Getting Started with VS Code and GitHub Copilot

This tutorial walks you through opening VS Code, finding the GitHub Copilot chat window, and running your first AI-assisted coding session - from data creation to a working web app.

📹 Prefer to watch? See the video walkthrough.

Prerequisites

  • VS Code installed (download here)
  • GitHub Copilot extension installed and signed in
  • Python installed on your machine (for the example app)

1. Opening a Folder in VS Code

Before you start, open the folder where you want to work.

  1. Launch VS Code.
  2. Go to File → Open Folder… (or press Ctrl+K Ctrl+O).
  3. Select a folder on your computer - for example, a new empty folder called my-first-copilot-project.
  4. Click Select Folder.

VS Code will now show the folder contents in the Explorer panel on the left.

Tip: Working inside an open folder gives Copilot context about your project - it can read and write files, run commands, and understand your code structure.

2. Opening the GitHub Copilot Chat Window

1779105481325

  1. If the window is not open, you can click the Chat icon in the top bar (it looks like a speech bubble).
  2. The Copilot Chat panel will open on the side of your editor.

3. Selecting an Agent

1779106033981

GitHub Copilot Chat is built around agents - each with a specific role. For this tutorial, use the Agent agent, which can take actions on your behalf: creating files, running commands, and building projects end to end.

  1. At the top of the chat panel, check the Agent Target dropdown and make sure Local [2] is selected. This means the agent runs interactively on your own machine with full access to your workspace.
  2. In the chat input box, click the agents dropdown [1] (it may show Ask by default).
  3. Select Agent from the list.
Agent Description
Agent Autonomously plans and implements changes across files, runs terminal commands, and invokes tools
Plan Creates a structured, step-by-step implementation plan before writing any code - hands it off to Agent when you're happy with it
Ask Answers questions about your code or general topics without making file changes

4. Selecting a Model

You can choose which AI model powers your Copilot session. Different models have different strengths.

  1. In the chat panel, click the model selector [3] (shown near the chat input box).
  2. Select Claude Sonnet 4.6 (or whichever model you prefer).

Tip: Claude Sonnet 4.6 is a strong general-purpose model well-suited for coding tasks and multi-step agent workflows.

4b. Choosing a Permission Level

When the Agent agent is active, it will ask your permission before running terminal commands or making changes. The permissions picker [4] lets you control how much autonomy it has:

Level Behaviour
Default Approvals Only read-only / safe tools run without confirmation - everything else prompts you
Bypass Approvals All tool calls are auto-approved; the agent may still ask clarifying questions
Autopilot Fully autonomous - auto-approves everything and continues until the task is done

For this tutorial, leave it on Default Approvals so you can see what the agent asks before it acts.

5. Your First Task: Create a Data File

Now let's put Copilot to work. With the Agent agent and Claude Sonnet 4.6 selected, type the following into the chat:

Please create a CSV file which contains the largest cities in the world, including their populations and coordinates

Copilot will:

  • Generate a cities.csv file with city names, populations, latitudes, and longitudes.
  • Write the file directly into your open folder.

You can review the file in the Explorer panel once it's created.

6. Your Second Task: Build a Web App

Now ask Copilot to build a visualization on top of that data:

Please create a Python web app which allows me to visualize and edit this data

Copilot will likely:

  1. Create a Python web application (e.g., using Flask or Streamlit).
  2. Set up a map or table view to display the city data.
  3. Add editing functionality so you can update the CSV entries.
  4. Tell you how to start the app - or you can ask it to start the app for you.

Letting Copilot Start the App

If Copilot gives you manual start instructions, you can simply ask:

Can you start the app for me?

Copilot will run the necessary terminal commands to launch the app and give you a local URL to open in your browser (typically http://localhost:5000 or similar).

7. Understanding Permission Prompts

When running in Agent mode, Copilot will sometimes ask for your permission before taking an action - such as running a terminal command, installing a package, or modifying a file.

Do not blindly accept these prompts. Take a moment to read what Copilot is asking to do. If you are unsure, ask:

What does this command do and why do you need to run it?

Copilot will explain its reasoning. This is a healthy habit that keeps you in control of your own environment.

Summary

You have now:

  • Opened a folder in VS Code
  • Found and configured the Copilot Chat window
  • Selected the Agent agent with the Local target
  • Used Copilot to autonomously create a data file and a working web app

Next Steps

  • Use the Plan agent to design a feature before building it - review the plan, then hand it off to Agent to implement
  • Learn about prompting techniques to get better results
  • Try adding a new feature to your cities app - just describe it to Copilot!