GitHub Copilot App: The new desktop platform for AI coding agents

GitHub Copilot: the new desktop app for AI coding agents

GitHub has launched the Copilot app, a desktop platform dedicated to managing AI coding agents, presented during the Microsoft Build 2026. This novelty extends the Copilot ecosystem beyond editorial integrations and command-line tools, offering a specific work environment to orchestrate multiple agents simultaneously.

Quick Answer

A control center for parallel sessions

The app provides a unified view called "My Work", which shows active sessions, open issues, pull requests, and background automations across connected repositories. Each session runs in its own git worktree, an isolated copy of the branch, to avoid interference between agent activities. Worktree management is fully automated, eliminating the need for manual configurations or cleanups.

Agent Merge: advanced pull request automation

The Agent Merge feature accompanies a pull request through all stages of the review and integration process. It monitors continuous integration checks, tracks required reviewers, addresses failed checks, and waits for merge conditions to be met. Developers can configure which operations the agent is authorized to perform, including resolving continuous integration issues, addressing reviewer feedback, and executing the merge.

Copilot Max: the advanced plan for intensive users

For more demanding users, GitHub has introduced Copilot Max, a premium subscription plan that offers advanced features for managing multiple agents. This plan is available as an upgrade from Copilot Pro, Pro+, and EDU plans.

Canvases: shared work surfaces

The app introduces canvases, shared work surfaces between humans and agents. A canvas can display plans, pull requests, browsing sessions, terminals, distributions, dashboards, or workflow states. Agents update the canvas as work progresses, while developers can modify, reorganize, approve, or redirect activities directly on the same surface. The chat interface handles instructions and discussions, while canvases contain the actual work for inspection and verification.

Code review: updates and new features

The Copilot code review system has introduced an intermediate level that directs pull requests to a higher reasoning model for greater precision and recall. Administrators can set the review effort level for individual repositories to "low" or "medium", matching the model's strength to the repository's risk and impact.

Two new features accompany this update: the /security-review skill, which provides a dedicated path for security-focused assessments, and the /rubberduck skill, now generally available, which uses multiple model families to critique implementations and detect new issues.

Copilot code review is now natively available on Azure DevOps as well, with the same one-click review features, inline comments, and committable fix suggestions provided on GitHub. Administrators can enable it on selected repositories.

SDK, CLI, and partner integrations

The GitHub Copilot SDK has reached general availability in Node.js/TypeScript, Python, Go, .NET, Rust, and Java. The SDK exposes the same agent-based runtime that powers the Copilot app, providing teams with a foundation to build internal tools such as code analysis utilities, release note generators, and agents integrated into support workflows.

The Copilot CLI has received a new terminal user interface in /experimental mode, with access to tabs for pull requests, issues, and gists. A voice mode uses local speech-to-text recognition, keeping audio on the local machine. The /every command schedules recurring prompts and background tasks.

Cloud automations allow agents to perform operations on schedules, respond to GitHub events, open issues, and post comments. The cloud agent requires permission before each write action by default, with an autopilot mode available once trust is established between teams. The Copilot cloud agent can present issues, start discussions, and respond to reviewers.

Memory++ and the /chronicle command

Memory++ and the /chronicle command provide continuity across devices and over time, allowing you to query the context of sessions started in the Copilot app, CLI, VS Code, or GitHub at later times.

Partner integrations for extended automation

Agent apps built by partners integrate with GitHub Copilot to help automate tasks, generate code, analyze context, and perform actions. You can use your favorite tools without leaving GitHub, assigning issues to new agents that fit your workflow. Among the partners mentioned are LaunchDarkly, Bright, Amplitude, Sonar, Endor Labs, Octopus Deploy, Packfiles, PagerDuty, and Amplitude.