A standalone application released in March 2026 by JetBrains, the company behind IntelliJ IDEA and PyCharm. JetBrains calls it an Agentic Development Environment. Its defining feature is the ability to run four different AI coding agents — OpenAI’s Codex, Anthropic’s Claude Agent, Google’s Gemini CLI, and JetBrains’ own Junie — side by side in a single window. Each agent runs as an independent task loop that does not interfere with the others, with work isolated through Git worktrees or Docker containers. Rather than stopping at “hand the task off to an agent,” Air is designed so that defining tasks, monitoring execution, and reviewing the resulting code all happen in the same place. It is currently offered as a public preview.
Key Features
- Four agents running in parallel: Codex, Claude Agent, Gemini CLI, and Junie can all run at once inside the same workspace. You do not have to wait for one agent to finish before handing another task to a different agent
- Per-task isolation: Each task is isolated in a Git worktree or a Docker container, structurally preventing the failure mode where several agents rewrite the same files and break each other’s work. Cloud execution environments have also been announced
- Write tasks with code awareness: Instead of a plain chat box, you compose instructions while referencing symbols and files directly, so “which class, which method” is never ambiguous
- Review in the same window: Diffs produced by agents can be inspected with language-aware code navigation. Being able to jump to definitions and find references while reviewing is very much the mark of an IDE vendor
- Workspace-level task management: You can see all in-flight tasks at a glance and keep several streams of work moving at once
- Not a replacement for your IDE: Air is positioned as a separate app for delegation and review, not as a substitute for IntelliJ IDEA and friends. Hand-written work continues in your existing IDE
Pricing
| Plan | Price | What it covers |
|---|---|---|
| Air itself (public preview) | Free | The Air app is free to use during the preview period |
| JetBrains AI Pro / Ultimate | See official site | Subscribers can use all four agents at no additional cost |
| BYOK (bring your own key) | Paid directly to each provider | Supply your own Anthropic, OpenAI, or Google subscription or API key and pay those providers at their standard rates |
Pricing is as of August 2026. The structure may change once the preview period ends, so check the official site and the JetBrains AI pricing page for the latest details.
Pros & Cons
✅ Pros
- Running multiple agents side by side cuts down waiting time itself
- Isolation via worktrees and containers is built in, so you do not have to assemble the riskiest part of parallel execution yourself
- Review is language-aware, so checking generated diffs is more than reading text side by side
- You are not locked to a single AI vendor — Anthropic, OpenAI, Google, and JetBrains models can be used according to the job
- No need to abandon your existing IDE; you can add just the delegation layer and try it
⚠️ Cons
- As a public preview, it may involve specification changes and bugs
- Using the agents requires either a JetBrains AI subscription or your own provider keys, so the real cost is not zero
- Some features assume Docker and Git worktrees, which raises the barrier for anyone unfamiliar with them
- Running several agents at once consumes a fair amount of machine resources
- The pricing model after the preview period ends is not yet settled
Comparison with Similar Services
| Criteria | JetBrains Air | Cursor | Claude Code | JetBrains IDE + Junie |
|---|---|---|---|---|
| Form factor | Standalone agent environment | AI-integrated editor | Terminal-based agent | Agent built into the IDE |
| Parallel agents | Four families at once | Mostly a single conversation | Single session focused | Single |
| Task isolation | worktree / Docker built in | Limited | Set up yourself | Limited |
| Model choice | Codex / Claude / Gemini / Junie | Multiple models | Anthropic models | JetBrains-provided models |
| Typical use | Delegating and reviewing parallel tasks | Hand-writing alongside AI | Delegating from the terminal | Staying inside a familiar IDE |
Who Is It For
- Developers juggling several changes at once who feel the queue of waiting AI tasks is wasted time
- Anyone who wants to pick between Codex, Claude, and Gemini per job but finds switching tools tedious
- Quality-conscious teams that want to review agent output properly before merging
- People already using IntelliJ IDEA or PyCharm who prefer to stay within the same vendor’s ecosystem
- Anyone who would rather not build their own isolation setup for parallel execution
Summary
JetBrains Air moves one step beyond “let an AI agent handle it” toward “run several agents at once and check the results properly.” The isolation through worktrees and Docker, and the language-aware review, are where JetBrains’ long experience building IDEs shows. It is still a public preview with an unsettled pricing model, so committing to it in production is premature — but for developers who already delegate work to AI daily, it is worth trying while the preview remains free.