GitHub Copilot Is Now Selectable in JetBrains Air for Local and Cloud Tasks on Copilot Plans; Air Neither Shows Copilot Usage Nor Caps Spending
GitHub and JetBrains say Copilot can run in Air locally or in cloud sandboxes billed to Copilot plans, but Air does not track Copilot credits, show its usage or set spending limits.
Editor's Note ·
- Clarification:
- The "What We Don't Know" section says that "both vendors direct users to GitHub for usage." Only JetBrains does so (its FAQ says "Check your usage on GitHub"); the GitHub changelog does not direct users to check usage on GitHub.
Overview
GitHub says developers can now use GitHub Copilot inside JetBrains Air. According to the GitHub changelog, “you can now delegate coding tasks, guide execution, and review results with GitHub Copilot in JetBrains Air,” using the Agent Client Protocol (ACP) for tasks “in your JetBrains IDE or in the cloud.” JetBrains published a companion post, Use Your GitHub Copilot Plan in JetBrains Air, saying the tasks run on an existing Copilot plan “not on JetBrains AI credits.” Both posts also list limits on what Air can see and control about Copilot usage.
What We Know
Local and cloud tasks
GitHub’s changelog lists three additions: a GitHub Copilot entry in Air’s agent picker that uses an existing Copilot subscription; parallel sessions with in-context review; and cloud tasks that run “through Air Teams, from your IDE or browser.” JetBrains says Air runs the official GitHub Copilot CLI over ACP, and that Copilot appears in the same task list as Claude, Codex and Junie. Local use does not need Air Teams, but a Copilot subscription is required.
For cloud tasks, GitHub lists three requirements: a JetBrains account in an Air Teams-enabled business organization, a GitHub Copilot subscription, and Copilot enabled by the Air administrator. JetBrains adds that cloud agents work from a remote branch, so changes must be pushed first, and that Air uses the GitHub connection under Settings | VCS Providers to sign in to Copilot when a task starts.
Where cloud tasks run
The GitHub changelog says that when a Copilot cloud task runs through Air, “Copilot works in a Docker container on a fresh virtual machine. All traffic passes through a proxy, and no state carries over between tasks.” JetBrains’ separate post on cloud agent sandboxing in Air describes the design: each task starts a VM that is “not reused between tasks,” with the agent in a Docker container behind a proxy. By default the proxy allows connections only to a preset list of trusted source control and package registry domains, such as npm; JetBrains says users can add domains, allow all domains, or turn internet access off. The same post says microVMs are “coming soon” and that cloud agents are currently available to organizational users, with a broader release for individual users planned later.
Automations
JetBrains says Air Automations can start cloud tasks on a schedule or when something happens in a repository, such as a new pull request, and that with Copilot selected, “each run uses your Copilot plan.” The post recommends a weekly dependency update as a first automation and warns that one reviewing every pull request “can drain your credits much faster, especially if the repository is busy.”
Usage and Spending Visibility
Both vendors state limits on usage tracking. Under “Cloud task considerations,” GitHub says Air’s governance and analytics features do not include Copilot usage, that Air administrators can disable Copilot through a governance policy but cannot set Copilot spending limits within Air, and that users cannot choose per task between a personal and an organization Copilot plan. It also says Air may only verify subscription or credit availability when the cloud environment starts, and that the task fails with an error in the session if they are unavailable.
JetBrains makes the same point in its FAQ: “Air doesn’t track Copilot credits for local or cloud tasks, and it can’t set a spending limit.” It adds that Air “doesn’t warn you when your credits are all used up” and that runs will “just start failing,” so usage has to be checked on GitHub.
Background
Air launched in public preview in March. The Register reported on March 10 that Air supported OpenAI Codex, Anthropic Claude Agent, Google Gemini CLI and JetBrains Junie, and that it could use ACP, which Zed and JetBrains sponsor. The outlet also described tasks that could run in a local workspace, a Git worktree, Docker or, in a future release, a cloud container. The Herald has previously reported on JetBrains opening early access for Air inside its IDEs, and, in an Analysis, on sandboxes and permission modes versus repository-borne attacks.
What We Don’t Know
- None of the cited sources gives a Copilot pricing or credit-consumption figure for Air tasks; both vendors direct users to GitHub for usage.
- The Register’s report predates the Copilot integration and does not address it, and the other three cited sources are from GitHub or JetBrains, so the claims about the integration rest on the vendors’ own descriptions.
- Neither vendor says whether Air will add Copilot usage to its governance and analytics features or add Copilot spending limits.