Kenhuang Academy
  • English

      • Introduction

      What is Copilot?

      GitHub Copilot is GitHub / Microsoft’s AI coding assistant. Officially it helps you write code faster so you can spend energy on problem-solving and collaboration—not merely “guess the next line.”

      Authoritative entry points: docs.github.com/en/copilot and the product page github.com/features/copilot.

      You can use Copilot to:

      • Get inline suggestions (gray ghost text) as you type in an IDE
      • Chat to explain code, write tests, and fix errors
      • Ask or edit from the terminal with Copilot CLI
      • Ask about a repository, PR, or Issue on github.com
      • Hand a task to Copilot cloud agent (coding agent): research the repo, change code, and open a pull request for you to review

      Where it shows up

      SurfaceWhat you are doing
      IDE extensionsVS Code, Visual Studio, JetBrains, Vim/Neovim, Xcode, Eclipse, Azure Data Studio
      github.comCopilot panel next to files, Issues, PRs; Agents page
      GitHub MobileChat on a phone
      Copilot CLIInteractive terminal session or a one-shot -p prompt
      GitHub Copilot appDesktop agent workbench (parallel sessions, PR lifecycle)
      Windows Terminal CanaryTerminal Chat (experimental, Canary only)

      This course follows VS Code + GitHub sign-in. JetBrains, the CLI, and github.com get their own chapters.

      flowchart LR
  You[Your goal] --> IDE[IDE complete / Chat]
  You --> Web[github.com]
  You --> CLI[Copilot CLI]
  IDE --> GH[GitHub repo]
  Web --> Agent[Cloud agent]
  Agent --> PR[Pull Request]
  CLI --> GH

      Three capabilities (mental model)

      CapabilityShapeWhat you do
      CompletionsGray suggestion at the cursorTab to accept, Esc to dismiss, keep typing to ignore
      ChatSide pane / inline chatSupply context (selection, @workspace, #file), then review the patch
      Agent / coding agentIDE Agent mode, or the cloud agent on GitHubDescribe the task; a human reviews the diff / PR—do not merge blindly

      Completions fit “I know what to write; type faster.” Chat fits “explain this, add tests.” The cloud agent fits “this can finish asynchronously and land as a PR.”

      Choosing a plan (no prices here)

      GitHub sells personal and organization plans. The familiar family names are Copilot Free, Copilot Pro, Copilot Business, and Copilot Enterprise. The official plans page may also list Student, Pro+, Max, and usage is metered in GitHub AI Credits.

      Do not memorize numbers from this course. Prices, monthly completion caps, credits, and which features Free omits always follow the live pages:

      Rules of thumb:

      • Personal trial → start with Copilot Free (limited features and usage)
      • Heavy individual use → see whether Pro or a higher individual plan is available to you
      • Team policy, audit logs, content exclusion → Business
      • Enterprise governance, larger allowances, GitHub Enterprise Cloud capabilities → Enterprise

      Students, teachers, and maintainers of popular open source may qualify for free upgrades—follow GitHub’s “Access Copilot …” docs, not third-party activation guides. If an org member cannot see a feature, an admin policy is usually off, not a missing extension. New purchases of some paid plans are sometimes paused; trust the billing page.

      Versus Cursor

      This site has a full Cursor course. One line: Copilot = stay in the VS Code / GitHub ecosystem; Cursor = an AI-native IDE.

      DimensionCopilotCursor
      ShapeHost editor + extension / GitHub cloud agentStandalone editor + sidepane Agent
      Strongest atIssue → PR, reviews on github.com, coexistence with existing VS Code extensionsLocal interactive edit loop, Plan Mode, rule files
      Migration costNear zero if you already use VS CodeFirst launch can import VS Code settings

      The full table is Copilot vs Cursor. Locked into VS Code and GitHub PR flow → learn Copilot first. Want planning, multi-file edits, and checkpoints in one window → read the Cursor course.

      Versus Claude Code

      This site also has Claude Code (Anthropic’s terminal agent, command claude).

      DimensionCopilotClaude Code
      Primary doorIDE extension + github.com + copilot CLITerminal claude
      GitHub-nativeStrong: Issue assignment, PR review, cloud agentVia gh, MCP, or your own wiring
      Best whenTeams already collaborating on GitHubLong autonomous terminal work, hooks, permission model

      All three can coexist: complete locally in Cursor or VS Code + Copilot; assign a well-specified Issue to Copilot cloud agent; send a long script to Claude Code. Write norms as files in the repo so the three stacks do not invent conflicting oral rules.

      A good fit, and when to be careful

      A good fit:

      • Speeding up everyday functions, boilerplate, and tests
      • Adding AI inside familiar VS Code / JetBrains without switching editors
      • Repos already on GitHub: PR summaries, reviews, assigning Issues to an agent

      Be careful:

      • Secret code: confirm org data policy, content exclusion, and whether the cloud agent may touch the repo
      • Free allowances run out and features drop—read the official usage docs first
      • PRs from an agent must be reviewed like any other contribution; GitHub says so too

      Next

      评论