Prepare your tools, accounts and spending limits

Goal: choose one learning environment, install only its required tools, and prove that you can run a small local check before starting an agent. Product guidance and official links were checked on 2026-09-08. Offers, prices, models and availability can change; the linked provider pages remain the authority.

External download, signup and product-documentation links on this page open in a new tab. Keep this guide open while following them. Internal lesson links stay in this tab. On the site, use Copy code on a command or prompt block; in a Markdown file, select and copy the block’s contents without the backticks. Never paste a whole page into a terminal.

[!IMPORTANT] The first local lab does not require a credit card, Azure, Codespaces, a paid Copilot plan or a published GitHub repository. A GitHub account and authorized Copilot access are needed for the live AI steps, not for the deterministic local tests. Creating an account does not complete a lab.

1. Choose one route

For live Copilot work, add the account and plan checks below. You do not need both editors or both local and cloud environments. A browser editor is not automatically a runtime, and Azure is not the service that bills Codespaces.

2. Create your personal GitHub account

1. Open GitHub signup.
2. For this independent workshop, use a personal email address you control and expect to keep. This is a learning recommendation, not a GitHub rule that forbids work addresses. Do not borrow an account or bypass an employer’s managed-account policies.
3. Follow the signup flow, choose your username and complete the requested verification. Never share your password, email verification code or recovery codes with an instructor or agent.
4. Verify your email by following GitHub’s message. If it does not arrive, check the spam folder and the address entered before requesting another message.
5. Enable two-factor authentication, preferably an authenticator or security key, and save recovery codes somewhere private.
6. Confirm that your avatar menu shows your own username. Use this account consistently when activating Copilot and signing in to your chosen client.

Checkpoint: you can sign in, your email is verified, and you know how to recover access. Your GitHub password is not a Git command-line credential. A local Git repository and a GitHub-hosted repository are different things; publication is optional and covered in the download guide.

Official help: create an account, verify email, and configure two-factor authentication.

3. Understand Copilot Free and the other plans

GitHub Free is an account plan; Copilot Free is an AI product plan. Signing in to GitHub does not prove that Copilot is enabled in your editor or CLI. Start with the free option if you are eligible and it covers the exercise.

These are published reference prices, not a checkout quote. Taxes, eligibility, allowances and offers can differ or change. An organization license is not a reason to purchase a second personal subscription.

Current plans use GitHub AI Credits for metered AI usage. A credit is not one question: usage depends on the model and tokens. Included and flex credits, available models, additional paid usage and organizational policy differ. Do not reuse old premium-request tables for new subscriptions; some existing annual subscriptions remain on explicitly documented legacy billing. The current Free documentation does not publish a numeric AI Credits allowance on the pages checked here, so this guide does not invent one.

1. Read the official plan comparison and the current purchase options.
2. While signed in to your personal account, follow the plan activation instructions. Select Free if eligible; do not select a paid trial or upgrade just to follow this workshop.
3. Review usage and billing controls before enabling additional paid usage. Copilot, Codespaces and Azure have separate allowances and controls.
4. If your account is managed, ask the administrator about access restrictions; do not create a second account to evade policy.

Copilot CLI is available with all Copilot plans, including Free, subject to usage limits and applicable policy. That does not make every model, cloud-agent operation or IDE integration universally available. If an allowance is exhausted, record the blocker, continue with local tests, and resume the AI steps when access is available; upgrading is not compulsory.

Eligible learners can check Copilot Student. Verification and activation are separate steps; do not assume an academic email alone grants the benefit. See also individual AI billing and organization billing.

4. Install VS Code Stable or Insiders

The editions can run side by side. Insiders is not a paid Copilot plan and does not grant extra account permissions. Do not assume that installing extensions or signing in in one edition prepares the other: check the edition and profile you actually use. Stable and Insiders use different Settings Sync services by default; sharing them is an explicit choice.

1. Choose your operating system and processor on the download page, then follow the installer. On macOS, move the application into Applications.
2. Open the selected editor. Use File > Open Folder to open only the extracted learning workspace, not your entire home folder.
3. Review the files before granting Workspace Trust. Trust enables execution; it is not a claim that downloaded code is safe.
4. Open the Copilot/Chat entry in that edition and follow the GitHub sign-in flow. Review requested access and confirm the intended username. If prompted for an extension, use the official GitHub publisher.
5. If the terminal command is unavailable on macOS, open the Command Palette and run Shell Command: Install ‘code’ command in PATH, or the corresponding Insiders command. Reopen the terminal afterward.

For Stable, run this in the terminal:

code --version

For Insiders instead:

code-insiders --version

After extracting a kit, change into the workspace named by its first-run guide and use one of these commands, not both:

code .

code-insiders .

Checkpoint: the intended edition opens the intended folder. A harmless read-only request succeeds in Copilot if you are doing the live steps. Use Start here for the first guided Ask, Plan and Agent session. Do not infer that a Preview feature exists in Stable merely because it appeared in an Insiders demonstration.

Official help: editor setup, Copilot setup, terminal commands, and Settings Sync.

5. Install only the selected lab’s runtime

All 14 adventures and the first hands-on lab use Node 24 for their local checks. For another hands-on kit, read its first-run guide before installing a second language. No global installation of a test framework is needed for the first Node exercise.

After installing Node, reopen the terminal and run:

node --version
npm --version

Expect Node’s version to begin with v24.. A version check proves the executable is found; the lab’s baseline is the next check. When using Git, also run:

git --version

For a C# kit only:

dotnet --list-sdks

Confirm that a compatible .NET 10 SDK is listed. For Python only, select the same interpreter in VS Code and the terminal and follow the Python setup lab’s virtual-environment instructions. On Windows, py may be the installed launcher instead of python. Do not install packages into an unrelated global environment.

6. Install the standalone Copilot CLI

Copilot CLI uses the command copilot. GitHub CLI uses gh. They are different products. The old gh copilot extension is not the installation route taught here. Choose one installer below; do not install every variant. The npm route requires Node 22 or later; this curriculum already uses Node 24. That Node requirement belongs to the npm installer, not every native installer.

macOS

If Homebrew is already available, use the command in the official installation guide:

brew install --cask copilot-cli

If you do not use Homebrew, the npm route below is an alternative once Node is installed. You do not need to install a package manager solely to follow this route.

Windows

Use PowerShell 6 or later; a current PowerShell 7 installation is a suitable choice. Windows PowerShell 5.1 and PowerShell 7 are different installations. See PowerShell installation if needed. Check the shell you opened:

$PSVersionTable.PSVersion

With WinGet available:

winget install GitHub.Copilot

Reopen PowerShell so it can see the updated PATH. If WinGet is unavailable, follow the approved installation route for your machine or use npm with Node.

Linux, or npm on macOS/Windows

With Node available, this is the cross-platform npm installation:

npm install -g @github/copilot

If installation reports a permissions error, use a user-owned Node installation or your administrator’s approved method. Do not solve it by running arbitrary downloads as an administrator.

As an alternative on macOS/Linux, download the official installer without executing it immediately:

curl -fsSL https://gh.io/copilot-install -o copilot-install.sh

Open the downloaded file in an editor, verify the official source and review its downloads and installation paths. Only if you approve that exact file, run:

bash ./copilot-install.sh

Verify, authenticate and start safely

In a new terminal:

copilot --version

Then change into the extracted kit’s intended workspace and start:

copilot

Inside the interactive Copilot session, not the operating-system shell:

/login

Follow the browser or device-code flow shown by your session, using your intended GitHub account. Do not paste tokens into lesson files or prompts. After authentication, inspect the session’s help:

/help

Use a read-only prompt before allowing changes:

Inspect this learning workspace and identify its starter files and existing tests.
Do not edit files, install packages, access secrets, publish code or run commands yet.
State the working directory, the proposed verification command and what it proves.

Checkpoint: you have the CLI version, the intended account and workspace, and an explanation you can check against real files. Review each permission request; do not enable blanket approval for a training exercise. Then follow the extracted-kit CLI workflow. Ask, Plan and Agent describe responsibilities here; the CLI need not present the same buttons, roles or handoffs as VS Code.

Official help: CLI installation, CLI authentication, and current command reference.

7. Use a Codespace instead of a local runtime, if useful

A GitHub Codespace is a hosted development environment: a Linux container on a virtual machine with files, tools and a terminal. You can use its VS Code web interface or connect from VS Code desktop. Code runs remotely, not on your laptop. You still need Copilot authorization for live AI steps.

Codespaces is not github.dev. The lightweight editor reached by pressing . on a GitHub repository does not provide compute or an integrated terminal for building and running these labs. Its documentation currently labels it public preview.

Check quota before creating the environment

Core-hours multiply running time by machine cores: 120 core-hours can mean up to 60 running hours on a 2-core machine, before other usage or storage limits. GB-month measures storage over time; it is not merely a disk-size allowance. Organization-paid environments and their policies can differ.

1. Read current Codespaces billing and check who will pay. With no valid payment method, use is blocked after the included allowance is exhausted.
2. If paid usage is enabled, review the budget and Stop usage when budget limit is reached setting. A budget alert alone is not a spending stop.
3. Review this repository’s container configuration and setup script before creating a Codespace. Provisioning downloads tools and consumes resources; a local ZIP is simpler if those tools are already installed.

Create and run one isolated exercise

1. Open the curriculum repository on GitHub.

2. Select Code > Codespaces, inspect the payer notice and choose the creation options. Check the branch and machine size; use a small available machine appropriate for the selected exercise.

3. Create the Codespace and wait for setup to finish. If setup fails, read the reported error; do not call a half-created environment ready.

4. In its terminal, from the curriculum checkout root, prepare only the first fixture in an unused directory outside that checkout:

   node scripts/prepare-hands-on.js --lab 01-interface --destination /workspaces/aca-01-interface
   

5. Open /workspaces/aca-01-interface as the folder in the editor and start a new terminal there. The preparation command creates a fixture copy, not a downloaded ZIP, so it does not create KIT-VERIFY.cjs.

6. Run the existing tests:

   node --test --test-concurrency=1 greeting.test.mjs
   

7. Expect 2 passing baseline tests. Record your actual result, then continue the investigation and change in Start here. If the destination already exists, preserve it and select a different unused name; the preparation tool deliberately refuses to overwrite it.

For the desktop route, install the official GitHub Codespaces extension and follow connecting with VS Code. CLI access, extensions and authenticated sessions must be checked inside the remote environment, not inferred from your laptop.

Stop or delete deliberately

At your Codespaces list, use the environment’s menu:

• Stop codespace stops compute but preserves files; storage is still counted.
• Delete removes the environment. Save, export or commit/push work to your authorized repository first, after reviewing the files. Deletion does not undo usage already incurred.
• Merely closing the browser tab does not stop the Codespace.

Official help: what Codespaces is, the github.dev distinction, creation steps, budgets, stopping, and deleting.

8. Azure personal free account: optional, with conditions

[!CAUTION] Do not create Azure resources to pass a local lab. Azure credit does not pay for a Copilot subscription or GitHub Codespaces. The GitHub cloud agent is also not a requirement to create an Azure subscription.

The Azure free-account offer advertises US$200 in credit for the first 30 days, for eligible new customers. This is not R$200, not cash paid to you, and not an unconditional benefit for every existing account. Normally only one free-account benefit is available per new customer.

1. Open the official Azure free-account offer and read the eligibility and country/region conditions.
2. If eligible and you want this optional cloud route, select the free-account signup, not a pay-as-you-go purchase by mistake. Use a personal Microsoft account or the supported GitHub sign-in option with an email you control.
3. Complete the identity checks. The offer requests a phone and a non-prepaid credit/debit card where supported; Brazil and Hong Kong require a credit card for this offer. A temporary US$1-equivalent authorization may appear. Use accurate details; do not enter a payment card into a chat or repository.
4. Check the actual subscription offer, remaining credit and expiration in the portal before deploying anything. If the benefit is not shown, stop and consult the provider rather than assuming credit exists.
5. Keep the free-account spending limit in place. Microsoft says it does not charge this free account unless you explicitly choose to upgrade to pay-as-you-go. When credit or the 30 days ends, services are disabled unless you opt to upgrade; this is not an instruction to upgrade.
6. If you later choose pay-as-you-go, review costs first. The free-credit spending limit is not available on pay-as-you-go; budget alerts are not a hard cap. Stop/delete only your own exercise resources after saving needed work.

Some items, including certain Marketplace products and support, are excluded. Check the offer terms, spending-limit behavior and what upgrading changes.

Eligible students can instead review Azure for Students: the published offer is US$100 for 12 months without a credit card, with its own age, institution, enrollment and regional requirements. Its eligibility is not the same as GitHub Education. Neither student benefit is guaranteed by this curriculum.

9. Final readiness checklist and troubleshooting

• I selected one environment and know where files and processes run.
• I verified my own GitHub account and intended Copilot plan for live AI work.
• My selected runtime is found in the terminal used for the lab.
• I downloaded one matching kit, extracted it into a new folder, and read its workspace and expected baseline instructions.
• I can distinguish a terminal command from a Copilot prompt or slash command.
• I know that an intentional starter failure is different from missing tools.
• I reviewed any optional cloud quota and know how to stop and remove my work.
• I recorded an actual local result, not a copied success message.

Continue to your first verified local change, or choose one adventure or hands-on ZIP. Full lessons include investigation, planning, implementation, a negative case, evidence and safe reset.