Rubber Duck Thursday - Hacking with MCP Servers + a surprise! 👀

Key Concepts

• GitHub MCP Registry: A centralized location for discovering and installing AI tools and MCP servers.
• MCP (Model Context Protocol): A protocol that allows LLMs to interact with external tools and services.
• GitHub Copilot: An AI pair programmer that assists developers with code completion, code generation, and code review.
• High Velocity Engineering: A disciplined approach to leveraging AI throughout the entire SDLC, focusing on providing strong context to GitHub Copilot.
• Playwright MCP Server: An MCP server that allows Copilot to interact with and test web applications using Playwright.
• Auto Model Selection: A feature in GitHub Copilot that automatically selects the best model for a given task.
• Advocate Hub: A content hub application built by the speaker to track developer advocacy content.

GitHub Updates and Announcements

• Copilot Code Review: Now available in JetBrains IDE and Visual Studio, allowing developers to get an initial review of their code before submitting it to their team.
• Anthropic Cloud Opus 4.1: Available in public preview in Visual Studio, JetBrains IDEs, Xcode, and Eclipse, allowing users to leverage Anthropic's flagship thinking model in GitHub Copilot.
• GitHub MCP Registry Launch: The official GitHub MCP registry is now available, providing a centralized location for discovering and installing AI tools and MCP servers. The registry is in public preview and currently features a curated set of MCP servers. The link to access the registry is gh.io/mcpregistry or github.com/mcp.
• Auto Model Selection for Copilot in VS Code: GitHub Copilot in VS Code now includes an "auto" option that picks the model for you, optimizing for model availability and routing to GPT-4, 4 mini, 4.1 sonnet, and sonnet 3.5. It respects user model settings and is available for Copilot free, pro, pro plus, and coming soon to Copilot Enterprise and Copilot Business.
• Verified Answers Generally Available in GitHub Discussions: A feature that allows users to verify answers in GitHub discussions.

GitHub MCP Registry Deep Dive

• The GitHub MCP registry aims to solve the problem of finding secure, valid, and working MCP servers, which are often scattered across community forums and random repositories.
• The registry provides a one-stop shop for connecting with MCP servers.
• Users can install MCP servers with a single click from the GitHub UI.
• It is crucial to read the documentation for each MCP server, as it may require additional steps such as obtaining an API key, running Docker, or starting a dev container.
• The speaker demonstrates installing the GitHub MCP server from the registry and using it to create an issue in a repository.
• The process to get other MCP servers added to the GitHub registry involves contributing to the official open-source MCP registry.
• The GitHub team is currently pulling in servers from this community-driven registry service and will likely have standards and guidelines for inclusion in the official GitHub MCP registry.

Auto Model Selection Demonstration

• The speaker attempts to demonstrate the auto model selection feature in GitHub Copilot but is unable to find the option in their VS Code.
• They plan to create a video demonstrating the feature and post it on LinkedIn.

Advocate Hub Project Update

• The speaker provides an update on the Advocate Hub project, a content hub application for developer advocates.
• The application allows users to paste a link to a content piece and have it automatically unfurled and captured into a dashboard.
• The speaker asks for feedback on what features to add to the application.
• GitHub Copilot is used to generate feature suggestions, including analytics and performance tracking, smart organization and discovery, and workflow and productivity enhancements.
• The speaker creates issues for each category of suggestion using the GitHub MCP server.
• The speaker implements a quick win by extending the content types to include podcast, live stream, webinar, newsletter, and speaking.
• The Playwright MCP server is used to ensure that the new categories were added to the app, demonstrating visual testing capabilities.

Playwright MCP Server Demonstration

• The speaker demonstrates the Playwright MCP server by asking Copilot to ensure that the new categories were added to the Advocate Hub application.
• Copilot navigates to the application, selects the dropdown menu, and checks to see that the categories are there.
• It also takes a screenshot of the application.

High Velocity Engineering with Michael

• Michael, a guest speaker, discusses how he is using GitHub Copilot and high velocity engineering to build his startup app, Go Glow Up, an AI productivity coach.
• High velocity engineering involves working with Copilot as a pairing session with a junior developer, providing strong context through instructions loaded into memory as part of the prompt.
• Michael's team is using GitHub MCP to run their full SDLC using chat, with 95% of the assets being prompt-driven.
• They are using a two-phase prompting pattern, where Copilot first creates a plan file and then executes the plan step by step.
• Copilot is used to generate the Terraform deployment for the infrastructure, including app insights, service plan, Cosmos DB, and OpenAI.
• The app is built using React Native in the front end, Azure functions in the back end, and Cosmos DB for the database.
• The AI coach model is trained to provide personalized productivity coaching.
• Michael emphasizes the importance of adding discipline to the AI-assisted development process and guiding the system rather than being a passive viewer.

Zero to Merged Contest

• The speaker encourages viewers to submit their projects to the Zero to Merged contest, where they can win a free ticket to GitHub Universe.
• The deadline for submissions is September 19th.

Conclusion

The session highlights the latest updates from GitHub, with a focus on the new GitHub MCP Registry and the power of MCP servers in connecting LLMs to external tools. The speaker demonstrates how these tools can be used to enhance developer productivity and streamline workflows. The guest speaker, Michael, showcases how GitHub Copilot and high velocity engineering can be used to build a full-scale SaaS application. The session emphasizes the importance of providing context to LLMs and adding discipline to the AI-assisted development process.