Skip to content
Follow
← all courses

◆ COURSE · 8 lessons · 126 min read · free

GitHub Copilot in Practice: Agentic Coding for Production Engineering Teams

For software engineers who already work in production codebases: use GitHub Copilot's interaction modes, prompting and context, the cloud agent and CLI, code review with custom instructions, and MCP to plan, build, review and validate real changes.

Who it's for

Software engineers who already work in production codebases and are adopting or expanding their use of GitHub Copilot, including its agentic and autonomous features; assumes existing programming and version-control fluency but not prior Copilot experience.

You'll need

  • professional software development experience
  • familiarity with git and pull-request workflows
  • access to a GitHub account with some tier of Copilot enabled

What you'll be able to do

  • Select and justify an appropriate Copilot plan and interaction surface for a given engineering context using current, non-legacy billing concepts.
  • Use chat, edit, and agent modes appropriately across supported IDEs, correctly distinguishing GitHub's canonical Edit mode from Visual Studio's separately branded Copilot Edits feature.
  • Delegate bounded engineering tasks to the Copilot coding/cloud agent and the Copilot CLI, evaluating results against current plan eligibility and billing behavior.
  • Write effective prompts and supply relevant context to Copilot using documented composition and context-provision technique, choosing an interaction mode and validation approach suited to the task.
  • Configure and apply Copilot code review with scoped custom instructions to improve review quality on AI-generated and agent-produced changes.
  • Configure MCP servers and correctly reason about tool-call approval behavior across autonomous and interactive Copilot surfaces.
  • Recognize which Copilot billing, plan, and product-naming facts are volatile or historical, and know where to look up current or legacy detail rather than relying on memorized figures.
  • Plan, execute, review, and validate an end-to-end agentic change to a realistic codebase, integrating context, agent delegation, review, and MCP practices learned throughout the course.

Lessons

  1. Lesson 1 · 15 minGetting Started with GitHub Copilot: Core Concepts and Interaction ModesMeet Copilot's four interaction surfaces (completions, chat, edit mode and agent mode), see which IDEs support each, and set up the reference project the whole course builds on.
  2. Lesson 2 · 12 minChoosing a Copilot Plan: Core AI Credits Billing ConceptsHow Copilot's usage-based AI Credits billing works, what the current plans include and what consumes credits, so you can choose a plan for the reference project.
  3. Lesson 3 · 11 minThe Copilot Cloud Agent and CLIDelegate work to Copilot's cloud agent, which runs in GitHub Actions and opens a pull request, and to the Copilot CLI: how each differs from local agent mode, who can use the cloud agent and how it is billed.
  4. Lesson 4 · 23 minReviewing AI-Generated Code: Copilot Code Review and Custom InstructionsSet up Copilot code review, see which customization files it reads, and scope custom instructions so AI-generated pull requests are reviewed against your team's standards.
  5. Lesson 5 · 18 minExtending Copilot with the Model Context Protocol (MCP)Connect Copilot to MCP servers: the client-server model, the GitHub MCP registry, the risks of tool confusion and prompt injection, and why tool-call approval depends on the surface you use.
  6. Lesson 6 · 17 minPrompting and Context Engineering: Writing Effective Instructions and Supplying Context to CopilotWrite clearer Copilot Chat prompts and supply the right context: GitHub's prompt-composition techniques, per-IDE context syntax, choosing a mode, validating suggestions, and the four context-supply mechanisms.
  7. Lesson 7 · 10 minReference: Legacy Billing Mechanics, Credit Rates, and Plan NuancesAn optional reference for volatile billing detail: the retired premium-request system, the October 2025 plan and pricing snapshot, per-model credit rates, and Free and Student plan nuances.
  8. Lesson 8 · 20 minCapstone: Supervising and Validating an End-to-End Agentic WorkflowRun one realistic change through the whole workflow: assemble context, implement with agent mode or the cloud agent using MCP, review it with scoped custom instructions, and validate it with real engineering checks.