Skip to content

GitHub Copilot CLI Workshop

A full-day, hands-on workshop that takes intermediate-to-advanced developers from "I have used Copilot in my editor" to "I can drive Copilot as an autonomous agent from my terminal, automate it in CI, and decide which Copilot surface to reach for in any situation."

This workshop is built entirely on GitHub's official primary sources (linked inline and collected in References). Where the product is evolving quickly, we say so and point you to the live command (for example /help or /model) rather than freezing a value that may change.

Last reviewed on 2026-06-29 against github/copilot-cli changelog version 1.0.65 (2026-06-24) and GitHub Blog Changelog entries through 2026-06-29. See References for the sources to monitor before running the workshop.


Who this is for

Aspect Detail
Audience Developers, tech leads, and platform engineers who already use GitHub Copilot in an IDE
Assumed knowledge Comfortable in a terminal, Git, GitHub pull requests, and basic CI concepts
Goal Understand the Copilot CLI deeply, compare it against the VS Code and SDK surfaces, and run eight production-grade demo scenarios end to end
Format Full day (~6 hours), lecture + hands-on, self-paced friendly

This is not an "intro to AI coding assistants" course. We assume you know what an LLM and an agent are. If you want that foundation, start with the Copilot SDK Tutorial in this same site.


What is GitHub Copilot CLI?

GitHub Copilot CLI exposes the same agentic harness as GitHub's Copilot coding agent in your terminal. In a local, synchronous session, it can read files, edit them, run shell commands, and operate on GitHub.com resources such as issues, pull requests, and Actions through natural language (About GitHub Copilot CLI; github/copilot-cli README).

graph LR
    Dev[You in the terminal] -->|natural language| CLI[Copilot CLI agent]
    CLI -->|read / edit / execute| FS[Local files & shell]
    CLI -->|GitHub MCP server| GH[GitHub.com: issues, PRs, Actions]
    CLI -->|/model| Models[GitHub-hosted or BYOK models]
    CLI -->|/mcp add| MCP[Custom MCP servers]
    CLI -->|/delegate| Cloud[Copilot cloud agent]

    style CLI fill:#bbdefb
    style GH fill:#c8e6c9

What it IS

  • A terminal-native agent that plans, edits, runs commands, and iterates — not just a chat box (Best practices).
  • IDE-independent: works the same over SSH, in a container, on a server, or in CI.
  • GitHub-aware out of the box: ships with the GitHub MCP server pre-configured, so issues, PRs, and Actions are reachable in natural language (Using Copilot CLI).
  • Scriptable: a single non-interactive command (copilot -p "…") makes it a building block for automation and CI/CD (About GitHub Copilot CLI).
  • Customizable and governable: custom instructions, custom agents, skills, hooks, MCP servers, and tool-permission controls.

What it is NOT

  • Not an autocomplete / inline-suggestion tool — that is Copilot in your IDE. The CLI is an agent, not a ghost-text completion engine (Copilot features).
  • Not a replacement for the VS Code experience — it complements it. Many teams use both (see Access Methods).
  • Not a hosted REST API or a model you fine-tune — to embed Copilot in your own program, use the Copilot SDK.
  • Not unattended by default — it asks for approval before running tools that can modify or execute files, unless you explicitly opt out (Security considerations).

The three access surfaces at a glance

The single most common question from experienced developers is "I already have Copilot in VS Code — why would I use the CLI?" This workshop answers that with a dedicated Access Methods chapter. The short version:

Surface Shape Reach for it when…
Copilot in VS Code (Agent mode, Chat, inline) GUI, IDE-integrated You are actively authoring code and want rich diffs, inline review, and editor context (Copilot features)
Copilot CLI Terminal agent You are on a server / SSH / container, automating in CI, or orchestrating multi-repo and long-running agentic tasks (About Copilot CLI)
Copilot SDK Library / API You are building a product that embeds the agent runtime (Copilot SDK Tutorial)

These are not mutually exclusive. The CLI, the IDE agent, and the SDK all read the same .github/ customization (instructions, agents, skills) — so investment in one pays off in the others (Copilot features).


Plans & licensing

Copilot CLI is available across Copilot plans, but some controls are plan-gated. Always confirm the current matrix on the official pages, since plans change.

For the authoritative plan comparison and pricing, see Plans for GitHub Copilot and Models and pricing.


Workshop agenda (full day)

graph TD
    A["Part 0 · Getting Started<br/>install · auth · modes · config"] --> B["Part 1 · Access Methods<br/>VS Code vs SDK vs CLI"]
    B --> C["Part 2 · Feature Deep Dive<br/>agents · skills · hooks · MCP · BYOK"]
    C --> D["Part 3 · Demo Scenarios (8)<br/>hands-on, end-to-end"]
    D --> E["Part 4 · Decision & Wrap-up"]
Part Chapter Time Outcome
0 Getting Started 45 min CLI installed, authenticated, all four interaction modes understood
1 Access Methods: VS Code vs SDK vs CLI 45 min A defensible decision framework with Pros/Cons
2 Feature Deep Dive 75 min Mastery of customization, agents, skills, hooks, MCP, sandboxing, BYOK
3 Demo Scenarios 3 hr Eight repeatable, high-value workflows executed hands-on
4 Decision Guide + References 30 min Take-home checklist and primary-source library

Times are guidance for a facilitated session; the material is also designed for self-paced study.


Companion video walkthroughs

Prefer to watch first, then do? These three official GitHub videos cover the same ground as this workshop and are great to play before (or alongside) each part. Full descriptions and timestamps-by-topic live in References → Talks & demos.

Video Best paired with
Ultimate GitHub Copilot CLI tutorial for beginners Getting StartedFeature Deep Dive
Build with the Copilot CLI — Mona Mayhem Demo Scenarios
Less // TODO: more done with GitHub Copilot CLI Feature Deep Dive + Copilot SDK Tutorial

These are narrated, hands-on demos by GitHub. This site adds the why, the primary-source citations, and a reproducible eight-demo storyline so you can practice each idea against your own fork.


How to use this site

  • Facilitators: each chapter is standalone. Project the page, run the commands live, and let attendees follow along in their own clone.
  • Self-paced learners: work top to bottom. Every demo lists its prerequisites and a copy-pasteable command sequence.
  • All eight demos share one subject apptemplate-typescript-react — and tell a connected story. Fork it first so you can reproduce each step against your own copy.

Ready? Start with Getting Started.