GHCP Across Multiple Surfaces

GitHub Copilot doesn't live in one place. Depending on where you are in your workflow — in a browser, an editor, a terminal, or building your own application — there is a dedicated surface designed for that context. Understanding which surface to reach for, and what it can do, is the first step to getting the most from Copilot.

Availability: Core Copilot features are available on all plans including Free; advanced capabilities vary by feature and may require Pro, Pro+, Business, or Enterprise.

The four surfaces at a glance

Surface	What it is	Best for	Plan requirement
GitHub.com	Browser-native Copilot at `github.com/copilot` — chat, agents, and project tools without opening an IDE	Issue management, coding-agent work, model comparison, rapid prototyping with Spark	Core chat: all plans incl. Free; coding agent: Pro/Pro+/Business/Enterprise; Spark: Pro+ and Enterprise (preview)
IDE	Copilot extensions for VS Code, JetBrains, Visual Studio, Eclipse, Vim/Neovim, Xcode	Code completion, inline chat, refactoring, debugging, local agent mode	All plans incl. Free
Copilot CLI	Terminal agent — plan, execute, resume, and delegate tasks directly from your shell	Autonomous task execution, long-running sessions, CI-adjacent workflows	All plans incl. Free
Copilot SDK	Programmatic API (Python, TypeScript, Go, .NET) exposing the same engine behind the CLI	Embedding Copilot agent workflows in your own apps and services	Paid plans for standard use; BYOK available

GitHub.com — your AI command centre

At github.com/copilot you get a browser-native Copilot experience with no IDE required. It handles coordination and exploration work that is awkward to do inside a code editor.

Capability	How to use it
File issues from screenshots	Drag a screenshot into the Copilot chat, then prompt: "Create a new issue using the 'bug' label…" — Copilot reads the image, generates a title and description, and applies your repo's issue template.
Assign coding agents	Tell Copilot: "Assign yourself to this issue and draft a fix." The agent analyses your codebase, identifies the root cause, and opens a draft pull request — entirely in a remote sandbox.
Prototype with GitHub Spark	Scaffold working code snippets and components, preview them live, and share via URL — without opening VS Code.
Switch AI models	Click the model name in any thread to switch between GPT-4.1, Claude Sonnet 4.5, Gemini 3 Pro, and many more — even mid-conversation. Each switch creates an independent response branch you can compare.
Navigate conversation branches	Each message is like a base commit — model switches create parallel "branches" so you can review different approaches without starting over.

Tip: Use GitHub.com for coordination (standup summaries, issue triage, cross-repo overview) and hand off implementation work to your IDE or CLI once the scope is clear.

IDE — where code gets written

The IDE surface is the most familiar entry point. Copilot extensions are available for VS Code, Visual Studio, all major JetBrains IDEs (IntelliJ, PyCharm, WebStorm…), Eclipse, Vim / Neovim, and Xcode.

Feature	What it does	How to trigger
Code completion	Inline ghost-text suggestions as you type, based on open files and context	Type normally; accept with `Tab`
Inline chat	Ask questions or request edits right between your lines of code; receive diffs inline	`Cmd+I` / `Ctrl+I`
Chat panel	Full conversation with slash commands: `/explain`, `/fix`, `/tests`, `/doc`, `/optimize`, `/new`	Open Copilot Chat sidebar
@workspace agent	Answers questions with awareness of your entire project, not just open files	Type `@workspace` in chat (VS Code / Visual Studio)
#file references	Scope Copilot to a specific file for targeted suggestions	Type `#` in the chat input and choose the file
Agent Mode	Multi-step autonomous file changes within your editor — Copilot plans, edits, and runs commands locally	Enable Agent Mode in VS Code / JetBrains / GitHub.com chat
Sparkles ✨	Quick AI actions surfaced inline — e.g. auto-generate commit messages in the commit input box	Click the ✨ icon where it appears

Tip: Keep only the files relevant to your current task open. Copilot only sees open files unless you use explicit #file references — more context is not always better when it is irrelevant.

Copilot CLI — the terminal agent

Copilot CLI is a full agentic runtime in your shell. Install it with npm install -g @github/copilot (or via WinGet on Windows: winget install GitHub.Copilot). Sessions persist on disk and automatic context compaction enables very long-running tasks — see the Compaction and Infinite Sessions in CLI and SDK post for details.

Command / action	What it does
`/plan`	Create an implementation plan before coding — Copilot outlines each step and asks clarifying questions before making any changes
`/model`	Switch the active AI model for the current session
`/fleet`	Break a complex task into independent subtasks that subagents run in parallel — the main agent orchestrates the workflow so large multi-step work completes faster
`/resume`	Return to a saved session, including coding-agent sessions started on GitHub.com
`/delegate`	Commit current changes to a new branch and hand off to Copilot coding agent — it opens a draft pull request and works asynchronously in the background while you continue other work
`/diff`	Review the changes made in the current directory
`/mcp`	Connect GitHub's native MCP server to work with issues, branches, and pull requests from the terminal
`/agent` + `/skills`	Define custom agents and scoped tool access via `AGENTS.md` — behaviour stays consistent across models and sessions
`Shift+Tab`	Cycle between standard, plan, and autopilot mode
`/experimental [on\|off]`	Toggle experimental preview features on or off

Copilot SDK — agents in your own apps

The Copilot SDK exposes the same agent runtime as a programmatic API. You define agent behaviour; Copilot handles planning, tool invocation, file edits, and more — no orchestration layer to build yourself. Currently in Technical Preview.

Language	Package	Install
Node.js / TypeScript	`@github/copilot-sdk`	`npm install @github/copilot-sdk`
Python	`github-copilot-sdk`	`pip install github-copilot-sdk`
Go	`github.com/github/copilot-sdk/go`	`go get github.com/github/copilot-sdk/go`
.NET	`GitHub.Copilot.SDK`	`dotnet add package GitHub.Copilot.SDK`

The SDK communicates with the Copilot CLI in server mode via JSON-RPC — the CLI must be installed separately as a prerequisite. A GitHub Copilot subscription is required for standard use; BYOK (Bring Your Own Key) lets you supply your own API keys from OpenAI, Azure AI Foundry, or Anthropic without a Copilot subscription.

Choosing the right surface