The life agentic

GitHub Copilot CLI

GitHub Copilot CLI lets you use GitHub’s agentic AI tools directly from the terminal. The command is copilot.

For students, this is often the easiest next step when GitHub Education and current Copilot availability line up: you can use GitHub’s AI tooling from the command line instead of only through a web page or editor.

If GitHub student access is unavailable or you want a different route to model access, see Model access and Claude Code.

Intended learning outcomes covered on this page

After working through this page, students should be better able to:

choose an appropriate tool for a task by distinguishing between coding-agent workflows and scripting or pipeline workflows
explain why repository-grounded terminal agents support different kinds of work than ordinary browser chat
use a coding agent in a way that separates planning from editing
recognise context-window limits and use strategies such as fresh sessions or compaction when context quality degrades
distinguish between persistent instructions and reusable memory or context
critically verify model output against files, tools, and project context before relying on it

What it is good for

GitHub Copilot CLI can help you:

ask questions about a codebase from the terminal
make small changes to files in a repository
explain code or shell commands
help with git and GitHub tasks
work interactively in a project directory

It is a good starting point for agentic workflows because it can inspect files, propose edits, and ask for permission before using tools that change the repository.

Before you install it

You should already have:

a personal GitHub account
active Copilot access, whether that comes through GitHub Education or another Copilot subscription
a terminal you are comfortable using

The official installation page is here:

Installing GitHub Copilot CLI

Installation

GitHub documents several installation methods. Common choices are:

Homebrew on macOS or Linux

brew install copilot-cli

npm on any platform

npm install -g @github/copilot

This installation path currently requires Node.js 22 or newer.

Install script on macOS or Linux

curl -fsSL https://gh.io/copilot-install | bash

If you are unsure, use whichever package manager you already use for the rest of your tools.

First start

Start the CLI with:

copilot

When you launch it, GitHub Copilot CLI will typically ask you to do two things:

confirm that you trust the current directory
log in with your GitHub account if you are not already authenticated

That trust prompt matters. Only start copilot inside a directory that you understand and trust.

A simple first workflow

Open a repository or a practice directory and try prompts like these:

Explain the structure of this project.

Show me which files I should read first to understand this codebase.

Find the main entry point and explain how the program starts.

Suggest a small improvement to the README.

These are good first prompts because they help you learn the tool without giving it a risky task.

How approvals work

When Copilot wants to use a tool that may change files or run commands, it asks for approval.

That is an important safety feature. Read the request before approving it.

As a beginner, a good rule is:

approve read-like actions freely only when you understand the context
be more careful with actions that write files, run package managers, or invoke git commands
avoid blanket approval until you are comfortable with the tool

Good habits

run copilot from a project directory, not from your home directory
start with explanation and exploration tasks before asking it to edit code
read any proposed command before approving it
use version control so you can inspect what changed
treat the output as useful but fallible

Useful features to learn early

The official docs describe a few features that are especially relevant for students:

interactive mode by starting copilot
plan mode for discussing an implementation before code is written
@path/to/file to include a file in the prompt
/compact and /context for context management
custom instructions from repository files such as AGENTS.md

Notice the pattern

If you compare Copilot CLI with the other coding-agent tools in this guide, a few recurring ideas start to appear:

Planning before editing: Copilot exposes this directly as plan mode.
Limited context: /context, /compact, and automatic compaction all exist because the session has a limited context window.
Work isolation: Copilot CLI puts less visible emphasis on subagents in the beginner workflow than Claude Code or OpenCode. That contrast matters too.
Instructions and memory: repository instructions such as AGENTS.md guide behaviour, while Copilot Memory is a separate layer for reusable context.

After you have seen the three main coding-agent guides, the page on Agentic concepts reconnects these recurring ideas.

Short version

Make sure you have active Copilot access, either through GitHub Education or another Copilot subscription.
Install GitHub Copilot CLI.
Run copilot inside a repository you trust.
Start with exploratory prompts.
Approve tool use carefully.

Next step

Official links

This site is open source. Improve this page.