How to Use the GitHub Copilot App: The Beginner's Guide
CodexFuse Team · 2026-11-02T19:45:49.085Z · 14 min read min read
The GitHub Copilot app: mission control for agentic software delivery. A beginner guide to download, first session, Projects, Automations, and costs.
The GitHub Copilot app is GitHub's standalone desktop app for agentic software delivery: instead of chatting with Copilot inside your editor, you open one surface where you start agent sessions across repositories, write specs, browse issues, create pull requests, and watch your agents work — from issue to merge in one place. This GitHub Copilot app tutorial walks the full beginner arc: downloading the app, signing in, running your first agent session, using Projects and Automations, switching devices mid-session, and what the cloud experience actually costs. If you have never used Copilot itself, start with our GitHub Copilot beginner guide — the app builds on the same agent, but the mental model is different enough that this guide earns its own page.
What Is the GitHub Copilot App, Exactly?
The GitHub Copilot app is a standalone mission-control app for working with Copilot's AI agents across your projects. It connects agents and tools across repositories, and the work goes beyond code — you can write specs, present design discussions, and analyze telemetry in the same surface.
The timeline matters because early coverage describes a thinner product. GitHub introduced the app in June 2026 at Microsoft Build as a technical preview for Pro, Pro+, Business, and Enterprise subscribers, then relaunched it in a redesigned, unified form at GitHub Universe on October 28, 2026 — that is the current app this guide covers.
Do not confuse it with Microsoft's consumer Copilot app, the chatbot built into Windows and copilot.com — that is a different product with the same first word. This guide is about GitHub's developer app only. It is also not the Copilot chat inside VS Code: the editor chat is where you pair-program line by line, while the app is where you delegate whole tasks — issues, sessions, pull requests — across projects. Code reviews live in a separate lane covered by our Code Review Max guide, which this article does not cover.
How Do You Download the GitHub Copilot App?
The GitHub Copilot app download starts from the download page linked in GitHub's official quickstart — the vendor does not publish a separate standalone download URL, so start from the quickstart and pick your build. Desktop builds exist for macOS, Linux, and Windows, and on Windows you can also run agent sessions inside a WSL 2 distribution.
| Client | What you get | Best for |
|---|---|---|
| Desktop app (macOS, Linux, Windows) | The full mission-control surface: sessions, Projects, Automations, Chats | Daily driving — this is the real app |
| GitHub Mobile | Session-continuity controls — find and steer the same session, not a standalone Copilot app | Checking a run from your phone |
| github.com | Start cloud sessions from the chat interface on the GitHub website | Quick sessions without installing anything |
| Copilot CLI and IDEs | Continue the same session in the terminal or your editor | Steering from where you already work |
GitHub's prerequisites are short: a GitHub account; a Copilot plan, or you can configure your own model provider with provider credentials such as an API key; and Git installed on your computer. Sign-in is the obvious flow — click "Sign in to GitHub" and follow the prompts — with two honest footnotes: GitHub Enterprise Server is not supported by the GitHub Copilot app, and if you have no Copilot plan the app lets you choose whether to sign up for a plan or continue with your own model provider (bring your own key).
How Much Does the GitHub Copilot App Cost?
The GitHub Copilot app is available for all Copilot plans, according to GitHub — but the full cloud experience is a paid-plan feature, and beginners should know the line before they fall in love with it. The honest summary, dated 2026-11-03: the app itself costs nothing extra on any plan, while cloud sandbox sessions — the runs that follow you across devices — belong to paid plans.
Free keeps the old chat experience. Independent pricing trackers report that Copilot Free excludes the cloud agent entirely: tokensave.app (October 11, 2026) lists Free at $0 with 2,000 code completions a month, a limited allowance of chat and agent use, and access to lightweight models such as Claude Haiku 4.5 and GPT-5 mini, while Pro at $10 a month adds unlimited completions, the cloud agent and code review, model choice, and 1,500 credits (1,000 base plus 500 flex). Pro+ is $39 with 7,000 credits and Max is $100 with 20,000 credits on the same tracker. Alternatively, the vendor lets you skip a plan and bring your own model provider key.
| Cloud sandbox meter | Vendor rate (USD) | What it covers |
|---|---|---|
| Compute | $0.000024 per vCPU second | Agent runtime in the cloud sandbox |
| Memory | $0.000003 per GiB second | RAM used by the running session |
| Storage | $0.005 per GiB month | Snapshot storage for stopped sessions |
For teams, the lines are policy switches. Business and Enterprise organizations get the app policy enabled by default, and the "Copilot in the cloud" policy controls whether members can start sessions in a cloud sandbox — also enabled by default. At Universe 2026, GitHub said Enterprise customers can use all Copilot features (app, code review, automations) without a separate subscription, paying for the AI credits they use — immediate through GitHub Sales, with self-serve coming later.
GitHub's own cost discipline tips are worth adopting from day one: match the model's capability to the task instead of defaulting to the biggest one, use Plan mode to validate scope before a run, scope ideas in Chats before opening a full session, start a new session per task rather than one endless thread, and run /chronicle cost tips to review what your sessions actually spent.
What Happens in Your First App Session?
Your first session is a seven-step loop, straight from GitHub's quickstart: install, connect a repository, pick a mode, prompt, iterate on the diff, and open a pull request. It is deliberately small — the point is to feel the loop, not to ship a feature.
- Install and sign in to the GitHub Copilot app.
- Connect a repository or local folder.
- In the sidebar, next to Sessions, start a session in the connected repository or folder.
- Select a session mode from the dropdown below the prompt field — for example, Interactive to work collaboratively with the agent.
- In the prompt box, paste this prompt: "Suggest a small low-risk code change in this repository, implement it, and explain the diff."
- Respond to any requests for input from the agent. After the agent makes changes, click Changes above the prompt box to view the diff, and keep iterating in the same session until you are happy.
- If you want to keep the change, click Create PR to open a pull request for review — then click PR above the prompt box to view it directly in the app.
| Session mode | How it works | Use it when |
|---|---|---|
| Interactive | You and the agent work collaboratively, back and forth | Learning the loop, or any task you want to steer closely |
| Plan | The agent plans first and you approve the plan before it runs | Risky or large changes where scope must be agreed first |
| Autopilot | The agent works fully autonomously | Small, well-understood chores you trust it to finish alone |
Before you prompt, pick your model in the picker — and note the one that is not a model at all. Copilot Fusion, generally available since October 28, 2026, is a routing and orchestration system: you select it, choose an effort level (Fusion Standard for everyday coding, Fusion High for challenging tasks, Fusion Max for the most demanding work), and it routes across models and subagents. GitHub's vendor claim, reported by Unite.ai, is that Fusion Max scored 60 on the Artificial Analysis Coding Agent Index — within two points of Claude Code with Opus 5.5 Max — at about half the average API cost per task; paid plans also get a limited-time 30% discount on model costs when using it. Or bring your own key and use your own provider.
One more beginner superpower: parallel workspaces. The app can run multiple isolated agent sessions at the same time, each with a dedicated git worktree and branch — so one agent can refactor while another writes tests, without stepping on each other's files. If worktrees are new to you, our Git beginner guide covers the branch and worktree concepts underneath.
What Does the App's Sidebar Actually Do?
The sidebar is the whole app, really: New, Pull requests, Issues, Automations, Customize, Search, Projects, Sessions, and Chats. Learn these nine and you can drive the entire surface.
Automations is the scheduling lane. You save recurring agent tasks and run them on a schedule or on demand — the prompt-plus-trigger workflows that keep working while you sleep. The full walkthrough lives in our GitHub Copilot Automations guide, so treat this tab as the doorway and read that guide before you build anything serious.
Projects organizes ongoing work with Copilot: within a project you start multiple sessions across repositories, track progress, share context, review artifacts, and set up recurring tasks — natively in the app since October 28, 2026 (public preview, per Unite.ai). Because project work lives in issues and pull requests across repos, the repo-workflow basics in our GitHub beginner guide transfer directly. Chats is the low-stakes sibling: brainstorm in conversation mode without creating a dedicated branch or workspace, which makes it the right place to scope an idea before you spend agent time on a full session.
Sessions is your mission-control list, Pull requests and Issues let you work the GitHub workflow without leaving the app, Search finds sessions and artifacts, and Customize tailors the experience. And /chronicle is your memory: session history plus the cost tips mentioned earlier, so you can see what you ran and what it spent.
How Is the GitHub Copilot App Different From VS Code?
VS Code is where you write code with Copilot looking over your shoulder; the GitHub Copilot app is where you manage the agents doing the work across every surface. Same agent family, different job — and that is the whole copilot app vs vscode debate in one sentence: the app assumes you will use both.
| Dimension | GitHub Copilot app | Copilot in VS Code |
|---|---|---|
| Role | Standalone mission control for agent work across projects | In-editor pair programmer for the file in front of you |
| Sessions | Parallel workspaces, each with its own worktree and branch | Sessions tied to the open editor workspace |
| Continuity | Follow and steer a session from any client — phone, web, CLI, IDE | Bound to the editor window (until Agent Host detaches it) |
| Modes | Interactive, Plan, Autopilot at the session level | Chat and agent mode inside the editor |
| Best for | Delegating multi-repo work: issues, sessions, PRs, telemetry | Writing and editing code with AI assistance inline |
The signature move is app-switching: start a cloud session from your phone at standup, steer it from VS Code at your desk, check it from github.com at lunch. GitHub's changelog calls this session continuity — the same session also continues in the Copilot CLI, and sessions started in a Slack or Teams DM can be continued in these clients. A companion deep-dive on session continuity is planned as its own guide; this page covers only the app surface. One related mechanism worth knowing: Agent Host isolates sessions from individual VS Code windows so your work continues after you close the project folder.
Two experimental edges to file away. Computer use, in public preview on macOS and Windows, lets local sessions interact with desktop applications — clicking controls, editing text — including legacy and GUI-only software with no API; each application requires your approval, and organizations can disable it by policy. And Fusion, covered above, is the model picker's routing system rather than a model.
What Mistakes Should Beginners Avoid?
Beginners lose the most time by treating the app like a fancier chat window. The seven mistakes below come from the vendor docs and the launch coverage — each one is a small habit change, not a feature to learn.
| Mistake | Why it hurts | Do this instead |
|---|---|---|
| Treating it like editor chat | Sessions create branches and worktrees; chat just answers | Think in sessions and artifacts, not messages |
| Expecting cloud sessions on Free | Free keeps the old chat experience — the cloud agent is a paid feature | Start on a paid plan, or bring your own model key |
| One giant session for everything | Context gets muddy and costs creep up | A new session per task; review with /chronicle |
| Skipping Plan mode on risky work | Autopilot can confidently do the wrong big thing | Plan first, approve the scope, then run |
| Ignoring computer-use approval prompts | The feature touches your desktop apps — approval is the safety rail | Approve each application deliberately; it is experimental |
| Never reviewing the diff | Agents write plausible-looking wrong code too | Check Changes before every Create PR |
| Starting from code instead of an issue | You skip the app's fastest flow | Browse Issues and start the session from the issue |
Sources
- GitHub Docs — GitHub Copilot app (copilot-surfaces/github-copilot-app): supported operating systems (macOS, Linux, Windows; WSL 2 on Windows), parallel workspaces, session modes, Automations, Chats, /chronicle, plan availability, computer use. Fetched 2026-11-02.
- GitHub Docs — Quickstart: GitHub Copilot app: download page, prerequisites (GitHub account; Copilot plan or own model provider; Git installed), sign-in flow, GitHub Enterprise Server not supported, seven-step first session. Fetched 2026-11-02.
- GitHub Docs — Copilot surfaces: the four surfaces (app, Copilot CLI, GitHub.com + GitHub Mobile, IDEs). Fetched 2026-11-02.
- GitHub Docs — Cloud sandboxes: "Copilot in the cloud" policy, metered billing (compute $0.000024/vCPU second, memory $0.000003/GiB second, storage $0.005/GiB month). Fetched 2026-11-02.
- GitHub Changelog, 2026-10-28 — Copilot in VS Code Universe releases: session continuity across VS Code, app, GitHub Mobile, github.com, CLI, Slack/Teams DMs; Agent Host; computer use experimental. Published 2026-10-28.
- Help Net Security, 2026-06-08 — GitHub introduces the Copilot app at Microsoft Build 2026: technical preview from May 2026 for Pro, Pro+, Business, Enterprise subscribers.
- Daily.dev — GitHub Universe 2026 summary (October 28, 2026): redesigned unified app, session continuity, Enterprise pay-per-AI-credit via GitHub Sales with self-serve later.
- Unite.ai — "GitHub Makes Copilot Fusion Generally Available at Universe 2026" (October 28, 2026): Fusion GA, model-picker effort levels, app/CLI/JetBrains availability, computer-use public preview, Projects natively in the app, 30% discount claim.
- The Deep View — "GitHub Unveils a Fix for AI's Token Economics Problem" (October 29, 2026): app-switching between phone and desktop, Fusion cost claims, Kyle Daigle's 50–60% cost-reduction quote.
- tokensave.app — Copilot pricing comparison (October 11, 2026): Free $0 (2,000 completions, limited chat/agent, lightweight models), Pro $10 (1,500 credits), Pro+ $39 (7,000 credits), Max $100 (20,000 credits).
- thecopilotstack.com — Copilot pricing tracker (September 2026): Free excludes the cloud agent; student tier includes it.
Frequently Asked Questions
Is the GitHub Copilot app free?
The app itself is available on all Copilot plans at no extra charge, but the cloud experience is not a Free feature. Copilot Free keeps the older chat experience with 2,000 monthly completions and lightweight models, and independent pricing trackers report that the cloud agent is excluded from Free entirely. Your options for the full app experience are a paid plan — Pro, Pro+, Max, or Enterprise — or bringing your own model provider key, which the app supports without a Copilot plan.
Can I use the GitHub Copilot app on my phone?
The github copilot app mobile question comes up a lot, and the answer is: there is no standalone GitHub Copilot app for phones. Mobile works through GitHub Mobile's session-continuity controls: you find the same cloud session you started on desktop and steer it from your phone. The practical pattern is start at your desk, check from the phone at lunch, and pick it back up anywhere — the session lives in the cloud, not on the device.
Is the GitHub Copilot app better than using Copilot in VS Code?
Neither is better — they do different jobs. The app is mission control for delegating work across repositories: sessions, Projects, Automations, pull requests, and telemetry. VS Code is where you write code with Copilot assisting inline. Most developers will use both, and session continuity means the same cloud session can move between them: start on the phone, steer in VS Code, check on the web.
What is Copilot Fusion in the app?
Copilot Fusion is a routing and orchestration system in the app's model picker, generally available since October 28, 2026 — not a model itself. You pick an effort level — Standard for everyday coding, High for challenging tasks, Max for the most demanding work — and it routes across models and subagents. GitHub's vendor claim is that Fusion Max scored 60 on the Artificial Analysis Coding Agent Index at about half the average API cost per task, and paid plans get a limited-time 30% discount on model costs while using it.
Does the GitHub Copilot app work with GitHub Enterprise Server?
No. GitHub's quickstart states plainly that GitHub Enterprise Server is not supported by the GitHub Copilot app. Enterprise Cloud organizations use it through the app policy, which is enabled by default, with the "Copilot in the cloud" policy controlling cloud sandbox sessions.
