick breaker walkthrough using Copilot agent mode, demonstrating how prompt framing, context, and custom instructions shape AI-generated output.
Rubber Duck Thursdays - Building a scoring system
GitHubChris introduces OctoSnap, a GitHub-themed memory card game built with Next.js, and dives deep into designing a scoring model with Copilot covering difficulty multipliers, time bonuses, and penalty mechanics.
Use GitHub Spark to create a podcast timer apps
GitHubDemonstrates GitHub Spark on mobile, building a podcast session timer app entirely through natural language prompts. The video covers iteratively adding timer controls (start, pause, reset), a reverse progress bar, a color-coded time-remaining indicator (green/amber/red), and AI-powered timer creation from a plain-language description—showcasing GitHub Spark's built-in LLM integration. It also demonstrates Spark's managed state storage and the share link for giving collaborators read or write access.