GitHub launches Copilot desktop app for coding agents

GitHub has launched a technical preview of a standalone Copilot desktop app, expanding its AI coding assistant beyond the editor and terminal. The Microsoft-owned company said the new app is designed to manage coding agents, issues, pull requests, and development sessions from a single interface.

The app lets developers start Copilot tasks from GitHub issues, prompts, or existing code sessions. According to GitHub, it also tracks progress across repositories and active agent runs, giving users one place to follow work that may otherwise be spread across editors, terminals, and browser tabs.

GitHub said the desktop app includes a unified inbox for issues and pull requests, side-by-side diff reviews, session history, repository context, and support for running multiple coding agents at the same time. Developers can inspect proposed changes, leave feedback, resume paused sessions, and move completed work into pull requests.

The app is built on GitHub Copilot CLI, the company’s terminal-based AI coding agent that reached general availability in February. The desktop client adds a graphical interface on top of those agent capabilities, letting developers supervise coding sessions and tasks without switching between tools as often.

The Copilot app is available for macOS, Windows, and Linux. GitHub said it is currently in public preview for Copilot Business and Enterprise subscribers, while Copilot Pro and Pro+ users can join a waitlist for early access. The company has not formally announced a full public launch date, though the announcement video references June 2, which may point to a wider rollout.

The release extends Copilot beyond the developer tools where it first appeared. Since launching in 2021, Copilot has mainly lived inside Visual Studio Code, JetBrains IDEs, and Visual Studio, with later expansion to GitHub.com, mobile apps, and Copilot CLI. The original Copilot experience focused on inline suggestions and chat help inside the editor, where developers wrote code locally while the assistant suggested completions or edits.

GitHub’s new app reflects a shift in the wider AI coding market toward autonomous coding agents that can operate across repositories, tasks, and cloud environments. That puts GitHub in more direct competition with Claude Code from Anthropic and OpenAI’s Codex, both of which have attracted interest by handling larger chunks of engineering work for developers.

GitHub has an obvious advantage in that much of the surrounding software development workflow already lives on its platform. Repositories, issues, pull requests, CI pipelines, and code review tools are already part of GitHub’s ecosystem, giving the company a way to connect coding agents more tightly to the rest of the development process.

Petter Arnesen, an Azure MVP and cloud architect who had early access to the app, described GitHub’s approach in a LinkedIn post as “probably the most interesting implementation” of an AI developer assistant he had tried. He said he used it for side projects and for agent-driven pull request review loops where Copilot could wait for feedback, address comments, and update PRs automatically.

Arnesen also said he would not yet “unleash this on production systems without supervision,” citing bugs during the preview period and a tendency for AI agents to produce overly complicated solutions without human oversight.

The launch also comes after GitHub made major changes to Copilot’s product and pricing model. In April, GitHub paused new sign-ups for some individual Copilot plans and introduced tighter usage limits for existing subscribers. Shortly after that, the company announced a move toward usage-based billing tied to tokens consumed by different AI models, including input tokens, generated output, and cached context.

GitHub also added more of the underlying agent infrastructure around Copilot in recent months. On Wednesday, the company introduced a REST API for launching cloud-based Copilot agent tasks, along with unified session views inside JetBrains IDEs. The desktop app brings those pieces together in a single product surface as GitHub pushes Copilot further toward agent-driven software development.