JetBrains Air adds support for more agents, local AI models, Docker on Windows, and more

Air, the agentic development environment from JetBrains, introduces major updates to expand flexibility and language support. Developers can now connect external coding agents through the Agent Client Protocol (ACP), opening integration with agents such as GitHub Copilot CLI, OpenCode, Pi, Cline, and more. This change allows users to manage multiple agents seamlessly within the same review and task workflow.

Alongside agent integration, Air now supports running local language models through popular runners like Ollama and LM Studio. This option lets developers harness models on their own machines, enhancing privacy and configurability.

Following these core changes, Air has also added Java and Kotlin language intelligence, enabling mixed-language project support using the IntelliJ IDEA code engine. This addition is currently in beta.

While macOS already supported Docker-based agent tasks, Windows users now gain comparable functionality. Agent tasks can run in Docker containers, providing isolation for dependencies and agent commands without touching the host environment. Docker Desktop is required to enable this feature.

Complementing these technical enhancements, Air has refined the user experience with features such as reviewing proposed agent changes before accepting them, viewing context usage for Claude Code, adjusting task effort levels, and keeping your Mac awake during extended operations.