JetBrains today is launching its Air agentic development environment into public preview. Air was created to enable developers to build tools around agents, guide them and fine-tune their outputs.
According to the company, Air is designed to end the current state of agent fragmentation, in which each agent runs in a different tool, with different setup, context and a lack of structural understanding of the code. It also eases the switching of agents across projects, where they are not being migrated but instead are a part of the workflow.
With Air, developers can define tasks by mentioning “a specific line, commit, class, method, or other symbol,” JetBrains wrote in its blog announcing Air. “As a result, the agent gets precise context instead of a blob of pasted text.” When the task is done, Air enables users to see changes not only at the code diff, but in the context of the codebase. Tools such as terminal, Git client and built-in preview also are enabled.
Air supports the Agent Client Protocol (ACP), an open standard that lets AI coding agents integrate with any code editor or IDE, and supports Codex, Claude Agent, Gemini and Junie. JetBrainssaid it will add support for other agents through the ACP Agent Registry. Users will be able to run agents locally, or isolate them in containers and Git worktrees for sandboxing and concurrent work, the company wrote.
Run agents locally by default, or isolate them in Docker containers and Git worktrees for sandboxing and concurrent work.
Air helps you to avoid the mess of having multiple windows and terminal tabs open for each task. You see one task (meaning one agent session) at a time. You’ll get a notification when another task needs your attention, so you can quickly switch to it while the agent keeps working. Air then helps bring your changes from a container or worktree to your main copy.
