GitHub Copilot joins JetBrains Air, with cloud tasks on your Copilot plan
GitHub Copilot is now available as an agent in JetBrains Air, the agent workspace that JetBrains ships as a plugin for its IDEs, and from November 3, 2026 it can run tasks in Air's cloud environments as well as on the developer's own machine. GitHub announced the change in its official changelog on November 3, and Microsoft's developer team published a companion post the same day that sets out how Copilot in Air compares with the GitHub Copilot plugin that many JetBrains users already have installed.
The integration is built on the Agent Client Protocol (ACP), the open protocol JetBrains uses to plug third-party coding agents into Air. Developers pick GitHub Copilot from Air's agent picker, sign in with an existing Copilot subscription and send a prompt. From there they can run several agent sessions in parallel, give the agent access to project context and IDE tools, and review the resulting changes next to their code before accepting them.
The billing rule is the detail most teams will look for first. According to GitHub, cloud tasks that run through Air "use your Copilot plan rather than JetBrains AI credits." A company that already pays for Copilot can therefore send work to Air's cloud runners without drawing down the JetBrains AI credit balance that covers JetBrains' own models and agents. The cloud option is limited to organizations that use Air Teams, the business tier of JetBrains' agent service, and it needs an administrator to turn Copilot on.
··········
WHAT GITHUB COPILOT DOES INSIDE JETBRAINS AIR.
JetBrains describes Air in its IDEs as a single place to run several coding agents side by side. Out of the box it supports Codex, Gemini, GitHub Copilot, Claude Agent and JetBrains' own Junie, and further agents can be attached through ACP, including ones already installed on the developer's computer. Air works with IntelliJ IDEA, PyCharm, WebStorm, Rider, CLion, RustRover, GoLand, PhpStorm and RubyMine, and is distributed both as a Marketplace plugin and inside the 2026.3 EAP builds. JetBrains still labels the product as early access, and GitHub's own instructions refer to the Air plugin as an alpha.
Copilot had already been available in Air for local runs before this week. Microsoft's post states that Copilot "will also be available through Air's Cloud Run experience, expanding beyond the existing Local Run integration." The November 3 change is therefore the arrival of cloud execution, together with a formal announcement of the integration on GitHub's side.
Inside Air, Copilot behaves as a delegated agent. A developer describes a task, Copilot plans and edits files, and the session stays visible in Air's workspace so that progress can be followed and steered while it runs. Several sessions can run at the same time, which is the pattern Air was designed around, and Air supports separate Git worktrees so that parallel agents do not overwrite each other's changes. Model selection is available in Air as it is in the Copilot plugin, although neither GitHub nor Microsoft lists which models appear in Air's picker for Copilot.
Microsoft's post also covers Next Edit Suggestions, the Copilot feature that predicts the next change a developer is likely to make. Starting with JetBrains IDE version 2026.2.4, Copilot can be chosen as the Next Edit Suggestions provider under Settings, Editor, General, Code Completion, Inline, where the user selects Copilot and switches the feature on.
··········
HOW COPILOT CLOUD TASKS RUN IN AIR TEAMS.
Cloud tasks are the substantive addition. A developer can start one from the browser at air.jetbrains.cloud, after connecting GitHub under Settings and VCS Providers and choosing a repository and a remote branch, or from the IDE by selecting Cloud in the Air plugin's environment selector. In both cases GitHub Copilot is picked as the agent and the prompt is sent as usual.
GitHub describes the execution environment in some detail. Each task runs Copilot in a Docker container on a fresh virtual machine, all network traffic passes through a proxy, and no state is carried over from one task to the next. Copilot can build the project, run tests and install packages inside that environment without touching the developer's local setup. Because the task runs remotely, it keeps going after the laptop is closed, and JetBrains says Air Teams lets users follow progress from the web, with mobile access scheduled for November 15.
........
Requirement | Local tasks | Cloud tasks |
Air plugin in a JetBrains IDE | Yes (alpha plugin) | Yes, or use the browser |
GitHub Copilot subscription | Yes | Yes |
JetBrains account in an Air Teams business organization | No | Yes |
Copilot enabled by the Air administrator | No | Yes |
Where the code runs | Developer's machine | Docker container on a fresh VM |
What the task consumes | Copilot plan | Copilot plan, not JetBrains AI credits |
........
The table reflects GitHub's changelog. JetBrains' own Air page states more generally that cloud runs require a JetBrains AI subscription, so organizations should expect to need a JetBrains business arrangement for Air Teams even though the Copilot work itself is charged to the Copilot plan.
··········
GITHUB COPILOT PLUGIN VS COPILOT IN JETBRAINS AIR.
Microsoft's developer team is explicit that the GitHub Copilot plugin for JetBrains remains "the complete" Copilot experience and that Air offers lighter, built-in access. The two can be installed side by side, and the comparison below follows the table published in Microsoft's post.
........
Capability | GitHub Copilot plugin | Copilot in JetBrains Air |
Positioning | Full Copilot experience | Lightweight built-in access |
Entry point | Chat panel, inline chat, completions, code review | Air agent picker via ACP, completions |
Model choice and parallel sessions | Yes | Yes |
Next Edit Suggestions | Yes | Yes, from IDE 2026.2.4 |
Skills, hooks, instructions, debug logs | Yes | No |
Inline agentic chat | Yes | No |
Local sandbox | Yes | No |
Enterprise managed settings and policy diagnostics | Yes | No |
Code review and commit messages | Yes | No |
Cloud execution | Copilot coding agent | Air Cloud Run (Air Teams) |
Subscription | GitHub Copilot | GitHub Copilot |
........
The gaps are meaningful for teams that have invested in Copilot customization. Repository instructions, custom skills and hooks configured for the Copilot plugin do not carry over into Air, and the Agent Debug panel is absent, which makes it harder to see why an agent took a particular path. Copilot code review and commit message generation also stay in the plugin. In practice the two surfaces suit different jobs: the plugin for day-to-day work inside one project with the full set of Copilot controls, and Air for running Copilot next to Claude Agent, Codex or Junie on parallel tasks and comparing their output in one workspace.
The cloud paths differ as well. With the plugin, cloud work goes to GitHub's Copilot coding agent, which runs on GitHub's infrastructure. With Air, the same subscription drives a task on JetBrains' Cloud Run infrastructure, managed from Air's workspace.
··········
WHAT CHANGES FOR COPILOT USERS AND ADMINISTRATORS.
For an individual developer with a Copilot subscription, the change is simple. Installing the Air plugin, selecting GitHub Copilot and signing in is enough for local tasks, and no JetBrains AI plan is needed for that path. JetBrains notes that installing Air through its in-IDE notification disables the separate AI Assistant plugin, which can be switched back on under Settings and Plugins.
For companies the picture is more involved, and GitHub lists several limits that administrators should read before enabling cloud tasks:
........
Area | What GitHub says |
Plan selection | Users cannot choose per task between a personal and an organization Copilot plan |
Start-up checks | Air may check Copilot subscription or credit availability when a cloud environment starts; if unavailable, the task fails with an error |
Permissions | Connecting Copilot to Air does not request GitHub Copilot permissions |
Governance and analytics | Air's governance and analytics features do not include Copilot usage |
Admin controls | Air administrators can disable Copilot through a governance policy |
Spending limits | Air administrators cannot set Copilot spending limits inside Air |
........
The last two rows define how cost control works. Air's governance tools, which JetBrains introduced to manage agents, access and spend across an organization, can switch Copilot off but cannot cap it. Any budget for Copilot usage started from Air has to be set on the GitHub side, through the organization's Copilot plan and its AI credit settings. Teams that recently moved to usage-based Copilot billing, covered in our report on GitHub Enterprise including Copilot with no seat charge, will want to confirm how Air cloud tasks appear in that usage, since Air's own analytics will not show them.
On the JetBrains side, the Air plans page dated November 4, 2026 lists a free Basic plan, a Business plan at $20 per AI seat per month with AI policies, bring-your-own-key providers, ACP-compatible agents, MCP servers and an Analytics API, and an Enterprise plan with custom pricing. One JetBrains AI credit equals one US dollar of provider-side model spend, and bring-your-own-key requests do not consume credits. JetBrains also says some cloud agent compute may be free during the early access period, without saying which or for how long. Neither company has published a separate price for running Copilot on Air's cloud infrastructure.
Readers comparing the total cost of coding agents across vendors can find plan prices for Copilot, Cursor, Claude Code and Codex in our guide to AI coding agent pricing. GitHub's own cloud execution option is described in our article on GitHub Copilot cloud sandboxes reaching general availability.
··········
GITHUB COPILOT IN JETBRAINS AIR: CONFIRMED FACTS AND OPEN QUESTIONS.
The confirmed elements come from GitHub's changelog and Microsoft's developer post of November 3, 2026, with supporting details from JetBrains' Air documentation. Copilot is selectable in Air through ACP, works with an existing Copilot subscription, supports parallel sessions and IDE tools, and from November 3 can run tasks in Air's cloud environments in isolated Docker containers on fresh virtual machines. Those cloud tasks are charged to the Copilot plan instead of JetBrains AI credits, require an Air Teams business organization and an administrator's approval, and fall outside Air's governance analytics and spending limits.
Several points remain undocumented. GitHub has not said which Copilot models appear in Air's picker, how a cloud task in Air is metered against a Copilot plan compared with the same work done by the Copilot coding agent, or whether premium request or AI credit consumption differs between the two paths. JetBrains has not published how long cloud compute stays free during early access. The feature set is also explicitly narrower than the Copilot plugin's, with no support for custom skills, hooks, instructions, code review or enterprise managed settings, and neither company has given a timeline for closing those gaps.
For JetBrains teams that already pay for Copilot, the practical result is a second cloud option for delegated tasks that keeps Copilot charges on the GitHub bill, at the cost of fewer controls than the plugin and of budget enforcement that still has to happen on GitHub.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org
