GitHub Copilot comes to JetBrains Air as a selectable agent
GitHub Copilot can now be picked from the agent picker inside JetBrains Air, running local and cloud tasks on an existing Copilot subscription. The integration is built on the Agent Client Protocol (ACP), and cloud tasks execute in a Docker container on a fresh virtual machine. There are gaps worth knowing about plan selection, analytics and spending controls.
Key points
- Copilot is selectable from Air's agent picker and reuses your existing Copilot subscription.
- The integration is built on the Agent Client Protocol (ACP) and runs alongside the existing GitHub Copilot plugin.
- Cloud tasks execute in a Docker container on a fresh VM with all traffic through a proxy; nothing persists between tasks.
- Cloud tasks are billed to your Copilot plan rather than JetBrains AI credits.
- Air cannot choose between your personal and organization Copilot plan for an individual task.
- Air admins can disable Copilot through a governance policy, but cannot set Copilot spending limits inside Air.
GitHub Copilot can now be selected as an agent inside JetBrains Air, JetBrains' agent-first development environment, connected through the Agent Client Protocol (ACP). Install the Air plugin in a JetBrains IDE and link a Copilot account and Copilot appears in Air's agent picker, running tasks locally while you track several sessions and review changes next to your code.
Cloud tasks need an Air Teams-enabled business organization, a Copilot subscription and an administrator who has allowed Copilot; they run in a Docker container on a fresh virtual machine behind a proxy, with no state carried between runs, and are billed to your Copilot plan instead of JetBrains AI credits. Three limits matter: Air cannot pick your personal or organization plan per task, its analytics exclude Copilot usage, and administrators cannot set Copilot spending limits inside Air.
Real-world use
- A developer working in IntelliJ or WebStorm who wants several agent sessions running in parallel.
- A team that already pays for Copilot Enterprise and wants cloud agent tasks without buying JetBrains AI credits.
- A project that needs builds, tests and package installs run in an isolated sandbox instead of a local machine.
- A team comparing Copilot with Air's other agents on the same repo to see which workflow produces better results.
Why it matters
Your Copilot subscription now stretches into JetBrains' agent environment, so the same quota covers two interfaces instead of a second plan. Factor in the missing per-task plan choice and spending controls before standardizing on it.