Skip to content

Latest commit

 

History

History
122 lines (80 loc) · 4.43 KB

File metadata and controls

122 lines (80 loc) · 4.43 KB

Exercise 17 — Create GitHub Issues via MCP

Duration: 5 minutes
Copilot Feature: GitHub MCP Server + Prompt File
Goal: Convert the implementation plan into GitHub Issues using the GitHub MCP and the pre-built issues prompt file.



🟡 OPTIONAL EXERCISE

This exercise is not required to complete the mandatory track. It requires an active GitHub repository and a Personal Access Token with repo and issues scopes. Complete it here if your environment is configured, or revisit it independently after the workshop.

Best after: Exercise 11  |  Return to mandatory track: Exercise 12 — Build APIs with Local Agent

Background

The Model Context Protocol (MCP) lets Copilot interact with external systems. The GitHub MCP server gives Copilot the ability to create issues, pull requests, branches, and more — directly from chat.

In this exercise, you'll connect the GitHub MCP and use the github-issues.prompt.md file to automatically create work items from your implementation plan.


Part A — Install the GitHub MCP Server (One-Time Setup)

Step 1 — Install the GitHub MCP

If GitHub MCP is already configured in your workspace, skip to Part B.

Open VS Code Command Palette (Ctrl+Shift+P) and search for:

MCP: Add Server

Select GitHub MCP from the registry:

  1. In the search box dialog, search for github
  2. Select the official GitHub MCP Server from the registry
  3. When prompted, enter your GitHub Personal Access Token
  1. Confirm the installation — VS Code will configure it automatically
  2. Reload the VS Code window — GitHub MCP will start automatically

For reference, see the GitHub MCP Server repository.

  1. Create a GitHub Personal Access Token at https://github.com/settings/tokens
    • Required scopes: repo, issues, read:org
  2. Reload the VS Code window — GitHub MCP will start automatically

Part B — Create a GitHub Repository (if not already done)

If you don't have a repository yet, create one:

Using the GitHub MCP, create a new repository named "itms-app" with:
- Description: "Intelligent Task Management System — Workshop App"
- Private repository
- Initialize with a README

Note your repository as [your-username]/itms-app — you'll need this.


Part C — Generate Issues from the Implementation Plan

Step 1 — Run the Issues Prompt File

In Copilot Chat, type / and select GitHub Issues Generator (from .github/prompts/github-issues.prompt.md).

Or manually send this prompt:

Read doc/implementation-plan.md and convert it into GitHub Issues for the repository [your-username]/itms-app.

For each task in the implementation plan, create a GitHub issue with:
- A clear title
- A description explaining what needs to be done
- Acceptance criteria as checkboxes
- Appropriate labels: backend, database, testing, infrastructure, security
- Milestone matching the Phase (Phase 0, Phase 1, etc.)

Start by creating the milestones first, then create issues for Phase 0 and Phase 1 tasks.
Use the GitHub MCP to create them directly.

Replace [YOUR-USERNAME] with your actual GitHub username.


Step 2 — Verify on GitHub

Open your repository on GitHub (https://github.com/[your-username]/itms-app/issues) and verify:

  • Milestones created for each Phase
  • Issues have labels (backend, database, etc.)
  • Issues have acceptance criteria as checkboxes
  • Issues are assigned to the correct milestone

Step 3 — Create a Project Board (Optional)

Using the GitHub MCP, add all the Phase 1 issues to a new GitHub Project board 
called "ITMS Sprint 1". Set up three columns: To Do, In Progress, Done.

Key Takeaway

The MCP transforms Copilot from a code editor tool into a project management tool. Going from a Functional Requirements Document to a structured, labelled, milestoned GitHub issue backlog in under 5 minutes is the kind of productivity multiplier that changes how teams operate. The same MCP can later be used to track progress, comment on issues, and link PRs — all from chat.


Return to Mandatory Track →: Exercise 12 — Build APIs with Local Agent