Extend GitHub Copilot coding agent with custom MCP tools

Key Concepts:

• GitHub Issues creation from Copilot Chat
• Copilot Coding Agent
• Product Requirements Document
• Notion integration
• Model Context Protocol (MCP)
• Pull Request creation and review
• GitHub Actions

Creating GitHub Issues from Copilot Chat

The video demonstrates the ability to create GitHub issues directly from Copilot Chat within github.com. The example used is the fictional "Copilot Airways" product team, where a product requirements document for user reviews on the travel guide is stored in Notion. Instead of manually creating the issue, the presenter shows how to ask Copilot to create a GitHub issue, including a link to the Notion document for traceability.

Copilot Coding Agent and Issue Assignment

The Copilot coding agent can be assigned issues directly from the Copilot Chat interface. The video shows Copilot acknowledging the assignment and initiating work by creating a pull request.

Model Context Protocol (MCP) for External Tool Access

The core of the demonstration revolves around the Model Context Protocol (MCP). MCP allows Copilot to access external tools, such as Notion, beyond its native capabilities (reading files, code edits, running builds, and tests). The video highlights that Copilot uses multiple "tool calls" to gather information from Notion, which is then incorporated into its workflow.

MCP Configuration in GitHub Actions

The video shows that the MCP servers and associated tools are configured within GitHub Actions during the Copilot setup process. This configuration enables Copilot to interact with external services like Notion.

Pull Request Review and Summarization

After Copilot completes its work, the generated code is available in a pull request. Given the potential size of the changes, the presenter highlights Copilot's ability to summarize the changes it has made within the pull request. This summarization provides reviewers with the necessary context for efficient code review.

Notable Quotes:

• "[Let’s ask Copilot to] create a GitHub issue with that information and a link to the Notion document, so they have traceability of the request."
• "Aside from native tools like reading files, making code edits, running builds, and tests, you can provide Copilot access to external tools using MCP, just like we are doing here with Notion."

Technical Terms:

• GitHub Issues: A feature in GitHub for tracking tasks, bugs, and feature requests.
• Copilot Chat: An interface within GitHub that allows users to interact with the Copilot AI assistant.
• Copilot Coding Agent: The AI agent within Copilot that can perform coding tasks, such as creating pull requests.
• Product Requirements Document: A document that outlines the requirements for a new product or feature.
• Notion: A popular workspace and note-taking application.
• Model Context Protocol (MCP): A protocol that allows Copilot to access external tools and services.
• Pull Request: A request to merge code changes from one branch into another.
• GitHub Actions: A continuous integration and continuous delivery (CI/CD) platform within GitHub.

Logical Connections:

The video logically connects the process of issue creation to the subsequent code generation and review. The MCP is presented as the crucial link that enables Copilot to access the necessary information from external sources (Notion) to complete its tasks effectively. The pull request summarization feature is presented as a way to streamline the code review process for changes made by Copilot.

Synthesis/Conclusion:

The video demonstrates how Copilot can streamline the software development workflow by automating tasks such as issue creation and code generation. The Model Context Protocol (MCP) is a key enabler, allowing Copilot to access external tools and data sources. The pull request summarization feature further enhances efficiency by providing reviewers with the necessary context to understand Copilot's changes. The main takeaway is that Copilot, with the help of MCP, can significantly improve developer productivity and collaboration.