GitHub Copilot joins JetBrains Air for local and cloud coding tasks

Expanding the Agent Workflow

JetBrains is expanding the capabilities of its Air platform by integrating GitHub Copilot as a primary coding agent. Air is designed to let developers hand off specific coding tasks to AI agents so they can run in parallel. This update allows users to manage these tasks and review the final changes in a single unified interface.

Users can now select GitHub Copilot from the agent picker alongside other models like Claude, Codex, and Junie. The integration is built to work with both local environments and the cloud-based Air Teams service. This means developers can choose where the heavy lifting happens based on the complexity of the task at hand.

The core functionality remains consistent with the JetBrains Air experience. You describe a task, the agent performs the work, and you review the results in the task list. By adding Copilot, JetBrains is providing a familiar tool for developers who already rely on GitHub's ecosystem for their daily coding needs.

How the Integration Works

The technical backbone of this integration is the Agent Client Protocol (ACP). This is an open protocol that JetBrains Air uses to connect to various AI agents. By running the official GitHub Copilot CLI over this protocol, JetBrains ensures that the agent behaves as intended by GitHub.

When you run a task locally, it operates directly within your JetBrains IDE. You simply need to sign in with your GitHub account when prompted by Air. The agent then works on your local project files, and you can monitor its progress and review its output within the Air interface.

For cloud-based tasks, the process involves a remote environment. When a task starts in the cloud, it initializes in a fresh virtual machine. The system uses a proxy to handle all traffic, ensuring that nothing carries over from one task to the next. This sandboxing prevents the agent from interfering with your local machine's resources or files.

Cloud Execution and Sandboxing

One of the primary benefits of using GitHub Copilot with Air Teams is the ability to run tasks remotely. You can start a cloud task from your IDE or the web and then close your laptop. The agent will continue to work on a remote branch using its own dedicated virtual machine.

Each cloud task runs inside a Docker container, which is a standard way to package software so it runs reliably in different environments. Because it is sandboxed, the agent can build, test, and install various packages without affecting your local setup. This is particularly useful for tasks that require significant compute power or complex dependency installations.

To use this feature, you must have a JetBrains account with a company that uses Air Teams and a valid GitHub Copilot subscription. You must also push your changes to a remote branch first, as cloud agents work from those remote locations rather than your local files.

Subscription and Credit Management

A notable detail of this integration is how it handles billing and usage. Tasks performed through JetBrains Air run on your existing GitHub Copilot plan. This means you do not use JetBrains AI credits to power these specific tasks; instead, they count against your GitHub subscription limits.

Because JetBrains Air does not have direct visibility into your GitHub account's remaining balance, it cannot display your current credit usage. The platform will check that you have an active subscription before starting a task, but it won't warn you if you are running low on credits. If your credits are exhausted, the tasks will simply begin to fail.

Developers are encouraged to monitor their usage directly on the GitHub website. This separation of concerns ensures that your Copilot plan remains the primary source of truth for your AI usage, regardless of whether the agent is running locally or in the cloud.

Automations and Next Steps

The integration also supports automated workflows for repetitive maintenance. You can set up automations to start GitHub Copilot tasks on a specific schedule or when certain events occur in your repository. For example, you can trigger a task every time a new pull request is opened.

JetBrains recommends starting with simple, well-defined tasks to get comfortable with the system. A good first project might be a weekly dependency update or fixing a flaky test. These types of tasks are easy to review and help keep your credit usage predictable as you begin using the tool.

While automated pull request reviews are possible, they can drain credits much faster if the repository is very active. By starting with smaller tasks, developers can see how the agent handles different types of work before moving to high-frequency automations. You can try this today by installing Air in your IDE and selecting GitHub Copilot from the agent picker.