Copilot App
The GitHub Copilot desktop app: parallel agent sessions with GitHub integration
title: "Copilot App" description: "The GitHub Copilot desktop app: parallel agent sessions with GitHub integration" section: "Copilot" readTime: "5 min" badge: "NEW"
GitHub Copilot App
The GitHub Copilot app is a standalone desktop application purpose-built for agent-driven development. It combines parallel agent workstreams, GitHub integration, and pull request lifecycle management, and it is built on GitHub Copilot CLI. It runs on macOS, Linux, and Windows.
The Copilot app is in technical preview and subject to change. It is available on all Copilot plans. For Business and Enterprise, the GitHub Copilot app policy must remain enabled (it is on by default and separate from the Copilot CLI policy). Content exclusion policies apply to Business and Enterprise users.
Isolated Sessions
You can run multiple agent sessions at once. Each session has its own git worktree and branch, so tasks do not interfere with each other. A session can run in a cloud sandbox, or you can configure a local sandbox to limit the resources agent-run tools can use on your machine.
Session modes:
- Interactive — collaborative, you steer as you go
- Plan — the agent plans and you approve before it executes
- Autopilot — fully autonomous
You can choose the model and reasoning effort for each session.
GitHub Integration
- Browse and find issues and start sessions from them
- Create and close pull requests and review pull requests
- View CI check results
- Search across repositories
More Features
- Automations — save recurring agent tasks and run them on a schedule or on demand
- Chats — brainstorm in a conversation mode without creating a branch or workspace
- Session history —
/chroniclegives insights from previous sessions, and/chronicle cost tipssurfaces expensive usage patterns - Canvases — custom, agent-driven artifacts where people and agents collaborate
- Computer use — on macOS and Windows, local sessions can interact with desktop applications that have no API, CLI, or MCP integration
- Customizations — global instructions, MCP servers, and agent skills
- Models — multiple LLMs, including your own provider through bring your own key (BYOK)
The app may generate code that matches or nearly matches public code, even when the "Suggestions matching public code" policy is set to Block. It also consumes AI credits, so match model capability to task complexity.
Copilot App vs VS Code
| Copilot App | VS Code | |
|---|---|---|
| Start from | GitHub issues, PRs, repositories | Files, selection, chat prompt |
| Session isolation | Own worktree + branch per session | Shared workspace (local sessions) |
| Built on | Copilot CLI | Editor, terminal, debugger, tests |
| Best for | GitHub-centric workflows, parallel async tasks | Active coding, file editing |
VS Code also offers an Agents window for managing sessions across projects, and Copilot CLI sessions can be steered remotely, see Agents.
Getting Started
- Download the app from the GitHub Copilot app repository (see the official docs)
- Sign in with your GitHub account
- Start your first session from an issue or an empty chat
- Use the Give feedback icon at the bottom left to report issues
Related
- Cloud Agent — Background agent that returns pull requests
- Agents — All Copilot agent types
- Copilot Memory — Preferences that persist across sessions