GitHub Copilot Academy · 100 lessons available
Master the GitHub Copilot stack
100 free lessons on GitHub Copilot, from first install to enterprise governance — every version-sensitive claim checked against GitHub's own documentation and re-checked on a schedule.
Already know your role? Choose a learning path.
Continue learning
path
Every lesson complete.
of lessons complete
- Published
- 100 lessons · 8 clusters
- Roadmap
- 8 clusters
- Plans covered
- 7 tiers · 5 paid
Start here
Choose your Copilot path
Three questions, answered in your browser. They pick one of the five role-based routes through the lessons that already exist — no new material, no account, nothing sent anywhere.
You chose
Recommended path
The progression
From first suggestion to agentic engineer
Eight clusters, in order. Each one assumes the one before it, so you are never asked to reason about MCP servers before you have accepted your first inline suggestion.
- Level 1Copilot Fundamentals13 lessons · BeginnerAvailable now
- Level 2IDE Workflows12 lessons · Beginner → IntermediateAvailable now
- Level 3Programming Languages13 lessons · IntermediateAvailable now
- Level 4DevOps & Infrastructure13 lessons · Intermediate → AdvancedAvailable now
- Level 5Copilot CLI12 lessons · IntermediateAvailable now
- Level 6Prompting & Customization12 lessons · IntermediateAvailable now
- Level 7Agents + MCP13 lessons · AdvancedAvailable now
- Level 8Security + Enterprise12 lessons · AdvancedAvailable now
- DestinationAgentic Software Engineer
What changed
Latest Copilot changes
Sourced and dated. Models retired, models added, and the corrections we made to our own data as a result.
- deprecatedMAI-Code-1-FlashDeprecated. Replacement named: MAI-Code-1.1-Flash.
- addedClaude Fable 5.1Added to the supported-models list.
- deprecatedClaude Opus 4.5Deprecated. Replacements named: Claude Opus 4.7, 4.8 or Opus 5.
Start here
Featured lessons
13 of the 100 published lessons, drawn from all 8 clusters: GitHub Copilot Fundamentals, Copilot IDE & Developer Workflow, GitHub Copilot Programming Languages, GitHub Copilot for DevOps & Infrastructure, GitHub Copilot CLI, Prompting, Models & Customization, Copilot Agents, MCP & Agentic Development and Copilot Security, Code Review & Enterprise.
- GitHub Copilot: The Complete GuideThe pillar overview of every Copilot surface — completions, chat, agent mode, the cloud agent, the CLI, models, plans and security — with links into each deeper lesson.
- GitHub Copilot Agent Mode ExplainedThe in-IDE autonomous loop: planning, multi-file edits, tool calls, terminal approval, iteration — and the tasks it handles badly.
- GitHub Copilot for IDEs: Complete GuideEvery Copilot feature compared across VS Code, Visual Studio, JetBrains, Eclipse and Xcode, with the differences that should actually influence which editor you use.
- GitHub Copilot IDE Workflow: From Idea to Pull RequestThe capstone: sixteen steps from a one-line requirement to a reviewed pull request, marking exactly which steps are Copilot's and which are yours.
- GitHub Copilot for Programming: Language GuideHow Copilot behaves differently across twelve languages — what type systems, package ecosystems and test tooling change about the review you owe generated code.
- GitHub Copilot for PythonType hints as context, virtual environments, pytest generation, Ruff and mypy as the review layer — and the mutable-default and unsafe-subprocess patterns to watch for.
- GitHub Copilot for JavaScriptThe language GitHub names as one of Copilot's best supported — and why that makes browser/Node confusion, unhandled promises and npm supply-chain risk the things to check.
- GitHub Copilot for SQLSchema as context, joins and window functions, dialect differences between PostgreSQL, MySQL, SQL Server and SQLite — and queries that return confident, wrong numbers.
- GitHub Copilot for DevOps EngineersThe validation model for AI-generated infrastructure: what Copilot genuinely helps with across twelve technologies, and the gates that decide whether any of it is trustworthy.
- GitHub Copilot for KubernetesProbes, resource limits, securityContext and NetworkPolicy — and a manifest that passed strict schema validation with twenty-one security findings in it.
- GitHub Copilot for TerraformVariables, for_each, modules and state, with fmt, validate and a policy scan as the gate — and why a security fix can add three new findings.
- GitHub Copilot for GitHub ActionsLeast-privilege permissions, SHA-pinned actions, OIDC and the script-injection vector in a pull request title — the one thing actionlint does catch.
- Build an AI-Powered CI/CD Pipeline with GitHub CopilotThe capstone: one repository where Copilot writes every artefact and validates none of them — application, image, Terraform, manifests and workflow, behind gates that ran for real.
Cluster 4
GitHub Copilot for DevOps & Infrastructure
Infrastructure is the code where being wrong is most expensive and where the language checks least. A Kubernetes manifest with a privileged container, no limits and a public load balancer is a completely valid Kubernetes manifest.
13 lessons, each with a repository-ready example and an honest record of which validation commands were actually run.
Open Cluster 4Validate before deploying
- Docker
hadolint, then docker build - Docker Compose
docker compose config - Kubernetes
kubeconform for schema, Checkov for policy - Terraform
terraform fmt, validate, then plan - OpenTofu
tofu fmt, validate, then plan - Ansible
ansible-playbook --syntax-check, then ansible-lint - Linux
shellcheck and bash -n; read every command first - AWS
JSON and IaC validation locally; never a live apply - Microsoft Azure
bicep build, then az deployment what-if - Google Cloud
Terraform validate; never a live gcloud mutation - GitHub Actions
actionlint, then read the permissions block - AI-Powered CI/CD
Every gate below, run in CI on every pull request
Cluster 3
GitHub Copilot Programming Languages
Copilot behaves differently in every language, and the difference is not a ranking — it is what the language lets you verify. Rust's borrow checker and TypeScript's compiler reject whole classes of wrong suggestion before a human reads them. Python, Ruby, shell and SQL do not.
13 lessons: one pillar and twelve languages, each with its own tooling, failure modes and security review.
Open Cluster 3The fastest honest check, per language
- Python
mypy . && pytest -q - JavaScript
node --test && npx eslint . - TypeScript
tsc --noEmit - Java
mvn -q verify - C#
dotnet build -warnaserror && dotnet test - C++
compile with -Wall -Wextra -fsanitize=address,undefined, then run the tests - Go
go vet ./... && go test -race ./... - Rust
cargo clippy -- -D warnings && cargo test - PHP
vendor/bin/phpstan analyse && vendor/bin/phpunit - Ruby
bundle exec rubocop && bundle exec rspec - Bash
shellcheck script.sh && bash -n script.sh - SQL
EXPLAIN the query, then run it inside a transaction you roll back
Cluster 2
Copilot IDE & Developer Workflow
Copilot is not the same product in every editor. Agent mode, edit mode, checkpoints, code review and custom instructions each vary by IDE — and some differences, like Xcode accepting only the first line of a suggestion with Tab, will catch you out on day one.
12 lessons, one per editor, ending with a complete idea-to-pull-request workflow.
Open Cluster 2Feature coverage across 18 tracked capabilities
- VS Code14 supported · 3 preview · 1 not supported
- Visual Studio16 supported · 0 preview · 2 not supported
- JetBrains9 supported · 7 preview · 2 not supported
- Eclipse8 supported · 3 preview · 7 not supported
- Xcode6 supported · 6 preview · 6 not supported
Counts are read from the same data the lessons use, and the bars restate the numbers beside them — nothing here is communicated by colour alone. This is not a ranking: preview features work today, and two of the tracked capabilities are upgrade agents that exist in exactly one editor each. Read the lesson, not the total.
Original research
Things we measured ourselves
Experiments run here, with the environment, the sample and the limitations stated so the result can be challenged. Published harnesses with no results yet are in the research index and deliberately not on this page.
Measured result
GitHub's Copilot Docs Contradicted Its Own Changelog on Retirement Day
When GitHub retires a Copilot model, does its reference documentation reflect that on the day the retirement takes effect?
Measured result
Who Actually Crawls a Developer Documentation Site
On a small, new technical documentation site, what proportion of HTTP requests come from people, and how does AI-assistant crawling compare to search-engine crawling?
Why this site
Why The Copilot Stack?
Structured learning
Lessons are ordered and cross-linked, so you build on what you already know instead of assembling an understanding from disconnected search results.
Practical developer examples
Real prompts, real commands, real configuration, and comparison tables you can act on — not paraphrased marketing copy about productivity.
Technical reference
Shortcuts, slash commands, CLI flags, model availability and plan capabilities, organised so you can find one fact quickly and leave.
Current information
GitHub changes Copilot constantly. Version-sensitive pages carry a "last technically verified" date and a review interval that a build-time script enforces.
Built for developers
Where a feature is limited, in preview, or genuinely a bad fit for the task, the lesson says so. Nothing here is padded to reach a word count.
Beyond the lessons
Where to go next
Which plan should I buy?
The Plan Finder asks about team size, budget and the controls you need, then names a tier and says why. Built on the maintained plan dataset, not a sales page.
How much of this do I actually know?
One assessment per cluster. Every wrong answer names the lesson to reread, and a result is a reading list rather than a score.
I would rather build something
Guided labs with prerequisites, validation and a cleanup step — and twelve companion project repositories to take away.
How does Copilot compare to the alternatives?
Comparisons against Cursor, Claude Code and Amazon Q, written with real prices and the reasons not to switch included.
I need the templates, not the theory
Pro is the implementation layer: instruction files for six stacks, the projects packaged for work use. The Academy stays free.
Begin with the fundamentals
Cluster 1 takes you from "what is this product" to reviewing an agent's pull request with a clear idea of what to look for. Cluster 2 then covers the editor you actually use.
The Copilot Stack is an independent educational resource and is not affiliated with, sponsored by, or endorsed by GitHub, Inc. or Microsoft Corporation. GitHub, GitHub Copilot, and related names and marks are trademarks of their respective owners.