GitHub Copilot in JetBrains Air: Cloud Tasks, Billing and Limits

GitHub's November 3 announcement brings Copilot cloud tasks to JetBrains Air, using an existing Copilot subscription. For developers already paying for Copilot, that creates a way to delegate work through Air without routing those tasks through JetBrains AI credits. For team leads, it also creates a billing boundary worth understanding before enabling unattended work.

The GitHub release notes identify several limits: Air does not include Copilot usage in its governance analytics, cannot impose Copilot spending caps, and offers no per-task selector between a personal and organizational Copilot plan. An administrator can disable Copilot through a governance policy. Connecting an account also does not guarantee a successful run: subscription or credit checks may happen when the cloud environment starts.

Air and the Copilot plugin serve different workflows

Copilot was already available locally through Air. Microsoft's comparison of Air and the Copilot plugin identifies November 3 as the expansion to Cloud Run. Air connects to Copilot through the Agent Client Protocol, or ACP, and provides a shared workspace for delegating tasks and reviewing agent output.

The existing GitHub Copilot plugin remains available. Microsoft recommends it for the fuller Copilot experience, including inline agentic chat, Copilot code review and enterprise managed settings. Those features should not be assumed to carry over just because Copilot appears in Air's agent picker.

For a team evaluating the change, list the features your developers actually use before switching their default workflow. Losing a familiar review or policy feature can cost more time than a new task interface saves.

Local setup is simpler than cloud setup

For local work, install Air in your JetBrains IDE, choose Copilot and sign in. Air Teams is not required. Cloud work needs a JetBrains account in an Air Teams-enabled business organization, a Copilot subscription and administrator enablement.

JetBrains' setup walkthrough explains that cloud tasks start from a repository's remote branch. Push the changes the agent needs before launching a task. In the browser, connect GitHub under Settings, then VCS Providers. From the IDE, select Cloud as the environment. The walkthrough also directs users to GitHub for Copilot usage information.

That remote-branch requirement matters when reproducing bugs. A prompt referring to an uncommitted local fix can send the agent looking at different code from the code on your screen. Include the branch and a reproducible test command in the handoff.

What cloud isolation does, and what to check yourself

According to JetBrains' sandboxing explanation, each task runs inside a Docker container on a fresh virtual machine. Network traffic passes through a proxy. The default configuration permits selected source-control and package-registry domains, with configurable network access. Tasks retain the developer's repository access boundaries.

My recommendation is to pilot this with a small test repair on a non-critical branch. Check whether dependencies install under the network policy, inspect the diff and test output, then reconcile usage on GitHub. Before adding recurring tasks, assign someone to watch that usage separately from Air's dashboards. Cloud isolation and spending oversight solve different problems; a successful run only establishes that the first workflow worked.

I'm Paras. I write about AI tools and the decisions developers and founders face when adopting them.

I help technical teams turn product details into clear writing, useful stories and founder-led distribution.

Book a 30-minute call.