Agent Upstart

Cool things people are building with agents — and practical help getting the most out of them.

Playbooks · Coding

Run Copilot as the agent in JetBrains Air

Written November 10, 2026

GitHub Copilot is now a selectable agent inside JetBrains Air, the agentic IDE JetBrains built on its earlier Fleet codebase. You pick it from Air's agent picker and run it on your own machine or in the cloud, on the Copilot plan you already pay for. The choice between local and cloud is the whole decision — and it is not a free one.

The problem

A JetBrains user who wants an agent on a task has two places the work can run: inside the IDE, on their own machine, or in a cloud sandbox. Air now offers both for Copilot, and the two paths differ in setup, billing, and what your admin can see. Picking without knowing the differences means finding them out mid-task.

This page is about choosing the agent and choosing where it runs. Following several agent sessions at once in Air is a different page: Copilot in JetBrains Air: follow agent sessions.

What shipped

GitHub's November 3 changelog makes Copilot available in JetBrains Air through the Agent Client Protocol (ACP), the open protocol that lets an editor treat an AI agent as a swappable part instead of a built-in feature. You select GitHub Copilot in Air's agent picker and use your existing Copilot subscription for coding tasks. You can track several agent sessions in parallel and review each session's changes next to your code. It sits alongside the GitHub Copilot plugin for JetBrains IDEs rather than replacing it — GitHub says to use Copilot in Air alongside the plugin and pick the workflow that fits.

Local: the short path

For local tasks, install Air in your JetBrains IDE, select GitHub Copilot, and connect your GitHub Copilot account. That is the whole setup: Air installed, Copilot picked, account connected. The work runs in your IDE, with your project context and IDE tools, and you review the edits next to the code.

Cloud: what it takes, and what it costs

Cloud tasks run through Air Teams. You need three things: a JetBrains account inside an Air Teams-enabled business organization, a GitHub Copilot subscription, and Copilot turned on by your Air administrator. You start a cloud task from the IDE, picking Cloud in the environment selector, or from a browser: open Air in the browser, connect GitHub under Settings > VCS Providers, choose a repository and remote branch, select GitHub Copilot, and send your prompt.

A cloud task runs in a Docker container on a fresh virtual machine. All traffic passes through a proxy, and no state carries over between tasks. Copilot can build, test, and install packages in that container without touching your local environment.

Cloud tasks bill against your Copilot plan, not your JetBrains AI credits. If you track spend, that is a different bucket from the one JetBrains users usually watch.

The fine print

Which to pick

Run a task locally when the work is ordinary, the machine is yours, and you want the shortest setup. Run it in the cloud when you want the task isolated from your machine — a fresh container that cannot touch your local environment — and your organization already has Air Teams set up with Copilot enabled. If the cloud path needs an admin to turn Copilot on first, that is a conversation before it is a setting. And if you juggle a personal plan and an organization plan, check which one a cloud task will bill before you run it, since Air will not let you choose.

Limits

Related playbooks

Sources

Back to: Coding playbooks hub · Also: Copilot in JetBrains Air: follow agent sessions