1. What the announcement establishes
The GitHub changelog says developers can select Copilot in Air’s agent picker, delegate coding tasks, guide execution, and review changes alongside their code. The integration uses Agent Client Protocol, or ACP. GitHub also describes support for multiple agent sessions, project context, and IDE tools.
The timing needs a distinction. Microsoft’s companion explanation describes Local Run as an existing integration and says Cloud Run becomes available starting November 3. The release therefore expands where developers can run Copilot tasks through Air; it should not be read as the first appearance of every local capability.
These are vendor descriptions of availability and behavior. We have not executed tasks through either route, and the announcement supplies no evidence that parallel sessions improve code quality or shorten a team’s review cycle.
2. Local access and cloud access have different requirements
For local tasks, GitHub directs developers to install Air in a JetBrains IDE, select Copilot, and connect their Copilot account. Cloud tasks have additional requirements: a JetBrains account belonging to an Air Teams-enabled business organization, a GitHub Copilot subscription, and Copilot enabled by the Air administrator.
Eligible users can start cloud tasks from a browser or select Cloud in the Air plugin’s environment selector. JetBrains’ Air product page also describes cloud availability as a gradual rollout. A solo developer should therefore verify actual account access before designing a workflow around browser-based delegation.
Our recommendation is to separate the decisions. Evaluate local delegation if that route is available and useful. Evaluate cloud execution only after confirming the organization requirements. A successful local session would not prove that the same account can launch a cloud task.
3. Air does not replace every Copilot plugin feature
Microsoft describes Copilot in Air as an integrated agent workspace for delegation, progress tracking, and review. Its comparison lists agentic coding, model selection, parallel sessions, and IDE tools for both Air and the separate Copilot plugin.
The same comparison lists differences that matter to an established workflow. Inline agentic chat, Copilot Code Review, commit message generation, and advanced agent customization are listed for the plugin but not included in the Air experience. GitHub explicitly says developers can use Copilot in Air alongside the plugin.
Before changing the team’s default interface, list the features people actually use during an ordinary task. If inline chat or built-in code review is central to that process, retain a route to those capabilities. Our coding-assistant evaluation guide can help frame that decision around concrete tasks instead of feature counts.
4. Fresh cloud tasks make reproducibility important
GitHub says each Copilot cloud task runs in a Docker container on a fresh virtual machine, with no state carried between tasks. JetBrains’ cloud execution explanation independently describes the same container-within-a-VM arrangement and says the VM is not reused between tasks. These remain vendor statements, not observations from our own execution tests.
For a small team, the operational implication is that yesterday’s successful setup should not be treated as proof that today’s task has everything it needs. Our proposed pilot would require the agent to identify the repository revision, preparation commands, checks attempted, and unresolved failures in its handoff.
Choose a task with a clear acceptance condition, such as a narrow bug fix with an existing reproduction. Ask a reviewer to reproduce the relevant check from the resulting revision. The regression-testing guide provides a useful starting point for deciding what evidence should accompany the change.
5. Air analytics leave a Copilot visibility gap
The announcement explicitly says Air’s governance and analytics features do not include Copilot usage. It also says users cannot select whether an individual task uses their personal Copilot plan or their organization’s plan. Teams should not assume that Air’s broader dashboard descriptions apply to this integration.
Another limitation appears at task startup: GitHub says Air may verify subscription or credit availability only when the cloud environment starts. A connected account can therefore still encounter a failed task and an error in the session. Account connection alone is insufficient evidence of execution readiness.
For an initial evaluation, we propose a simple task log recording successful starts, startup failures, completed changes, rejected changes, and reviewer time. That log would help assess the workflow; it would not replace missing product usage analytics or establish which plan an individual task used.
6. Expand parallel work only when review keeps up
Start with one bounded task and one named reviewer. Record the time spent clarifying the request, inspecting the diff, reproducing checks, and correcting the result. These are proposed measurements, not outcomes demonstrated by this release.
Only then try two independent tasks in parallel. Avoid using tightly coupled changes for the first comparison: that would mix agent coordination problems with the question of whether Air makes task review easier. Our AI code review workflow offers acceptance criteria that can remain consistent across both trials.
Air is worth evaluating when your team already delegates coding work and wants a common place to follow sessions and inspect changes. Keep the pilot narrow if cloud access is unavailable, required plugin features are missing, or reviewers accumulate unfinished work. Expand when repeated tasks produce understandable diffs and reproducible checks without an expanding review backlog.
Sources & further reading
Primary references used for product and technical information. Features and terms can change; check the provider before committing.
- GitHub Changelog: GitHub Copilot is now available in JetBrains AirAccessed 2026-11-04
- Microsoft Developer Blogs: GitHub Copilot in JetBrains AirAccessed 2026-11-04
- JetBrains AirAccessed 2026-11-04
- JetBrains: How Cloud-Based Agent Sandboxing Works in AirAccessed 2026-11-04