GitHub Copilot CLI: 터미널의 AI 코딩 파트너

DEV Community · Bruce Wong · 2026-06-11 개발(SW)

1. What is Copilot CLI?

In February 2026, GitHub officially released Copilot CLI to General Availability (GA). It is not a command-line port of the VS Code plugin, but a standalone, terminal-native AI coding agent.

GitHub’s positioning is clear:

“Copilot CLI has grown from a terminal assistant into a full agentic development environment—one that plans, builds, reviews, and remembers across sessions, all without leaving the terminal.”

All Copilot subscribers can use it (Free/Pro/Business/Enterprise) at no extra cost.

2. How is it different from VS Code Copilot?

This is the most confusing part. Many people think CLI is just “Copilot from VS Code moved to the terminal.” It is not.

2.1 Positioning differences

Dimension VS Code Copilot Copilot CLI Runtime Inside the IDE Terminal / command line Interaction Inline editor completions, side Chat Full-screen terminal chat Core strength Real-time coding assistance Complex task planning & execution Context switching Within the IDE No need to leave the terminal

2.2 Feature comparison

Feature VS Code Copilot Copilot CLI Code completion ✅ Real-time inline ❌ Not the main focus Agent mode ✅ Yes ✅ Yes, and more powerful Multi-step complex tasks ✅ Supported ✅ Better suited Git operations ⚠️ Limited ✅ Native integration GitHub.com operations (PR/Issue) ⚠️ Requires browser switch ✅ Direct operations Plan Mode (plan first, then execute) ✅ Yes ✅ Shift+Tab, more complete Cross-session memory ✅ Yes ✅ Stronger Repository Memory CI/CD integration ❌ Difficult ✅ -p programmatic calls Shell command execution ⚠️ Indirect ✅ Native support

2.3 One-sentence summary

VS Code Copilot is “your co-pilot while coding” — real-time completions, quick explanations, single-file refactoring.

Copilot CLI is “a full-stack engineer that works independently” — you can say “refactor this entire module,” go do something else, and it reports back when done.

3. When to use CLI vs VS Code?

Scenarios for VS Code Copilot

Daily coding with real-time completions
Quick refactoring within a single file
Explaining selected code functionality
Reviewing and confirming changes directly in the editor

Scenarios for Copilot CLI

Complex multi-step tasks: creating complete features, cross-file changes
DevOps / automation: generating CI/CD configs, Terraform scripts
GitHub operations: bulk PR management, code reviews, issue creation
Background tasks: long-running code analysis, security scans
Pure terminal environments: SSH remote servers, Docker containers, WSL
Scripted integration: embedding AI capabilities into automation workflows

4. Copilot CLI’s three killer features

4.1 Cloud Delegation (&)

The most unique feature. Prefix your prompt with & to push the task to the GitHub Cloud Agent:

Runs remotely on GitHub Actions runners
Automatically creates a copilot/ branch and Draft PR
Continues even if you shut down your local machine
Perfect for long-running tasks (code analysis, large-scale refactoring)

In contrast, Autopilot Mode (–autopilot) runs locally, is visible in real time, but stops if your local machine shuts down.

4.2 Programmatic Mode (-p)

Ideal for integration into scripts and automation workflows.

# Single-shot call, great for script integration
copilot -p "Show me this week's commits and summarize them" --allow-tool='shell(git)'

# Pipe input
./script-outputting-options.sh | copilot

Enter fullscreen mode Exit fullscreen mode

Value: Can be integrated into CI/CD, cron jobs, and shell scripts.

4.3 Plan Mode — plan first, then execute

Press Shift+Tab to enter Plan Mode. Copilot will:

Analyze your request
Ask clarifying questions
Build a structured implementation plan
Wait for your confirmation before coding

Value: Avoids the “AI blindly changes code” problem by communicating requirements first.

5. Installation and quick start

5.1 Installation

这是提取出来的 Markdown 表格内容：

Platform Command macOS brew install copilot-cli Linux brew install copilot-cli or curl -fsSL https://gh.io/copilot-install | bash Windows winget install GitHub.Copilot npm npm install -g @github/copilot

需要我帮你保存成 .md 文件吗？如果要的话，文件名和路径有偏好吗？
Article content

5.2 Authentication

copilot
/login 
## 5.3 Quick start

Enter fullscreen mode Exit fullscreen mode

# 1. Enter your project directory
cd my-project

# 2. Start interactive mode
copilot

# 3. Initialize custom instructions (helps Copilot understand your stack)
/init
# Generates .github/copilot-instructions.md with project context

# 4. Start building
"Create a React component for user authentication"

# 5. Run in the background (non-blocking terminal)
& "Analyze all security vulnerabilities in this codebase"

Enter fullscreen mode Exit fullscreen mode