Getting Started

Part 0 of the workshop. By the end of this chapter you will have Copilot CLI installed and authenticated, understand the four interaction modes, and know where its configuration lives. Budget ~45 minutes.

Prefer to watch first?

The official Ultimate GitHub Copilot CLI tutorial for beginners (GitHub) demonstrates everything in this chapter — install, /login, folder trust, and your first interactive and -p prompts — in a few minutes, then continues into the modes and slash commands you will meet next. See References → Talks & demos for what each companion video covers.

Prerequisites

Requirement	Notes
Active GitHub Copilot subscription	Required for API access. An org/enterprise admin can disable the CLI by policy (README)
Supported OS	Linux, macOS, or Windows (PowerShell v6+ and WSL) (About Copilot CLI)
A terminal & Git	You should be comfortable with both
(Optional) Node.js	Only needed for the `npm` install path

Install the Copilot CLI

Pick one of the following. All four are official (README, Installing GitHub Copilot CLI):

# Install script (macOS / Linux)
curl -fsSL https://gh.io/copilot-install | bash

# Homebrew (macOS / Linux)
brew install copilot-cli

# WinGet (Windows)
winget install GitHub.Copilot

# npm (any platform; requires Node.js)
npm install -g @github/copilot

Need to test prerelease features? Each channel has a prerelease variant: @github/copilot@prerelease, brew install copilot-cli@prerelease, or winget install GitHub.Copilot.Prerelease (README). Use prerelease builds only for validation or demos that explicitly require them.

Verify the install:

copilot --version

First launch & trust

Launch the CLI from inside a folder that contains code you want to work with — not your home directory (Security considerations):

cd ~/projects/my-repo
copilot

On first launch you will see an animated banner (re-show it any time with --banner) and a trusted-directory prompt (Using Copilot CLI):

1. Yes, proceed
2. Yes, and remember this folder for future sessions
3. No, exit (Esc)

Why the trust prompt matters

During a session, Copilot may read, modify, and execute files in and below the launch directory. Only trust locations whose contents you trust. Do not launch from your home directory or any folder with untrusted executables or secrets (Security considerations).

Authenticate

If you are not already signed in, the CLI prompts you to run /login.

Interactive: device flow (recommended)

> /login

Follow the browser flow. Your token is stored automatically and reused across sessions (README). Confirm which GitHub account is active at any time with /user (CLI command reference).

Headless / CI: Personal Access Token

For non-interactive environments, use a fine-grained PAT with the "Copilot Requests" permission (README):

Create a token at https://github.com/settings/personal-access-tokens/new.
Under Permissions, add Copilot Requests.
Expose it via an environment variable. Recent CLI builds also support COPILOT_GITHUB_TOKEN, which was added to avoid collisions with tools that already use GH_TOKEN or GITHUB_TOKEN (copilot-cli changelog 0.0.354). In CI, prefer COPILOT_GITHUB_TOKEN; if you set more than one token variable, confirm behavior with copilot help environment.

export COPILOT_GITHUB_TOKEN="github_pat_xxxxxxxx"

We use exactly this mechanism in Demo 4 · CI/CD automation.

The main interaction modes

These are the modes you need for the workshop. In interactive sessions, Shift+Tab cycles between agent modes such as ask/execute, plan, and autopilot; programmatic mode is started from the shell with -p/--prompt (About Copilot CLI; Best practices).

graph LR
    Ask["Ask / Execute<br/>(default)"] -->|Shift+Tab| Plan["Plan mode"]
    Plan -->|Shift+Tab| Auto["Autopilot<br/>(experimental)"]
    Auto -->|Shift+Tab| Ask

Mode	What it does	Best for
Ask / Execute (default)	Conversational; asks approval before each tool that modifies or runs files	Learning, sensitive changes
Plan	Analyzes the request, asks clarifying questions, writes a structured `plan.md`, and waits for your approval before coding	Complex, multi-file work
Autopilot (experimental)	Keeps working autonomously until the task is complete	Routine, well-scoped tasks
Programmatic	`copilot -p "…"` runs one prompt and exits	Scripts, CI/CD, automation

Enable experimental features (including autopilot) with --experimental or the /experimental slash command; the setting is then persisted in your config (README).

# Programmatic example — summarize this week's commits
copilot -p "Show me this week's commits and summarize them" --allow-tool='shell(git)'

Each prompt you submit consumes one premium request (README).

Configuration layers & precedence

Copilot CLI composes configuration from global (you) and repository (project) sources, with session flags on top. Understanding this hierarchy is essential before the demos.

graph TD
    S["Session — highest priority<br/>--model, --experimental, /model, flags"] --> R["Repository<br/>.github/copilot-instructions.md<br/>.github/instructions/**/*.instructions.md<br/>AGENTS.md, .github/agents, .github/skills"]
    R --> G["Global — user level<br/>~/.copilot/settings.json<br/>~/.copilot/copilot-instructions.md<br/>~/.copilot/agents, mcp-config.json"]

Key facts to internalize (Best practices, Using Copilot CLI):

Custom-instruction files now combine rather than using priority-based fallback. Repository instructions take precedence over global ones on conflict.
The global config directory is ~/.copilot/ (override with the COPILOT_HOME environment variable). User settings are stored in settings.json; MCP, LSP, agent, instruction, and session-state files live alongside it. The changelog is the best source for new workspace-level config files such as .github/mcp.json (copilot-cli changelog).
Edit settings with /settings. It opens a searchable dialog, supports inline values such as /settings autoUpdate true, and can reset a key to its default (GitHub Blog Changelog: /settings).

We dissect each layer in the Feature Deep Dive.

Verify your setup

Run this checklist inside a trusted repo. If every step works, you are ready for the demos.

# 1. Confirm version & auth
copilot --version

# 2. Inside a session, list everything available
> /help

# 3. Confirm which GitHub account is signed in
> /user

# 4. See and switch models
> /model

# 5. Inspect context usage
> /context

# 6. Confirm the GitHub MCP server is wired up
> /mcp

Discover commands live

The product changes weekly. Instead of memorizing a command list, use /help (in session) and copilot help <topic> where <topic> is one of config, commands, environment, logging, or permissions (Best practices).